WITHDRAWN FROM THE ACTION OF GRAVITY. 671 



but the latter does not amount to one second, and there cannot 

 be any doubt that it would be going too far to extend the com- 

 plete duration, and a fortiori, the portion which terminates at the 

 rupture of the lines, to two seconds. Thus in the case we have 

 just considered, the sum of the length of a constriction and a 

 dilatation becoming about four and a half times greater, the 

 partial duration becomes at least four times less. 



67- But if, in reckoning the absolute duration in the case of 

 one of our short cylinders of oil, we only obtain with regard to 

 this liquid one upper limit, and this much too high, the cylinder 

 of mercury in § 55 (which cylinder is formed in the air, and the 

 length of which in proportion to the diameter is sufficient for the 

 divisions to have assumed exactly, or very neax'ly, their normal 

 length), will furnish us, on the contrary, in regard to this latter 

 liquid, with a limit which is probably more approximative and 

 which will be very useful to us. 



First, in the case of this cylinder, the diameter of which, as 

 we have said, was 2\ millims., the transformation does not take 

 place in a sufficiently short time for us to estimate with any ex- 

 actitude the total duration of the phsenomenon ; I say the total 

 duration, because in so rapid a transformation it would be very 

 difficult to determine the instant at which the rupture of the lines 

 occurs. To approximate as closely as possible to the value of 

 this total duration, I have had recourse to the following process. 



By successive trials, I regulated the beats of a metronome 

 in such a manner, that on rapidly raising, at the exact instant at 

 which a beat occurs, the system of glass strips belonging to 

 the apparatus serving to form the cylinder (§ 50 and 51), the suc- 

 ceeding beat appeared to me to coincide with the termination 

 of the transformation ; then having satislied myself several times 

 that this coincidence appeai-ed very exact, I determined the 

 duration of the interval between two beats, by counting the 

 oscillations made by the instrument during two minutes, and 

 dividing this time by the number of oscillations. I thus found 

 the value 0"*39 for the interval in question. The total duration 

 of the transformation of our cylinder of mercury may therefore 

 be valued approximatively at 0"'39, or more simply, at 0"*4. 



But the entire convex surface of this cylinder is not free, and 

 its contact with the plate of glass must exert an influence upon 

 its duration, both directly as well as by the increase which it 



2 z 2 



