...rigid.

Although this behaviour has been referred to as `surface melting' in some studies[109,110,111] we prefer not to use this terminology because the behaviour does not evolve with increasing size to the surface melting familiar for bulk surfaces[160].

.
.
.

...s.

In their formulation of the superposition method Hoare and McInnes missed out the effect of symmetry[147]. This explains why in the brief mention of the results that they obtained, the method predicted that for 13 the cluster would reside in the icosahedral global minimum up to a reduced temperature of 0.6. (13 in fact melts at $\sim$0.29). Without the inclusion of symmetry, the density of the states of the icosahedron is overestimated by 120 times.

.
.
.
.

...method

The 75-atom Marks decahedron, though, has been found by a global optimization method developed by David Wales where a decahedral seed is introduced at the centre of the cluster.

.
.
.
.

...PES.

For the distributions and averages of b_d, b_u and $\Delta E$ the degenerate rearrangements (that join permutational isomers of the same minimum) are excluded from the sample because they contain no information about structural relaxation, and simply obscure the changes to the topology of the PES. The fraction of transition states that mediate degenerate rearrangements varies from 38% for $\rho_0$=3 to 1% for $\rho_0$=14.

.
.
.
.

...minimum.

This result is based upon an analysis of the 13 PES similar to that performed for 13 in §6.3. Using n_ev=15, 1467 minima and $12\,435$ transition states were found.

.
.
.
.

...feasible.

For this purpose, the points files of all the global minima will soon be made available on the world wide web at URL <http://brian.ch.cam.ac.uk/ jon/jon.html>

.
.
.
.