Barrett — On some novel Thermo- Electric Phenomena. 129 



The value of the scale-reading of the galvanometer in microvolts was 

 accurately determined by means of a standard cell and potentiometer, one scale 

 division being found to be equal to 26-5 microvolts. With a similar couple 

 Professor Callendar found the E. M. F. in microvolts at three points ought to be 

 as follows : — 



100° C. . . . 650 microvolts 



445° . . . 3630 „ 



1000° . . . 9550 „ 



the other junctions being in ice. My own determinations at 100° C. and 445° C. 

 gave almost identically the same values in microvolts ; this may be accidental, but 

 the variations of thermo-electric power in diffei'ent s^iecimens of platinum is 

 probably very slight. I therefore took Professor Callendar's value in microvolts 

 for 1000° C. in preference to the rather different number I obtained.* The results 

 are plotted in the curve shown in Plate IX«., marked " pyrometer couple." 



We will now return to the thermo-electric behaviour of this nickel-manganese- 

 iron alloy. An iron wire, drawn from the purest commercial iron, was coupled 

 with a wire of the alloy. The wires were twisted together at the junction and 

 then brazed. After insulation with asbestos they were lashed alongside of the 

 platinum- rhodium pyrometer couple, and the pair of couples were then inserted 

 in the centre of a thick iron tube held horizontally in, and heated by, a gas 

 furnace, the temperature of which could be raised to a white heat. The cooler 

 junctions of both couples were kept in ice, and pairs of readings were taken as 

 the temperature was raised, and again as it was lowered ; the rise and fall being 

 slow and steady in both cases. As the readings of each couple were taken alter- 

 nately, to obtain a true comparison three readings were necessary in each case, 

 first, of one couple A ; then, of the other B ; then, of A again ; the mean of the 

 first and last readings of A being comparable with B. The galvanometer being 

 extremely dead-beat, all three readings could be taken closely together. This 

 comparison was repeated two or three times every minute during heating and 

 cooling, thus several series of readings for about every 10° C. rise or fall of 

 temperature were obtained ; the readings which corresponded to similar tem- 

 peratures in heating and cooling being remarkably concordant, allowance being 

 made for a very small constant difference to be mentioned in the sequel. 



The results are plotted in the second curve shown on Plate IXa. It 

 will be seen that, up to a temperature of 320° C, the E. M. F. of the nickel-steel 

 alloy and iron couple rose rapidly ; it then remained absolutely constant until 



* As (within the limits of the scale) the E. M. F. of the thermo-couple is directly proportional to the 

 scale-readings of the galvanometer, it is easy, when the E. M. F. in microvolts at 100° and 445° C. is 

 found, to determine the higlier points in the curve by means of the ratio given above : the scale-readings 

 being plotted as ordinates and the temperatures as abscissae. 



U2 



