112 On the Construction of Platinum Thermometers. 



impossible, to attain an accuracy of the order of a tenth of a 

 degree even at temperatures as low as 200° C. 



On the other hand, when a platinum thermometer has once 

 been annealed (a process which can be completed in a few 

 minutes), its zero will not be found to vary even by a 

 hundredth of a degree — if it is properly constructed — when 

 used at any temperatures up to 500° C. ; and in some 

 thermometers which I have used at temperatures as high as 

 1300° C, I have not found any changes of zero amounting to 

 more than a tenth of a degree. 



Sensitiveness and rapidity of observation are often of con- 

 siderable importance. A mercury thermometer with an open 

 scale must necessarily contain a relatively large mass of 

 mercury, since there are objections to unduly diminishing the 

 bore of the tube. Thus it cannot be made so sensitive as a 

 platinum thermometer in which the actual mass of metal may 

 be very small. A still more important point is that, owing to 

 the imperfect elasticity of the glass, it is desirable to allow 

 the mercury thermometer some time to settle before taking 

 a reading. This is of course unnecessary in the case of the 

 platinum thermometer. 



In ordinary work with mercury thermometers it is gener- 

 ally necessary to expose some portion of the stem to the air. 

 If a volume of mercury corresponding to any considerable 

 number of degrees be thus exposed, the error becomes serious. 

 It may amount to as much as 10° at 350° C. This correction 

 is so uncertain that it is now generally considered inexpedient 

 to attempt to apply it. The only way to avoid it, and at the 

 same time to secure a sufficiently open scale for accurate work 

 without unduly increasing the length of the stem, is to use a 

 series of thermometers of " limited scale. " Each of these 

 must have at least two points of its scale specially determined. 

 It has hitherto been the custom to graduate such thermo- 

 meters by means of substances of known boiling- and freezing- 

 points ; but, as Griffiths * has shown, the graduation may be 

 much more easily and accurately effected by comparison with 

 a single platinum thermometer, a method which has the 

 further advantage of saving the trouble of calibrating the 

 stem. 



It is easy to construct a platinum thermometer with a scale 

 of 1 centim., or even 10 centim., to the degree, whose reading- 

 shall be entirely independent of the length of stem immersed. 

 The divided wire on which the readings are taken need not 

 be more than 20 centim. in length, and can be made to corre- 

 spond to any part of the scale by simply unplugging a few 

 * B. A. Report, 1890. 



