decreases in a gradual and perfectly regular manner. For example, one
cylinder which gave 0.03 per cent. of air by volume, after three months'
constant use gave 0.02 per cent. The advantage of using CO_{2} from this
source is obvious when compared with the difficulty of evolving a stream
of gas of constant composition from a Kipps or Finkener apparatus. A
micrometer screw, in addition to the main valve of the CO_{2} cylinder, is
useful for governing the rate of flow. A blank experiment should be made
to ascertain the amount of air in the CO_{2} and the correction made in
the readings afterwards.

[Footnote A: _Jour. Soc. Chem. Ind._, June 30, 1902, p. 819.]

[Illustration: Fig 51.--Will's Apparatus for Testing Nitro-cellulose]

~Measurement of Pressure and Rate of Flow.~--Great attention is paid to
the measurement of the rate of flow of gas, which is arrived at by
counting with a stop-watch the number of bubbles of gas per minute in a
small sulphuric acid wash bottle. A mercury manometer is introduced here,
and is useful for detecting a leak in the apparatus. The rate of flow that
gives the most satisfactory results is 1,000 c.c. per hour. If too rapid
it does not become sufficiently preheated in the glass spiral, and if too
slow there is a more rapid decomposition of the nitro-cellulose by the
oxides of nitrogen which are not removed.

~Decomposition Tube.~--This is of the form and dimensions given by Dr Will
(15 mm. wide and 10 cm. high), the preheating worm being of the thinnest
hydrometer stem tubing. The ground-in exit tube is kept in position by a
small screw clamp with trunnion bearings.

~Bath.~--To permit of two experiments being carried on simultaneously, the