PETRCE. — TEMPERATURE COEFFICIENTS OP MAGNETS. 555 



cut on a milling machine, and a rectangular cavity, open to the outside 

 air but closed to the inside of the box, was constructed by soldering two 

 thin pieces of brass into the end and top of the slot. Into this cavity a 

 set of forms carrying thin coils of the shapes needed, fitted exactly. The 

 box itself, and the cover, were mounted on the face plate of a lathe and 

 turned off smooth, so that when a piece of rubber packing was inserted 

 between the two, and the whole was screwed together, the case thus made 

 was water-tiofht. The box was mounted on a wooden frame which had 

 sliders for the forms which carried the coils. The magnet to be tested 

 was fastened firmly in place by a holder not shown in the figure, and the 

 box was connected with a set of pipes so that cold water, warm water, or 

 steam could be sent through it at pleasure. 



The temperature coefficient of a bent cast iron magnet, as defined 

 above, generally increases with the temperature, but for purposes of com- 

 parison, we may use the mean value A' of this coefficient between 10° C. 

 and 100° C. 



Three magnets of the form marked 1, chilled by Mr. Thompson and 

 weighing as much as 1250 grams each (nearly three pounds), gave for K 

 the values 0.0003G, 0.00037, and 0.00034 respectively ; another magnet 

 of the same pattern treated by a maker of hardened cast iron machinery, 

 yielded the value 0.00082. Whatever the secret process employed in 

 this last case, the resulting magnet was by no means so useful as those 

 made from castings chilled in the manner described above. 



Uncliilled castings make very undesirable magnets, for the temperature 

 coefficients are usually five or six times as large as in the case of chilled 

 magnets, and it seems impossible to get their magnetic moments really 

 permanent. Curiously enough the chilling process makes a casting less 

 brittle than before, and causes the grain of a fracture to be finer and 

 more uniform. 



The values of K seem to indicate that the whole interior of the casting 

 is affected by the chilling, whereas it is extremely difficult to harden a 

 thick piece of steel uniformly. It did not appear that a magnet made 

 up of a lot of thin plates chilled separately had a smaller temperature 

 coefficient than a solid magnet of the same dimensions. 



Castings of the shapes marked 3, 4, and G weighed about 260 grams, 

 160 grams, and 500 grams, respectively, and yielded for K the values 

 0.00040, 0.00040, 0.00031. The actual temperature coefficients at low 

 temperatures are always less than these mean values, and in the case of 

 the last mentioned form the coefficient is not greater than 0.00013 between 

 10° C. and 40° C. I have myself never found a value quite so small as 



