64 Mr. A. Stansfield on some Improvements in 



of uncertainty is introduced into the results. In addition to 

 this, the deflexion of the galvanometer for the highest tem- 

 perature to be measured must be limited to the width of the 

 photographic plate employed if the zero of the scale is to be 

 recorded on the same plate. In measuring high temperatures, 

 therefore, the sensibility of the galvanometer must be small, 

 and consequently the record will be on a small scale. 



For general use, these disadvantages are quite outweighed 

 by the fact that the whole record of the cooling of an alloy 

 can be obtained in one curve, but for special purposes, such 

 as obtaining accurate measurements of melting- or boiling- 

 points and investigating the exact shape of the u cooling- 

 curve " of a metal or alloy at the temperature of freezing or 

 of other molecular change, a modification of the method 

 becomes necessary. In order to obtain large-scale records of 

 these changes, Roberts-Austen employed a more sensitive 

 galvanometer, its zero position being adjusted by trial until 

 the spot of light from its mirror fell upon the photographic 

 plate when the thermo-couple was at the particular tempera- 

 ture at which a record of the cooling was desired. The zero 

 position of the enlarged record was often many feet from the 

 plate, and, as the galvanometer had to be deflected through a 

 large angle, its indications became untrustworthy, and it was 

 necessary to employ simultaneously a second ordinary 

 galvanometer to give a measurement of the temperature at 

 which the enlarged record was taken. The two galvanometers 

 were connected in parallel and arranged one behind the 

 other*, so that they both recorded their deflexion on the same 

 moving plate. A record taken in thi3 way is given in fig. o. 



It represents the melting and solidification of an ounce of 

 pure gold ; the thermo-junction was immersed in the molten 

 gold without any covering, and had therefore scarcely any 

 lag other than that produced by the galvanometer itself. 

 The curves A and B were produced by the less sensitive 

 galvanometer, and the scale on the left has been obtained by 

 measuring the ordinates of these curves from the datum on the 

 original plate. The scale for the more sensitive records C and 

 D is 22 times as great as the other, and it is obtained from 

 the latter by comparing the slopes of the two records. 



The curves A and C represent the melting, and the curves 

 B and D the solidification of the gold. The irregularities 

 in the melting record probably indicate the fall of pieces of 

 still solid gold into the molten mass. On account of this 

 irregularity cooling-curves are usually employed, instead of 



* Roberts-Austen, Proc. Inst. Mecli. Eng. 1893, p. 243, and fig. 4, 

 pi. xxxvii. 



