Practical Thermometric Standard. 537 



manner which it is often impossible to calculate. In the 

 case of the platinum- thermometer at 1000°, the rise of zero 

 is permanent and is easily corrected. At lower temperatures 

 the changes of zero of a platinum-thermometer are generally 

 negligible. At temperatures above 1100°, it is best to use 

 a smaller fundamental interval, and to dispense with the mica 

 frame. Fused silica appears to be the most suitable material 

 for the containing-tubes, in cases where a tube form of ther- 

 mometer is required. Such tubes are difficult to make in the 

 present state of the art, but are tougher and more impervious 

 than porcelain, and are free from the objectionable glaze. In 

 many experiments {e.g. radiation, and flame temperatures) 

 no containing-tube is required, and the naked wire may be 

 used successfully at temperatures very near its melting-point, 

 provided that care is taken not to strain it or expose it to 

 metallic vapours 



For verifying the uniformity of the standard sample of 

 platinum, and for comparing the scales of different specimens, 

 it is proposed to employ the method of comparison originally 

 devised in 1886. In this method, the wires to be compared 

 are wound side by side symmetrically in a double screw-thread, 

 so as to be always at the same mean temperature, even if the 

 temperature distribution is not quite uniform. The pair of 

 coils are enclosed in a single tube, which is heated in a muffle 

 of concentric tubes arranged so as to secure great steadiness 

 and uniformity of temperature. The difference between the 

 resistances can be very accurately observed, as it remains 

 small and nearly constant throughout the scale, provided that 

 the fundamental intervals of the two wires have been adjusted 

 to approximate equality. The comparison of the scales of 

 different wires can be effected by this method with an 

 accuracy which greatly exceeds that of any other thermo- 

 metric comparison. 



For standardizing individual platinum-thermometers, it is 

 generally sufficient, owing to the simplicity of the difference- 

 formula and to the regularity of the scale, to take an obser- 

 vation at a single secondary fixei point (in addition to the 

 primary fixed points 0° and 100° 0.) for the determination 

 of the difference-coefficient. Thus, by a single observation, 

 the whole scale is calibrated to a very high order of accuracy 

 for a range of 500° or 1000°. 



The boiling-point of sulphur is probably in most cases the 

 host secondary fixed point to choose for this de ermination. 

 It has hitherto been very generally employed for this purpose, 

 and it is therefore yjroposed to assume this point as the basis of 

 the British Association scale. The numerical values of tem- 

 peratures expressed on the proposed B. A. scale will depend 



