THERMO-ELECTRIC DIAGRAM FROM - 200° C. TO 100° C. 769 



focus and the directrix. As a fact, this property was the earliest one employed in 

 attempting to determine the positions of these curves, and the present A-curve was 

 thus drawn and tested in 1898. 



The equation (60) is, however, to be preferred to equation (59), both because it was 

 directly calculated from the observations and because it gives a smaller probable error 

 than (59). In the most convenient form for calculation equation (60) becomes 



E= -15868-074 + OT6O13760 2 .... (63) 



where 



= ^ + 314-0942 ...... (64) 



Platinum (No. 1). 



It may be convenient to consider the other platinum observations here, for the 

 sake of comparison with Callendar's platinum, although the No. 1 observations were 

 not examined till last year. 



The A-curve on fig. 15 was drawn with the same precautions as for the other 

 metals, from the whole thirty recorded observations. . The working co-ordinates were 



T = ? + 200, H = E/100 + 310. 



The raid-points of twelve chords of this curve parallel to the axis of t were noted, 

 and plotted on the diagram (above the short line a, to which they are parallel) ; 

 they were 



H= 10- , 60- , 110- , 160- , 210- , 260- , 310/ , 320* 330-* 340* 350-* 360- 

 T= 156-4, 154-3, 1527, 151-1, 149-3, 147-4, 146-0, 1455, 144-8, 144-5, 144-2, 144-5. 



These lay so nearly in a straight line that it was permissible to take the curve as a 



parabola. To save labour in calculating the locus of these points, only eight of them 



were employed, the four with the asterisk being omitted. The calculation by least 



squares gave the equation 



H + 29-535T = 4916-877 ..... (65) 



leading to these errors in the values of H, 



+ -78, - -43, - -15, + -12, - -20, - -81, + -06, + -63, 



whose sum is zero, thus verifying the calculation ; and producing a probable error 

 of zt'35. From this we get the cotangent of the angle of slope of the axis to be 

 29-535 (an angle of about 1° 56'). 



Following the same process as for Callendar's platinum, the equation connecting 

 E and t, in its most convenient form, is 



E= -1,303,998-87 -2953-50+124,357-O63V0 ) 

 where = ^ + 499-02596 j . 



t being degrees centigrade from the freezing-point, and E measured in C.G-.S. absolute 



units. 



The sum of the errors was 6"87 — 7'05, which, being practically zero, verifies the 

 TRANS. ROY. SOC. EDIN., VOL. XLVII. PART IV. (NO. 25). 113 



