THERMO-ELECTRIC DIAGRAM FROM -200° C. TO 100° C. 785 



In fig. 22 I have plotted these two sets of observations for Pt-Fe, and Ag-Fe 

 as ordinates to temperatures as abscissae. In the case of Pt-Fe, I sheared the curve 

 directly from the " plot " as shown, at the same time multiplying by five ; the result is 

 the sheared curve. A nearly vertical row of fourteen black dots gives the mid-points 

 of as many horizontal chords. These showed a slight slant. I therefore drew carefully 

 a set of parallel chords of the original curve. Parallelism was obtained thus : — Pins 

 were driven into the paper at A and B, a ruler laid against them, and the chord drawn ; 

 each pin was moved up 2 mm. and another chord similarly drawn ; and so on. Seven 

 such chords, with their mid-points, are shown ; the straight line extending up and down 

 from them shows very clearly that these observations of Professor Tait give a slant 

 parabola (or at least conic). The ten observations are marked on the curve, and lie on 

 it remarkably closely. 



With the same method I plotted the Ag-Fe observations (above the Pt-Fe), and 

 by means of pins at C and D drew seven parallel chords, marking their mid-points 

 as before. These mid-points, again, clearly lie on an inclined straight line, and 

 demonstrate a parabola with a non-vertical axis. The observations are marked along 

 the curve. The ordinates in both cases are galvanometer deflections, and the tempera- 

 ture, in both cases, is measured from a mean zero at 12*5° C. 



To examine this point further, I took a set of experiments of Messrs Holborn 

 and Wien, # and a set of experiments of Professor BARUs.f The former was a set 

 of observations on "Thermo-couple A," the first series given on p. 127 of their paper. 

 The curve shown on fig. 23 having so great a ,sweep, and so slight a curvature, was 

 difficult to construct so as to maintain both constancy of curvature and constancy 

 of change of curvature. But all the recorded observations are marked along it by 

 means of small crosses, so that I must leave it to others to say if the curve fairly 

 represents the observations. The curve produced is a facsimile of the original curve 

 employed. At A and B is shown the pin method of getting parallel chords, eleven in 

 number. Short lines are drawn perpendicular to the chords at the points where they 

 cut the curves ; the mid-point of each chord bisects the distance between these two end- 

 marks. The locating of these end-marks can be done with considerable precision by noting 

 the positions of two ordinates (one on each side of the required end-mark) on which the 

 chord and the curve are separated (say) by one millimetre. Beginning with the longest 

 chord, and excepting the 8th, 9th, and 10th mid-points, the remaining eight lie very nearly 

 on such a line as the one shown. Here, again, the evidence of a slant parabola seems clear. 

 Similarly with Professor Barus' observations, I took the first set, Series I. on p. 15 

 of his paper. The same difficulties occurred here as with Messrs Holborn and Wien's 

 curve ; but the same care was exercised, and the six mid-points of the six parallel 

 chords show an unmistakably slant parabola. At C and D are shown the arrangements 

 for the pin method of drawing the parallel chords ; and the observations and end-marks 

 are given as in Holborn and Wien's curve. 



* Wied. Ann., xlvii. (1892), pp. 107-134. t Phil. Mag., Ser. V., vol. xxxiv. pp. 1-18. 



TRANS. ROY. SOC. EDIN., VOL. XLVII. PART IV. (NO. 25). 115 



