A physically based mean-field theory of criticality and phase separati
on in the restricted primitive model of an electrolyte (hard spheres o
f diameter a carrying charges +/- q) is developed on the basis of the
Debye-Huckel (DH) approach. Simple DH theory yields a critical point a
t T = k(B)Ta/q(2) = 1/16, which is only about 15% above the best rece
nt simulation estimates (T-c,T-sim = 0.052-0.056) but a critical dens
ity rho(c) = rho(c)a(3) = 1/64 pi similar or equal to 0.005 that is m
uch too small (rho(c,sim) = 0.023-0.035). Allowing for hard-core excl
usion effects reduces these values slightly. However, correction of th
e DH linearization of the Poisson-Boltzmann equation by including pair
ing of + and - charges improves rho(c) significantly. Bjerrum's theor
y of the (required) association constant is revisited critically, Ebel
ing's reformulation is strongly endorsed but makes negligible numerica
l difference at criticality and below, The nature and size of the asso
ciated, dipolar ion pairs is examined quantitatively and their solvati
on free-energy in the residual fluid of free ions is calculated on the
basis of DH theory. This contribution to the total free energy proves
crucial and leads to a rather satisfactory description of the critica
l region. The temperature variation of the vapor pressure and of the d
ensity of neutral dipolar pairs correlates fairly well with Gillan's n
umerical cluster analysis. Possible improvements to allow for larger i
on clusters and to better represent the denser ionic liquid below crit
icality are discussed. Finally, the replacement of the DH approximatio
n for the ionic free energy by the mean spherical approximation is stu
died. Reasonable critical densities are generated but the MSA critical
temperatures are all 40-50% too high, in addition, the predicted dens
ity of neutral clusters seems much too low near criticality and, along
with the vapor pressure, appears to decrease too rapidly by an expone
ntial factor below T-c.