dissociation. This
would account for an increase of S, and a diminution of the ratio S/s,
with rise of temperature which apparently occurs in many vapours. The
experimental evidence, however, is somewhat conflicting, and further
investigations are very desirable on the variation of specific heat
with temperature. Given the specific heat as a function of the
temperature, its variation with pressure may be determined from the
characteristic equation of the gas. The direct methods of measuring
the ratio S/s, by the velocity of sound and by adiabatic expansion,
are sufficiently described in many text-books.

S 19. _Atomic and Molecular Heats._--The ideal atomic heat is the
thermal capacity of a gramme-atom in the ideal state of monatomic gas
at constant volume. This would be nearly three calories. For a
diatomic gas, the molecular heat would be nearly five calories, or the
atomic heat of a gas in the diatomic state would be 2.5. Estimated at
constant pressure the atomic heat would be 3.5. Some authors adopt 2.5
and some 3.5 for the ideal atomic heat. The atomic heat of a metal in
the solid state is in most cases larger than six calories at ordinary
temperatures. Considering the wide variations in the physical
condition and melting points, the comparatively close agreement of the
atomic heats of the metals at ordinary temperatures, known as Dulong
and Petit's Law, is very remarkable. The specific heats as a rule
increase with rise of temperature, in some cases, e.g. iron and
nickel, very rapidly. According to W.A. Tilden (_Phil. Trans._, 1900),
the atomic heats of pure nickel and cobalt, as determined from
experiments at the boiling-points of O2, and CO2, diminish so rapidly
at temperatures below 0 deg. C. as to suggest that they would reach
the value 2.42 at the absolute zero. This is the value of the minimum
of atomic heat calculated by Perry from diatomic hydrogen, but the
observations themselves might be equally well represented by taking
the imaginary limit 3, since the quantity actually observed is the
mean specific heat between 0 deg. and -182.5 deg. C. Subsequent
experiments on other metals at low temperatures did not indicate a
similar diminution of specific heat, so that it may be doubted whether
the atomic heats really approach the ideal value at the absolute zero.
No doubt there must be approximate relations between the atomic and
molecular