forms of
suspension when employed for any kind of torsion balance, from an
ordinary more or less "astatic" galvanometer to the Cavendish
apparatus. In the first place the actual strength of the fibres under
longitudinal stress is remarkably high, ranging from fifty to seventy
tons weight per square inch of section, and even more than this in the
case of very fine threads; the second and more important point in
favour of quartz depends on the wide limits within which cylindrical
threads of this material obey the simplest possible law of torsion,
i.e. the law that for a given thread carrying a given weight at a
given temperature and having one end clamped, the twist about the axis
of figure produced by a turning moment applied at the free end is
proportional simply to the moment of the twisting forces, and is
independent of the previous history of the thread.

It is to be noted, however, that the torsional resilience of quartz as
tested by the above law is not so perfect but that our instrumental
means allow us to detect its imperfections, and thus to satisfy
ourselves that threads made of quartz are not things standing apart
from all other materials, except in the sense that the limits within
which they may be twisted without deviating in their behaviour from
the law of strict proportionality by more than some unassigned small
quantity, are phenomenally wide.

A torsion balance--if we except the case of certain spiral
springs--is almost always called upon for information as to the magnitude
of very small forces, and for this purpose it is not essential merely
that some law of twisting should be exactly obeyed, but also that the
resistance to twisting of the suspension should be small.

Now, regarded merely as a substance possessing elastic rigidity,
quartz is markedly inferior to the majority of materials, for it is
very stiff indeed; its utility depends as much as anything upon its
great strength, for this allows us to, use threads of exceeding
fineness. In addition to this it must be possible, and moreover
readily possible, to obtain threads of uniform section over a
sufficient length, or the rate of twist per unit length of the thread
will vary in practice from point to point, so that the limits of
allowable twist averaged over the whole thread may not be exceeded,
and yet they may be greatly overpassed at particular points of the
thread.

It is interesting to note that in the case of quartz we not only hav