OF THE NATIONAL PHYSICAL LABORATORY. 



359 



It will be observed that both the formulae just quoted do not apply at lower 

 temperatures, being nearly 40 C. in error at 0, and that therefore extrapolation 

 downwards even over a narrow range is not permissible. The error of the formula 

 for NPL 2 was determined to be 4 at 200. 



XV. Determination of the Fixed Points and 100 and Sulphur Boiling-point. 



The determinations of the fixed points and 100 for the gas and platinum 

 thermometers were made in baths specially built for eacli kind of instrument. The 

 ice-points were taken in glass vessels of a capacity of about G and '3 litres respec- 

 tively, consisting each of an inverted glass bell-jar with draining arrangement below, 

 and surrounded by a thick packing of cork clippings. Very little melting of ice took 

 place even in 12 hours, the upper surface of the ice -being protected by a thick felt 

 covering wrapped round the stem of the instrument. 



The block-ice previous to use was always well washed and finely divided by an ice 

 plane, and was repeatedly tested for dissolved impurity, the method adopted being to 

 ascertain the amount of chlorine present by addition of silver nitrate to the drainings. 

 It was found satisfactory, except on one occasion. 



The steam-point apparatus for the gas thermometer was of the usual type with 

 concentric tubes, but was arranged to be easily changed from the vertical to the 

 horizontal position by suitable couplings of wide compo tubing, connecting it to the 

 boiler. The steam- and zero-points for any series of comparisons were always taken 

 with the reservoir in the position in which it was used in that series, and in all cases 

 the amount of stem emergent was made as nearly as possible the same as in the 

 comparisons. The steam bath as arranged for the horizontal position is shown in 



