460 Prof. R. Threlfall on the Measurement 



incline to think that nothing is better than a trace of slow- 

 drying white paint. Amongst other experiments I tried 

 cementing two very thin glass disks together, selected so as 

 to mutually correct each other by the drying of the cement. 

 I also tried a method of using plaster of Paris. As plaster of 

 Paris expands on setting I covered the back of a thin mirror 

 with a layer of it about ^ m - i n thickness on drying, this of 

 course forced the mirror into a concave form. The back of 

 the plaster was rubbed away on a fine tile till it was only 

 about ^ in. thick, and the mirror still remained very con- 

 cave. Since cementing magnets on to disks with shellac 

 varnish invariably forces the mirror to become convex, I 

 hoped that I should obtain a correction of the concavity pro- 

 duced by the plaster by the convexity which the shellac tends 

 to provoke. My anticipation was completely realized ; the 

 mirror on examination turned out all that could be desired, 

 but, alas, was too heavy for the purpose for which I required 

 it. I can, however, most strongly recommend the process to 

 anybody who desires a mirror to be flat and does not mind it 

 being heavy. The best way is to use very little plaster and 

 then to leave the mirror concave. This concavity can be 

 removed by painting on small successive films of shellac ; it 

 must be remembered that shellac films go on contracting for 

 several days after they cease to be sticky. Equally good 

 results can of course be obtained by cutting out (with a 

 rotating tube and emery) disks of the size required from 

 previously examined thin sheet glass. The surfaces generally 

 require regrinding. The advantage of the process described 

 is that it enables thin cover-slip glass, which is generally to 

 hand, to be kept flat. 



The mirror having been mounted on the steel disks, these 

 last were magnetized by the apparatus mentioned above. In 

 carrying out this operation the following precautions have to 

 be observed: — 



1. The distance between the pole-pieces requires to be the 

 same for each magnet. This was attained by setting them 

 to touch a carefully prepared brass rectangular bar. 



2. To annul the effect of any small outstanding differences 

 between the cores of the magnets the current was supplied to 

 them in multiple arc, and was strong enough to magnetize 

 the cores beyond the saturation point. 



The approximate moment of inertia of the magnet system 

 was easily calculated, and it was found that the astaticism 

 was at least ten times as good as the best I had been able to 

 obtain with small bar-magnets mounted on mirrors or mica, 

 and magnetized with a small horseshoe magnet. Of course, 



