In the work of Formoso et al. [
52], molecular dynamics calculations were performed using certain simulation constrains, such as the size of the molecular model or the simulation time, which could be too restrictive. In this study, with the aim of using more realistic conditions, different simulation parameters have been evaluated to see their influence in the non-covalent interactions among chains. In particular, we have tested the following: (i) the initial molecular guess, (ii) the length of the chains, (iii) the size of the simulation box and (iv) the simulation time. In order to do such analysis, the R
-PD-SS and R
-PD-SS (urea- and urethane-containing diphenyl disulfides) molecular systems have been chosen. For the validation, we make use of the number of hydrogen bonds and their impact in
. This parameter is based on the distance between dichalcogenides and is defined as [
52]:
Using the radial distribution function of the S or Se atoms, three regions may be defined to calculate : the reacting region (), where dichalcogenides are close enough to undergo the exchange reaction (R ≤ 4.5 Å), the neighboring region (), where dichalcogenides are far to react but with a non-negligible probability to approach the reacting region (4.5 < R < 20 Å), and the external region (), where dichalcogenides are neglected, R > 20 Å. Note that the external region is not included in the definition of in order to have a size-independent parameter. The amount of chalcogen atoms located in each region is calculated by integration of the radial distribution function within the limits defined above. These limits are defined for disulfide compounds and will be different for diselenides, as it will be discussed hereafter.
2.2.1. Initial Guess and Chain Length
In the previous work, the model comprised a single chain formed by two phenyl disulfide monomers (D
and D
units in
Figure 2, top) linked by a urea-methyl-urea or urethane-methyl-urethane unit (C
in the same Figure). This chain was then replicated in space in order to generate a 3D structure. For that, four replicas of the chain were placed along two spatial directions using translation vectors. The resulting initial conformation, before the equilibration step, is depicted in
Figure 2 (bottom left). It can be observed that the use of translation vectors yields a laminar-like periodic structure that may introduce a structural bias. In this work, we have considered a random initial guess (
Figure 2, bottom right) in order to have a more realistic description of a complex polymeric system. Besides, the chain length has also been increased by including one additional –(CH
)– group between both disulfides and the polyurea chain (C
unit).
The results of the simulations for the two initial structures (periodic and random) and the two molecular chains (with and without the additional methyl group, labeled as long and short, respectively) are collected in
Table 1, where the total number of hydrogen bonds (HB
) and the maximum (HB
), minimum (HB
) and average (HB
) number of hydrogen bonds at a given step of the simulation are provided, together with
. Besides, two simulation times (10 and 30 ns) are considered, that will be discussed later.
Considering the initial guess for the short chain, the total number of hydrogen bonds is similar for both systems but the distribution is different. See
Table 1 and also
Figure 3 (top), where the normalized percentage of hydrogen bonds is represented. In the periodic structure, the hydrogen bonds are mainly formed between the same components of different chains, that is, both the disulfide units and the linking chain tend to form HBs with their counterparts in the surrounding chains (D
-D
, D
-D
and C
-C
), see
Figure 2, bottom left. This introduces a bias that will affect the equilibration step and, therefore, the rest of the simulations. On the other hand, in the random conformation, each unit can establish HBs with any other unit of the surrounding chains that is close enough, and may be considered as a more realistic situation. This change in the distribution is responsible of the observed decrease in
for the random conformation. In the periodic model, disulfide bonds are located periodically along the system, favoring the D
-D
and D
-D
interactions and maximizing the value of
. On the other hand, the interactions between disulfides laying on different sides of the chain (D
-D
) will be more scarce, contributing to a lowering of this parameter. In the random conformation, however, an opposite situation may take place. Since the disulfides will not necessarily have another disulfide nearby, the probability of the D
-D
interaction is lower and, hence,
will also be smaller. Nevertheless, the probability of the D
-D
interaction is higher, contributing to enlarge
. The net effect, as it is observed in
Table 1, is that
is lower for the random conformation. This effect is found to be independent of the length of the chain and the simulation time. This means that the use of a periodic model induces an arbitrary preference in the interaction between adjacent disulfides (D
-D
) and
may be overestimated.
If a long chain is considered (with an extra –(CH
)– group), notable differences are found between the periodic and the random systems regarding the hydrogen bonds. In both cases, a decrease of the total number of hydrogen bonds is observed, but it is particularly pronounced for the random system (from 730,119 to 595,587, after 10 ns). A larger chain unit (C
) means a separation of the groups that can form HBs, both intermolecular and intramolecular, reducing the possibility of interactions and, therefore, the total (HB
), maximum (HB
) and average (HB
) number of hydrogen bonds are decreased. In the periodic system, this effect is small, since the laminar structure is kept and the starting HBs (from the D
-D
and C
-C
interactions) remain. The observed reduction (from 746,461 to 707,860, after 10 ns) may be ascribed to the changes in the geometry of the chains induced by the new methyl group. In fact, this effect is largely mitigated in the longer (30 ns) simulation of the periodic structure, since a longer simulation time allows the structure to reorganize and recover the lost HBs. The differences between the periodic and random structures are clearly observed in
Figure 3 (bottom), in the normalized percentage of hydrogen bonds after 10 ns (left) and 30 ns (right) of simulation time.
Comparing the radius of gyration of the short chains (
Figure 4, left), the periodic system oscillates close to 14 Å, while the one corresponding to the random system appears around 12.5 Å. This suggests that the periodic structure somewhat retains the laminar-like structure after the equilibration, while the random structure acquires a more coiled conformation. When the system is built with long chains, the same feature is observed (
Figure 4, right), but with larger values.
As a summary, we can conclude that the periodic starting structure is less suitable to perform the simulations, since its periodicity introduces a bias that may affect to the estimation of .
2.2.2. Size of the Simulation Box
Another important parameter is the size and shape of the simulation box, which determines the density of the system. In the simulations of Ref. [
52], a square prism of 40 × 30 × 30 nm was used and now it has been compared to a cubic box of two different sizes: 40 × 40 × 40 nm and 50 × 50 × 50 nm. The results are collected in
Table 2 and
Figure 5, where the radial distribution function of each sulfur with respect to the others is represented, excluding their counterparts in the disulfide S-S bond. Despite the number of hydrogen bonds remains basically unchanged, the shape and the size of the box have a remarkable impact on the radial distribution function and, thus, in
. This suggests that there is a frontier effect that may be avoided using larger and regular boxes.
In conclusion, having all these considerations in mind, a random conformation with long chain have been selected to carry out the molecular dynamics simulations, using a cubic simulation box of 50 × 50 × 50 nm, with a simulation time of 30 ns for each system, given that the longer the simulation time allows the exploration of a bigger number of conformations.