318 Intelligence and Miscellaneous Articles. 
equal to 3s as has been pointed out by Simon: this relation is nearly 
2 for the chloride of iron and 1 for our other liquids. 
3. The wide tubes give a value which is comprised between the 
two preceding values when the latter differ, and equal to them when 
they coincide; this is what takes place with alcohol, and it is for 
this reason that the only experiment of verification cited by Laplace 
and Poisson gave a result which accords exactly with the formula: 
it would not have been so if, in this experiment, Gay-Lussac had 
made use of water instead of alcohol. It will be seen also why 
Frankenheim* found that experiment disagreed with the formula, 
even when he employed tubes of an internal diameter of 14 millims. 
4. In proportion as the radii of convex cylinders diminish in de- 
parting from the plane where this radius is infinitely great, the volume 
raised continues diminishing for the first two liquids ; with the others 
this diminution commences at a certain limit of curvature, and in- 
creases gradually and apparently indefinitely. Amongst the liquids 
which I have tried, ether is that which presents the most constant 
volume; but unfortunately the results relating thereto are less cer- 
tain than the others, notwithstanding the precautions which I took 
to diminish the evaporation during the experiment. And, in any 
case, it is not to the absence of viscosity that this constancy would 
have to be attributed; comparative experiments with pure water and 
gum-water having shown that viscosity, although retarding the 
movement when the equilibrium is established, has no sensible in- 
fluence upon this definitive state. 
To explain these facts, one might be tempted to admit that the 
angle of contingence varies with the curvature of the wall; but it 
may be demonstrated that this is not the case. Thus, if we consider 
only the menisci of water and chloride of iron raised by one plane, 
for which we should haye already ¢ < 90°, the area of the section 
would be 
a Cs sin ¢, 
pan 
and the maximum ordinate 
H=av 2 sing >a sin ?; 
whilst experiment coUstantly gives 
V2A>H. 
I also show that for ¢=90° we have the coordinates of the centre 
of gravity, 
v=2H, y=2H—~2 y~ 019595 H; 
— 3 ? y 3 3 => Jv 5 
in proportion as ¢ diminishes, the centre of gravity removes from 
the axis of the ordinates, whilst it is in reality more approximated 
to this than it would be according to this formula. 
* Poggendortfi’s Annalen, vol. Ixxii. p. 191, 
