PKOFESSOI; H. B. DIXON ON THE MOVEMENTS 
O O i 
• )o4 
not always symmetrical. Fig. 25 (Plate 12) shows the two flames meeting symmetri¬ 
cally, and giving rebound waves faster and brighter than the original waves before 
collision. 
Fig. 26 shows the flame coming from the right hand to be much brighter and 
travelling much faster than that from the left hand (although the collision occurs 
only a little to the left of the centre of the tube). Fig. 27 shows these phenomena 
reversed. On rej^eating the experiments it was found that the faster flame had 
usually been affected by some impulses causing a sudden increase in its brightness 
and velocity, and also j^roducing a backward wave (analogous to a reflected wave). 
Sometimes the impulses were exhibited by the left-hand flame, as in figs. 28 and 29 
(Plate 13); sometimes by the right-hand flame, as in figs. 30 and 31. In some 
experiments a single impulse only is observed, as in figs. 29 and 31 ; in others several 
such impulses can be traced, as in fig. 30. Figs. 32 and 33 show both flames affected 
by these impulses ; in fig. 32 symmetrically, in fig. 33 unsymmetrically. 
The explanation of these apj^earances that first occurred to us was that the flame 
was preceded bv invisible sound-waves, travelling more quickly than the flame in its 
initial phases; that these sound-waves became visfl3le as soon as they met the flame 
moving towards them in the opposite direction (as in our previous experiments on 
sound-waves), and that, on the other hand, the visible flame meeting the sound-wave 
was affected by the sudden increase of pressure, and continued its journey with 
greater speed and luminosity. This explanation was at once destroyed when Tre 
found similar impulses in a flame which was sent through the apparatus in one 
direction only (the tap B being closed), as is shown in fig. 34. 
It next occurred to us that these impulses might be due to the explosion catcliing 
up its own sound-waves. If sound-waves are propagated through the gas from the 
j)oint of ignition, the flame might lag behind the sound-waves at first and catch them 
after a run more or less prolonged. The sound-waves when overtaken might cause 
reflected sound-waves (made visible in tlie luminous gases), and the explosion itself 
might become more intense owing to the collision."^ Many experiments were under¬ 
taken to verify or disprove this hypothesis, but it will not be necessary to reproduce 
the large number of photographs we took in the course of this investigtition. ATe 
found that no addition to the length of the lead pqoe between the firing point and the 
bifurcation, and no addition to the length of the two arms affected the result. Finally, 
it was found that the explosion was affected as it passed through the junctions between 
the lead and the glass, and the impulses recorded in our j^hotographs were due to the 
detonation-wave, damped down at the junction, being regenerated by fits and starts. 
One of the photographs showing this most plainly is reproduced in fig. 35. This was 
obtained in the following way:—Three straight glass tubes were fixed horizontally 
one above another and were joined together by curved lead tubes, into which the 
* Le Chatelier has published a similar hypothesis to explain the apparent break in the curve -Nvhere 
the detonation is started; see p. 347. 
