CONDUCTIVITIES OE IRON AND COPPER. 
573 
within a quarter of a degree. A much greater error than this would have no effect 
on the results given below, which are for temperatures below 200° C. 
The rates of cooling of the bars at different temperatures were determined by 
heating a piece of each bar, about a foot long, in a sand-bath to a temperature higher 
than required, and observing its rate of cooling under the same conditions as those to 
which the experimental bar was exposed. For this purpose it was suspended in the 
position originally occupied by the long bar, the thermo-electric junction inserted in a 
hole drilled at its middle point, and the deflection of the galvanometer observed at 
regular intervals during the cooling. It \vas found that the observations of the rate 
of cooling were not trustworthy for the first few minutes after it had commenced, so 
that it was necessary to heat the bar to a temperature considerably higher than the 
highest at which the rate of cooling was required. With this precaution it was found 
that the data obtained for rates of coolino- under similar conditions agreed "well 
^ O 
together—for example, the data given in columns a and h of Ta,ble I. 2 were obtained 
on different days but under similar conditions. Experiments showed that the presence 
of a thin film of oxide on the surface of the copper bar had no perceptible effect on its 
rate of cooling, for it was found that the film did not form unless the heating was 
continued for a long time, so that by varying the time of heating in the sand bath 
it was possible to observe the cooling with the surface in the diflerent initial stages of 
oxidation, and the rate of cooling was found to be practically the same in all cases. 
In reducing the insults, the graphical method was largely used—the distribution of 
temperature along the bars and the cooling observations were plotted in the usual 
way, and from the curves so obtained differential curves were plotted showing, in tlie 
one case, the gradient of temperature at any point on the bars, and, in the other, the 
rate of cooling at any temperature. In the case of the curves showing the stationaiy 
state of the bars, the scale adopted was such that abscissae represented actual distances 
along the bar, and the lengths of the ordinates in millimetres ga.ve the corresponding 
temperatures in scale divisions. In the differential curves, however, showing the 
gradient of temperature at any point on the bar, the ordinates were drawn on ten 
times this scale, so that one millimetre represented about ^-Q-th of a degree Centigrade. 
The cooling curves were drawn on a somewdiat larger scale in overlapping sections. 
The scale varied, according to the data to be plotted, from one in which 4 centims. 
represented one minute and 5 millims. one scale division (1 millim. to -^vjth of a degree 
Centigrade) to one in which 1 centim. represented one minute and 4 millims. one scale 
division. Owing to the size of the scale, it w^as necessary to draw the curves in 
sections, and. experience showed that the sections must overlap considerably, in order 
to secure continuitv. The differential curves showinof the rate of coolino’ at anv 
temperature were drawn with abscissae on the same scale as the observational curves, 
but the ordinates were in each case on a scale ten times as great. As it is impossible 
to reprciduce these curves on the scale to which they were drawn, two specimen curves 
are given on a reduced scale in figs. I and II (p. 5'JO). Fig. I shows the up})er portion 
