534 Prof. H. L. Callendar on a 



The symbol pt' here denotes temperature on the S2ale of the expansion 

 of platinum, taking 1000° C for the fundamental interval, so that each 

 degree corresponds to one thousandth part of the expansion between 0° and 

 10U0 C C. In order to reduce the meldometer reading to this scale, it is 

 only necessary to observe the readings corresponding to the F.Ps. of gold 

 or silver, which are 1071° and 9-j5° respectively on the expansion scale. 

 Let e' be the reading iu scale-divisions corrected to0° O. for the expansion 

 corresponding to the F.P. of guld. Reduce all the readings (corrected 

 to 0") in the proportion of 1071 to e '. This gives the temperature pt' in 

 degrees on the expansion scale. Add to these the differences given by 

 the ordinates of the difference-curve (1), or by the formula, to find the 

 corresponding values of the temperature t on the centigrade scale. It 

 mast be remembered that the abscissa of the difference-curve is t and 

 not pt', but it is very easy to guess a first approximation to t by inspec- 

 tion. It is also very easy to redraw the curve in terms of pt' as abscissa 

 if desired. 



Similarly for calibrating a thermocouple (Pt-Ir or Pt-Rh), observe the 

 thermo-E.M.P. e" (corrected to 3 C), in scale-divisions of the potentio- 

 meter or galvanometer, corresponding to the F.P. of gold. Reduce all 

 the readings to the platinum thermoelectric scale pt" by multiplying by 

 1076 and dividing by e". Add the corresponding ordinates of the 

 difference-curve (3). This method is justifiable because the 10 per cent. 

 Pt-Ir and Pt-Rh couples give similar difference-curves, as shown by 

 direct comparison of the two, and also by the observations of Barus and 

 Le Chatelier, which are indicated by the points marked ( X ) and ( + ) 

 respectively in fig. 1. The smooth curve was deduced from a comparison 

 with the platinum resistance-thermometer (H. M. Tory, B. A. Rep. 1897), 

 and agrees very closely with the observations of Holborn and Wien. The 

 agreement of the observations is satisfactory when we consider that those 

 of Barus are affected by the uncertainty of the fixed points selected for 

 comparison, and those of Le Chatelier also by the scale errors of the 

 galvanometer. The following are the observations given by Le Chatelier 

 (Comptes Rendus, 189o) : — 



Temperature *! 100° 218° 306° 357° 445° 961° 1061° 



Deflexions 5 145 2P8 26 33 8P5 917 



Scaled" 59° 170° 255° 304° 386° (953°) 1071° 



Difference*-^'' 4 1° 48° 51° 53° 59° 8° -10° 



The temperatures t are taken on the proposed B.A. scale, and not as 

 originally given by Le Chatelier. The last two lines are calculated by 

 the method above des2ribed, assuming the F.P. of silver. An approxi- 

 mate difference-formula for the thermocouple between the limits 700° 

 and 1200° is 



t -pt"= 0-00023 t(t- 1000). 



In applying this method to thermocouples it must be remembered that 

 variations in the quality of the platinum may produce changes of over 

 100 per cent, iu the thermo-E.M.F. (Stansneld, Phil. Mag. "1898, xlvi. 

 p. 71.) The effect of these variations at temperatures near 1000° C. is 

 minimized by the simple method of calibration above described ; but 

 there may remain discrepancies of 5° or 10° at 300°-400° C. These may 

 be avoided by observing also the S.B.P. at 444'5° C, and calculating a 

 logarithmic reduction formula, such as that proposed by Stansfield, which 

 agrees very fairly with the difference-curve above given, 



