and Maximum of Magnetism of Nickel and Cobalt. 331 



tizing-forces fi is increased to about three or four times its value 

 at 15° C, and the maximum value of fju is increased from 222 

 to 315. When 23 has a value of 32,000, p is not affected by 

 this change of temperature, seeing that the two curves coincide ; 

 but above that point fju is less at 220° C. than at 15° C. In 

 other words, if nickel is heated from 15° C. to 220° C, the mag- 

 netization of nickel will increase if the magnetizing -force is small, 

 but will decrease if it is large. It is impossible to say at present 

 whether increase of temperature above 220° will always produce 

 effects in the same direction as below it or not. 



These remarkable effects of heat, it seems to me, will, when 

 followed out, lead to the discovery of most important connexions 

 between heat and magnetism, and will finally result in giving 

 us much more light upon the nature of heat and magnetism, 

 and that equally important question of what is a molecule. To 

 accomplish this we must obtain a series of curves for the same 

 ring between as wide limits of temperature as possible. We 

 must then plot our results in a suitable manner ; and from the 

 curves thus formed we can find what would probably happen if 

 the temperature were lowered to the absolute zero, or were in- 

 creased to the point at which nickel is said to lose its magnetism. 

 In such inquiries as these the graphical method is almost inva- 

 luable, and little can be expected without its aid. 



In applying the formula to this curve, we do not find so good 

 an agreement as at the lower temperature. I do not consider 

 this conclusive that the formula will not agree with observa- 

 tion at this temperature ; for I have noticed that the curves of 

 different specimens of iron and nickel seem to vary within a 

 minute range, not only in their elements but also in their form. 

 This might perhaps be accounted for by some small want of ho- 

 mogeneity, as in the case of burning in iron and nickel ; but at 

 present the fact remains without an explanation. But the 

 amount of the deviation is in all cases very small when all the 

 precautions are taken to insure good results. The nature of the 

 deviation is in this case as follows : when the constants in the 

 formula are chosen to agree with the observed curve at the ver- 

 tex and at the two ends, then the observed curve falls slightly 

 below the curve of the formula at nearly all other points. In a 

 curve plotted about 5 inches high and broad, the greatest dis- 

 tance between the two curves is only about T V of an inch, and 

 could be much reduced by changing the constants. For the 

 benefit of those who wish to study this deviation, I have calcu- 

 lated the following values, which will give the curve touching 

 the vertex and the two ends of the observed curve of Table III. 

 They are to be used by plotting in connexion with that Table. 



