OF OASES AT HIGH EXHAUSTIONS. 
405 
be real, though the slight uncertainty as to the second makes it doubtful. If it is real, 
it is probably due to the heating of the hinder face of the mica by conduction from the 
front face ; so that the permanent difference of temperature is a little less than was 
the difference at a somewhat earlier stage. 
658. There is great satisfaction to the mind when it is possible to plot a long series 
of figures as a curve on paper. Indeed, this is almost the only way in which the 
meaning of the results can be fully realised. It has been found impracticable to place 
the curves together in one table, inasmuch as the diagrams necessary to enable the 
results to be understood, and without which columns of figures are almost valueless, 
have to be on a scale of increasing magnification. On diagram A (Plate 56) are shown 
the results recorded in the first part of the table down to a pressure of Off 6 millim. 
plotted as a curve (marked “ air ”), the ordinates being the logarithmic decrements, 
and the abscissae the pressures in millims. of mercury. The total barometric height, or 
one atmosphere, is represented by a length of 152 millims. : each millim. on the scale 
therefore represents 5 millims. pressure of mercury. Gaps in the series have not been 
filled up by interpolation. 
Starting from the logarithmic decrement 0'1124 at 760 millims., it diminishes 
very regularly, but at a somewhat decreasing rate. Between 50 millims. and 3 millims. 
the direction is almost vertical, and it will be noticed that a great change in the 
uniformity of the viscosity curve commences at a pressure of about 3 millims. (shown 
in the upper part of the curve on diagram A), where the direction suddenly changes 
and goes off almost horizontally. At this point the previous approximation to 
Maxwell’s law begins to fail, and further pumping considerably reduces the loga¬ 
rithmic decrement. 
659. To render the gradation of the air curve more uniform, it is now necessary to 
considerably magnify the scale. Diagram A is on so small a scale that 152 millims. 
represent the whole barometric column, but the scale to which diagram B is drawn is 
that of 1 millim. to 5 millionths of an atmosphere, so that the whole atmospheric 
pressure would be represented by a vertical height of 200 metres. One-thousandth 
part of this, representing the top 200 millims. of the scale, is all that I give in 
the compass of this diagram. At the lower part of diagram B (Plate 57) is given 
a highly condensed scale of the viscosity curve between 760 millims. and 0ff6 millim. 
To bring this condensed scale to the same proportion as that of the rest of the diagram 
it would have to be 200 metres instead of only 10 millims. long. 
Starting from 1000 M the diminution of viscosity is very slight until the exhaustion 
reaches about 250 M ; after that the viscosity gets less with increasing rapidity, and 
falls away quickly after 35 M is reached. 
660. It will be seen, from the almost horizontal character of the upper part of the 
curve, that the scale is not even now large enough to bring it out properly. I have 
therefore in diagram C (Plate 58) again lengthened the scale, so that one atmosphere 
is represented by 5000 metres, and of this I have taken the top 300 millims., giving 
