Interstellar grains
are composed by a rocky core (usually amorphous
silicates) covered by an icy mantle, the most abundant molecule being
H
2
O followed by CO, CO
2
, NH
3
, and
also radicals in minor quantities. In dense molecular clouds, gas-phase
chemical species freeze onto the grain surface, making it an important
reservoir of molecular diversity/complexity whose evolution leads
to interstellar complex organic molecules (iCOMs). Many different
models of water clusters have appeared in the literature, but without
a systematic study on the properties of the grain (such as the H-bonds
features, the oxygen radial distribution function, the dangling species
present on the mantle surface, the surface electrostatic potential,
etc.). In this work, we present a computer procedure (ACO-FROST) grounded
on the newly developed semiempirical GFN2 tight-binding quantum mechanical
method and the GFN-FF force field method to build-up structures of
amorphous ice of large size. These methods show a very favorable accuracy/cost
ratio as they are ideally designed to take noncovalent interactions
into account. ACO-FROST program can be tuned to build grains of different
composition mimicking dirty icy grains. These icy grain models allow
studying the adsorption features (structure, binding energy, vibrational
frequencies, etc.) of relevant species on a large variety of adsorption
sites so to obtain a statistically meaningful distribution of the
physicochemical properties of interest to be transferred in numerical
models. As a test case, we computed the binding energy of ammonia
adsorbed at the different sites of the icy grain surface, showing
a broad distribution not easily accounted for by other more size limited
icy grain models. Our method is also the base for further refinements,
adopting the present grain in a more rigorous QM:MM treatment, capable
of giving binding energies within the chemical accuracy.