282 BRIDGMAN. 



In addition to the pressure measurements, the thermal e.m.f. of 

 each of the specimens was found at atmospheric pressure against lead. 

 The same potentiometer was used as for the pressure measurements, 

 and no special comment is necessary. One terminal of the couple 

 was kept in an ice bath; the other was placed successively in baths at 

 25°, 50°, 75°, and 95°. Twelve couples were made up and measured 

 simultaneously. As a matter of fact, the readings were actually 

 made against copper; the e.m.f. against lead was obtained by a sub- 

 traction. The e.m.f. against temperature of each of these couples 

 could be represented within the limits of error by a power series in t 

 (three terms were always sufficient). These formulas are given in 

 the following as a further means of identifying the metal. By one 

 and two differentiations the Peltier heat against lead and the Thomson 

 heat at atmospheric pressure may be at once obtained. These are 

 also listed in the following. A positive e.m.f. means that the positive 

 current flows from lead to the metal at the hot junction. 



Methods of Computation. 



The thermal e.m.f. in the pressure circuit was first computed for 

 each reading, applying all corrections, including temperature correc- 

 tions. The points were then plotted on a large scale, thermal e.m.f. 

 against pressure, and smooth curves drawn through the points at 

 each temperature. In this way four smooth curves were obtained 

 for each substance, or really five curves, because the curve correspond- 

 ing to 0° coincides with the pressure axis. From this bundle of smooth 

 curves the thermal e.m.f. could be read as a function of temperature 

 at constant pressure. The curves were in this way replotted, giving 

 thermal e.m.f. against temperature at constant pressure. Six pressures 

 were chosen, 2000, 4000, 6000, 8000, 10000, and 12000, giving six 

 smooth curves of thermal e.m.f. as functions of temperature. The 

 smoothing so far was done graphically. A further smoothing was next 

 performed by calculation. The ordinates were read from the six 

 smooth curves at intervals of 10°, and the differences of successive 

 ordinates computed. The curves were now further adjusted so that 

 the successive differences should lie on a smooth curve when plotted 

 on a scale 10 times as large as the original curve. This second ad- 

 justment was of course performed so as not to produce any changes 

 greater than the uncertainties in the original plot. The smoothed 

 first differences, obtained in this way, are equal, without appreciable 



