Condition of Iron hy Heating to Redness and Cooling. 481 



always somewhat smaller (at most by 0'007) than at first, and 

 only reached the initial value three times, i. e. the mean density 

 of the unattacked and newly tempered bar ; on the average it 

 was smaller by 0*004. With the bar (b) the density at first 

 diminished rapidly, then increased, and then gradually sank 

 with small variations to a constant value, greater than the 

 initial value. 



With the other two bars (c) and (d\ which were cooled 

 in colder water, the density alternately diminished and in- 

 creased, and the changes took place much in the same way, 

 a maximum or minimum occurring about the same densities 

 of the bar. In both cases the density remained lower than 

 the initial value ; on the average it was about 0*015 less. 



Of these four series of observations the second leads to 

 very remarkable results. If we calculate the densities of the 

 layers removed between the first and second observation, and 

 between the second and third observation, we find the values 

 to be 16 and 2. The thickness of each of the two layers 

 would be about 0*002 millim. 



Although, on account of the smallness of the quantities 

 dissolved off, these numbers may be affected by errors of re- 

 latively large magnitude, yet, unless we suppose gross errors 

 of observation to exist, which I do not consider to be pro- 

 bable, the fact remains that between an external dense layer 

 and a central portion of nearly uniform density there existed 

 a layer of extraordinarily small density. Apart from this 

 peculiar condition of the surface, the distribution of density 

 is quite regular — the density decreases until it reaches a con- 

 stant value. Consequently we have a dense envelope en- 

 closing a less dense kernel. 



The first series of observations, which relate to a bar tem- 

 pered in warm water, shows an extraordinarily small difference 

 in density. The greatest difference between any two obser- 

 vations amounts to seven in the third decimal place, which 

 may certainly be attributed to accidental want of homogeneity. 

 From the fact that the density becomes less upon the whole, 

 we may certainly conclude that this bar must have possessed 

 a density decreasing, with small variations, from outside 

 towards the interior. 



The same conclusion follows from the measurements made 

 with the two pieces (c) and (tZ), which were tempered in cold 

 water. With them, however, the difference between the 

 density of the exterior layer and the interior is much greater 

 than with the bar (a), but the irregularities in the distribution 

 of density are also much greater. 



I have calculated the densities of the layers, dissolved away 



