Chemistry Reference
In-Depth Information
as discussed in former sections of this chapter. One may wonder, what is then the real
value of such computational exercises?
To be sure, fundamental physical principles should be kept as a guideline of all
simulations as much as possible; all detail of a simulation must be thoroughly
described and carefully specified, in particular with respect to possible violations of
such principles and their consequences. Old theories should not, however, be
considered as gospel just because they have been around for a long time - especially
classical nucleation theory, a construction based on macroscopic equilibrium quan-
tities mostly without any atomic detail, or many aspects of formal statistical mech-
anics, hardly applicable to multiform real chemical systems in condensed phases.
The force fields must at least be validated by comparison with some experimen-
tal data. Results must not be too dependent on small details in the overall formu-
lation. Of particular concern is reproducibility: ideally, the simulation code should
be deposited and made available to the community for independent verification
(a request that is rarely met in publications, in part due to protection of intellectual
property, but also because many codes are in-house packages undergoing con-
tinuing revision and updating, well known only to its authors and lacking a proper
documentation). However, once these basic requirements are met, it should always
be kept in mind that a simulation is not necessarily reality: the result is what
might
happen under the specified computational conditions, and not necessary what
does
happen. A properly validated simulation should not be used just to reproduce other
experimental data: that would be at best a welcome, further validation, but not real
research. The true value of a simulation is the production of new data or new
insights, in the form of molecular trajectories, suggesting
possible
scenarios for
chemical processes at the molecular level.
A related issue concerns the representation of the system. Undoubtedly, more
accurate simulations can be (and are) conducted on the so-called Lennard-Jones
fluid, an ensemble of hard- or soft-sphere pseudo-molecules, for which the connec-
tions to orthodox statistical mechanics are more strict. Although in this way one
may correctly reproduce the phase diagram of, say, argon, this would hardly be
considered as significant progress in chemistry. Computing resources are perhaps
better used in attempting an approximate sketch of the evolution of real molecules
than in super-accurate simulations on unrealistic systems, even at the price of
entering a realm where statistical mechanics loses much of its formal strength.
In this perspective, “realistic” is not a proper adjective for a simulation: “infor-
mative” is probably more appropriate. For example, the simulation of the early
stages of crystallization from the melt for
n
-hexane described above is properly
validated (the force field correctly describes the density of the liquid as a function
of temperature, the distribution of
gauche-trans
isomerism in the liquid, and
the sublimation enthalpy of the crystal). It deals with a real molecule, including
the torsional rearrangement of the chain. And yet, if a “realistic” simulation of the
crystallization process of
n
-hexane is required, the computing time needed would
be of the order of years, if at all possible. The simulation can, however, be
considered informative, and hence useful, in that it provides a working hypothesis,
a new insight that might stimulate further thought and perhaps even the planning of
