696 
substance, and not on external circumstances. This is not the case 
with the further course of the curve. This course depends on the 
properties of the tube in which the crystallisation takes place, and 
can be calculated when the necessary data about the crystallizing 
substance and the tube are known’). 
As it is of importance to determine the accurate value of the 
temperature of the boundary, it is necessary to consider the sources 
of errors that may play a part in the measurement. 
The vertex of the registered curve in Fig. 1 is sharp, and makes 
the impression that phenomena of inertia have not played an im- 
portant part even in the quickest part of the process. This sharpness 
is, however, only apparent. When the quickness of the registering 
drum is increased, and the sensitiveness of the galvanometer dimi- 
nished, the top of the curve appears to be more or less rounded, 
as is seen in Fig. 2. 
Fig. 2. 
This rounding is not found in Fig. 1, because there the scale of 
the figure is smaller in horizontal direction, and larger in vertical 
direction than in Fig. 1. The cause of the rounding lies in the 
finite dimensions of the thermo-element. 
By extrapolation of the two branches of the curve, as is 
indicated by the dotted line in Fig. 2, an ideal curve may be traced, 
which would give the course of the deflection of the galvanometer, 
when the thermo-element was infinitely small and the galvanometer 
infinitely quick. That this extrapolation cannot lead to a much higher 
temperature than is shown in the figure, may appear from the 
following considerations and experiments’). 
1. The inertia of the galvanometer cannot give rise to an appre- 
ciable error. The temperature changes during its adjustment (0,07 sec.) 
by only a very small amount, even at the moment of its most 
rapid ascent or descent. 
2. As may be understood a priori, the breadth of the thermo- 
ble. 
4) This appears, however, most convincingly from the agreement between the 
calculated and the observed temperature (see p. 701). 
