We present a new
method for computing chemical potential differences
of macroscopic systems by sampling fluctuations in small systems.
The small system method, presented by Schnell et al. [Schnell et al.,
J. Phys. Chem. B, 2011,
115
, 10911], is used to create
small embedded systems from molecular dynamics simulations, in which
fluctuations of the number of particles are sampled. The sampled fluctuations
represent the Boltzmann distributed probability of the number of particles.
The overlapping region of two such distributions, sampled from two
different systems, is used to compute their chemical potential difference.
Since the thermodynamics of small systems is known to deviate from
the classical thermodynamic description, the particle distributions
will deviate from the macroscopic behavior as well. We show how this
can be utilized to calculate the size dependence of chemical potential
differences and eventually extract the chemical potential difference
in the thermodynamic limit. The macroscopic chemical potential difference
is determined with a relative error of 3% in systems containing particles
that interact through the truncated and shifted Lennard-Jones potential.
In addition to computing chemical potential differences in the macroscopic
limit directly from molecular dynamics simulation, the new method
provides insights into the size dependency that is introduced to intensive
properties in small systems.