The bulk melting of solids is well understood thermodynamically. However, thermodynamics, being a macroscopic theory, does not reveal the microscopic behavior or the mechanism of the melting transition, i.e., how the liquid nucleates at the melting point. A first order phase transition like melting cannot transform continuously from the solid to the liquid but must do so discontinuously. It is expected that melting will be an inhomogeneous nucleation process where the nucleation starts about particular sites and grows to encompass the whole sample. Molecular dynamic simulations[1] have shown that grain boundaries and surfaces can act as nucleation sites in agreement with experiments[2,3,4]. However in the thermodynamic limit where the sample dimensions become extremely large the effects of such boundaries must become negligible and a nucleation site that scales with the volume is required. The most likely fluctuations that can lead to nucleation sites are those that are short ranged (of the order of interatomic distances). These scale correctly with the volume. We have found experimental evidence[5,6] that around some impurities there exists already below the melting point a somewhat more disordered region which one expects would have a smaller barrier to fluctuate to a nucleation site than the undisturbed bulk.
To experimentally investigate such nucleation sites of melting
requires a short ranged probe. Thus, diffraction techniques which are
mostly sensitive to long range order are not the best probes to
investigate the nucleation mechanism. The short range probes that we
employed to investigate the nucleation of melting mechanisms were the
Mössbauer effect (ME)[5] and the x-ray absorption fine
structure technique (XAFS)[6]. The criterion usually used by
physicists to distinguish between a solid and its liquid is the
presence of long range order in the former and its disappearance in
the latter. The short range probes are not sensitive to long range
order and a different criterion must be employed to distinguish
between a solid and a liquid for these techniques. Macroscopically,
the most obvious difference between a solid and its liquid is the
response under shear stresses. Liquids have no resistance to static
shears, and, at the melting point where they coexist, the viscosity of
the liquid is many orders of magnitude less than the solid. By the
Stokes-Einstein relation this difference in viscosity is due to a
typically 10
increase in the diffusion rate in the liquid relative
to the solid. A second striking macroscopic difference is the
increased entropy in the liquid as revealed by the latent heat of
melting.
Both these features can be detected by a local probe. In the case of
the ME the increased hopping rate of the liquid is typically so great
that the spectral intensity becomes negligible. In the case of XAFS,
as shown for liquid Pb[7], the increased hopping rate
decreases the signal by the fraction of the time an atom is hopping
from the site about which it vibrates. In liquid Pb, an atom spends
about a half of its time vibrating and the rest
diffusing[7]. The mechanism of diffusion in a liquid is
quite different than in a solid because it is a high entropy process
involving many atoms flowing around the diffusing atom resulting in
the 10 increase in its rate. The local probe can distinguish a
liquid like region by either detecting an enhanced hopping and/or an
increased entropy density.
In the case of a 1 at% Sn alloy in Pb, the ME showed[5]
that the spectral intensity of the Sn atoms drops precipitously at
around 150K, about 1/4 of the melting temperature. This rapid drop is
too steep to be explained by anharmonicity of the vibration of the Sn
atoms. It is caused by an anomalously rapid hopping of the Sn atoms
about a lattice site within a ``bubble" containing a large entropy
density involving a coherent motion of 
20--40 surrounding
atoms. XAFS measurements[6] also indicated premelting
phenomena about the Hg atoms in a 4 at% alloy of Hg in Pb starting
around 400K. Again an anomalously rapid hopping of the Hg atoms
localized about its original lattice site is indicated by these
measurements.
In this paper we will present more detailed measurements in these systems which will relate to concerns expressed by J. Mullen[8], and Martin and Singer[9] about the original interpretation of the results .We also present some results on impurities in a Ag host to ascertain the generality of the results observed in the Pb host. The outlineof the paper is as follows. The next Section II gives the experimental results. A discussion and interpretation of the experiments are given in Section III. A summary and conclusion are given in Section IV.