116 MR. F. E. SMITH ON THE CONSTRUCTION OF 



the change. The observations (;i large number of coils weiv employed) show that the 

 three coils assumed to be constant are liable to considerable fluctuations in value, 

 while Flat appears to be a remarkably constant coil ; this is shown in the ' B. A. 

 Report' for 1903. Consequently, the values assigned to the resistances in 1888 were 

 probably too small by O'Ol per cent. Correcting for this, the length of the mercury 

 column becomes 106 '29 centims. (the probable error of the investigation being 

 0'004 per cent., the third decimal figure has been discarded). The value given on 

 p. 114 is 106'292 centims. ; thus the two results are identical to O'Ol per cent. 



The definitions of the mercury column representing the ohm are different in the 

 two cases. In 1888 the definition necessitated a knowledge of the density of 

 mercury, while at the present time it is in terms of a length and a mass. The value 

 taken by Dr. GLAZEBROOK for the density of mercury at C. was 13 '5957 grammes 

 per cubic centimetre ; the difference in the two definitions is therefore small, being 

 equal to one part in one hundred thousand. The identity of the results to O'Ol per 

 cent, is not therefore affected. 



Constancy of Mercury Standards. 



Some information respecting the stability of the tubes may be gleaned from ice- 

 point determinations of thermometers. The permanency of the mercury resistances 

 depends entirely on the freedom of the glass from strain, or, if in a strained condition, 

 the maintenance of that strain. Now the walls of a mercury tube are much thicker 

 than the walls of a thermometer bulb. Probably, therefore, complete recovery from 

 strain will be more difficult in the former case than the latter. On the other hand, 

 the distorting forces to which a thermometer bulb is subject are much greater and 

 more variable than those acting on a mercury standard. Under similar conditions, 

 therefore, a mercury tube is probably not more liable to change than the bulb of a 

 thermometer ; also, the condition of a tube when a measurement of its resistance is 

 being made is very similar to that of a thermometer which has been raised to the 

 steam-point shortly before an observation of its zero. The constancy of such zero 

 readings should therefore be comparable with the constancy of the resistance measure- 

 ments. In the case of well-annealed Verre dur thermometers, Dr. J. A. HARKER, 

 of the Thermometric Department of the Laboratory, assures the writer that the 

 difference between such zero readings, spread over a considerable number of years, 

 will not average, in the general case, more than a few hundredths of a degree centi- 

 grade. Since a change of 0'1 C. corresponds to an alteration in the capacity of the 

 bulb of 0'002 per cent., changes of considerable magnitude in the standards of 

 resistance are not anticipated. 



The constancy of the mercury standards of the Phys.-Tech. Reichsanstalt is shown 

 in vol. iv., ' Wissenschaftliche Abhandlungen der Phys.-Tech. Reichsanstalt.' The 



