Molecular Simulations of Retention in Chromatographic Systems: Use of Biased Monte Carlo Techniques to Access Multiple Time and Length Scales - Multiscale Molecular Methods in Applied Chemistry - page 192

surface shown in Fig. 3 , which shows little dependence of solvent composition.
Therefore, retention loss when highly aqueous solvents are used is more likely due
to pore dewetting as opposed to chain collapse.
The end-to-end vector discussed above provides a picture of the overall align-
ment of the chains, but more local information on individual segments within the
chain can be gleaned from the orientational order parameter S i along the chain
backbone:

;

1
2

3cos 2

S i ¼ y i 1 (3)

where y i is the angle between the i th 1-3 backbone vector in the C 18 chain and the

normal to the silica surface. This order parameter is equivalent to the experimen-
tally observable NMR order parameter for deuterated alkyl chains [ 59 , 60 ]. Figure 4
shows this order parameter for each 1-3 vector along the chain backbone and
Table 1 gives S n , the value of the order parameter averaged over all 16 1-3
backbone vectors. Looking at the order parameter along the chain backbone one
sees a similar trend for all four solvent systems. The order parameter is large and
positive for the first few backbone vectors and reaches a minimum somewhere near
vector number 10, and then approaches zero beyond this minimum. Thus, the initial
portion of the chain shows a significant preference to align itself away from the
silica surface while the terminal portion of the chain is oriented more randomly.
Despite the similarity in shape, the curves are shifted upward as acetonitrile
concentration is increased, again indicating an increase in chain alignment.
Now that structural characteristics of the RPLC system have been discussed, a
discussion of the retention mechanism can begin. To describe the mechanism of
analyte retention in RPLC one needs to know, with high resolution, the preferred
locations and orientations of the analyte molecule within the stationary phase. The
simulations described here are able to yield this type of data directly. The preferred
locations of the analytes are described through the z -dependent distribution coeffi-
cient profiles, or K ( z ) plots, shown in Fig. 5 for n -butane and 1-propanol. These
profiles are analogous to the (experimentally measurable) distribution coefficient
for transfer from mobile to stationary phase but offer much more detailed informa-
tion on where retention occurs within the stationary phase. Larger values of K ( z )
correspond to more favorable (lower free energy) locations of the analyte within the
stationary phase. In examining these profiles, one of the most striking features is the
large dependence of the analyte distribution coefficient on z. From this, it is clearly
evident that the stationary phase is not a homogeneous medium into which analytes
partition nor a nonpolar surface to which analytes adsorb. Rather, the stationary
phase is a microheterogeneous medium with multiple preferred regions for the
analytes.
For n -butane the K ( z ) profiles show a bimodal distribution in all solvent systems.
There is one peak in the center of the bonded phase ( z 8 ˚ ) and another in the

interfacial region. The peak in the center of the bonded phase remains rather sharp
regardless of solvent composition. However, the shape of the peak in the interfacial