2. Methods and Simulation Details
A simulated non-ionic amphiphilic AB diblock copolymer chain consists of a solvophilic (soluble) block (A) and a solvophobic (insoluble) block (B), and the chain length is fixed at
. The length of the solvophobic block,
, varies from 16 to 40, and the length of block A ,
, varies accordingly from 48 to 24. All chains are simulated on a cubic lattice. Each lattice site is occupied by either a solvent molecule (S) or by a monomer (A or B), except the sites excluded by the pore wall. Note that this exclusion has nothing in common with the entropic depletion effect, i.e., with the decrease in concentration of polymer chains in the vicinity of impermeable inert walls due to the reduced number of possible chain conformations close to the wall [
31]. This entropic effect plays a non-negligible role in the studied systems and appreciably lowers the concentration of polymer beads close to the inert wall, but as the cylindrical wall generally passes between the lattice points, the distance of some lattice points from the wall is
, and these positions are excluded by the definition of the interacting potential (Equation (
5)). We apply the single-site bond fluctuation model, which generates 26 possible bonds (connecting the adjacent monomers) fluctuating between 1 and
by the permutation of the vectors of (±1, 0, 0), (±1, ±1, 0) and (±1, ±1, ±1) [
51]. Note that neither the overlap of polymer beads (when two or more monomers occupy the same lattice site) nor the intersection of bonds is permitted [
52,
53]. A bead (monomer or solvent) interacts with its neighbors in the range of distance,
, from 1 to
, and the pairwise potential energy,
, is defined as
where
and
stand for the solvent (S) or the monomers (A and B), and
represents the interaction strength. The bulk mobile phase is modeled as an unconfined cube with the dimensions of
(
, in lattice units), and the pore (stationary phase) is modeled as an impermeable cylindrical tube with a variable diameter,
. The schematic representation of the employed bulk-pore model is shown in
Figure S1 in the Supplementary Material. For a given pore diameter,
, we change
to keep the volume ratio
. For both the bulk and the pore, periodic boundary conditions (PBC) are imposed in all directions except the impermeable shell of the cylindrical pore. The pore is filled with solvent molecules, and the copolymer chains entering the pore interact with the wall, i.e., with the inner surface of the pore. The bead–wall interaction,
, is defined as
Here, the subscript represents A or B or S; the subscript W stands for the pore wall; and is the normal distance from the bead to the concave inner surface. Note that and stand for an inert wall and for an attractive wall interacting with species , respectively. This employed square potential implies that neither monomers nor solvent molecules can occupy the lattice sites of . In this simulation study, the values of all interaction parameters are expressed in the energy units, , where is the Boltzmann constant and stands for the temperature of reference state.
We use the modified dynamic configuration-bias Monte Carlo (CBMC) algorithm based on the method originally developed by Siepmann and Frenkel [
54]. The original CBMC employs the implicit solvent, which means that each unoccupied lattice position implicitly contains a solvent bead. Simultaneously, all interaction parameters with solvent are by definition zero, i.e.,
, and the nonzero interactions between polymer beads,
(
), indirectly model the solvent quality [
55,
56,
57,
58]. The choice of zero interaction parameters with solvent simplifies the acceptance criterion due to the fact that the transition probability between two states can be expressed as the ratio of the Boltzmann factor and the Rosenbluth weight reflecting the interaction energy [
54].
The CBMC method has to be modified when applied to copolymers in selective solvents. As described in our earlier papers [
59,
60], we use the variant with tailored weights, reflecting the fact that the interactions of beads A and B with the solvent differ. In this case, the transition probability contains the Boltzmann factor of the generated polymer conformation, which reflects the energy difference between the trial and original self-avoiding walk (SAW), including the energy contribution caused by the exchange of polymer beads and solvent molecules (see
Figure 1).
A detailed explanation of the weighting factors used together with the justification of the modified acceptance criterion can be found in the
Supplementary Material. The application of the modified CBMC is formally analogical to that of the original method. We express the modified weights by the energy difference parameter
, where S denotes the solvent and
,
denote A, B, and W (pore wall). Then, the energy difference between the two systems differing in the bead type in a given lattice position
i can be expressed as the sum of all
of its neighbors (as exemplified in
Figure 1), i.e.,
.
Contrary to the original CBMC method proposed by Siepmann and Frenkel [
54], which models the solvent quality indirectly by one parameter only, our modified approach requires the setting of several interaction parameters,
, which are neither simply related to parameters used in the original CBMC nor to the Flory–Huggins interaction parameter [
61]. They are loosely related to the Lennard–Jones (LJ) interaction parameters (to both homo- and cross-interactions) [
62]. Nevertheless, as our model does not include the explicit solvent–solvent interactions, i.e.,
, the used
are not directly proportional to the LJ parameters, which describe the interactions in vacuum and are used in explicit solvent model systems. The parameters used in the modified CBMC have to be estimated independently on the basis of mapping the data onto the experimental association behavior. Note that
still describes a good solvent condition analogously to the original CBMC calculation, but the value
already models a poor solvent and hence the combination of the two values above, i.e.,
and
, provides the selective solvent. The low value
reflects the fact that in the used variant of single-site bond fluctuation model, all 26 neighbors (up to the distance
, see Equation (
4)) interact equally with the segment/solvent in a given lattice position, and the effect of interactions is therefore more than 4 times stronger than that in common CBMC simulations on the simple cubic lattice, where the number of interacting neighbors is only 6. In this paper, we do not emulate the behavior of any particular copolymer-solvent system but focus on the general behavior of self-assembling copolymers in selective solvents. Because the simulations in bulk yield a reversible equilibrium of well-defined core-shell micelles with unimer chains, i.e., the study faithfully emulates the behavior of real micellizing systems and fulfills the basic prerequisite of the working hypothesis, we are of the opinion that the interaction parameters have been set appropriately.
As previously mentioned, for the simulated amphiphilic diblock copolymers, we set
to model a good solvent condition for block A and
to describe the insolubility of block B. We set
to describe the interaction between the incompatible A and B monomers. The parameters of other pairwise interactions are
. For the polymer–wall interaction, we use
to model the SEC mode, and we use negative values to describe the attractive interaction of polymer beads to the inner surface of the cylindrical pore. Besides the local move, i.e., the deletion and regrowth of chains in the same box, we employ the swap of polymer chains between two boxes (akin to the Gibbs ensemble Monte Carlo simulation [
63]) to enable their partitioning between two phases. We always performed at least
CBMC steps consisting of molecule swaps and local moves with the ratio of 1:1 for each simulation run, except for some special cases when the simulations needed to be prolonged. A simulation trajectory was divided into 20 blocks to estimate the standard deviations of computed quantities. Additionally, we tested each simulation starting from several different initial configurations to confirm the reliability of the results.
We assume that two copolymers belong to the same micelle if two insoluble beads (B) from different chains are at neighboring lattice sites. This association criterion was also proposed and used by other authors [
58,
64,
65,
66,
67].
In addition to studies of the bulk-pore partitioning, we performed several simulations separately in the bulk and in the pore to calculate the excess chemical potential of diblock copolymers,
where
is temperature and the constant
represents the total number of interacting neighbors in the model used (
). Note that the ensemble average of the Rosenbluth weight of the generated ghost chain,
, is normalized by the ideal gas part,
. We performed up to
CBMC steps for these individual runs and computed
every
steps by generating
ghost chains.
In this study, the effective volume of the pore,
, is represented by the number of lattice sites in the pore that can be occupied by polymer or by solvent beads, i.e., the lattice sites of
(
is the distance from the wall). Obviously, the effective volume of the bulk can be expressed as
. To discuss the partitioning of a multiphase system encompassing the equilibria described by Equations (
1)–(
3), we define the effective partition coefficient of copolymers,
, as
where
and
represent the total equilibrium concentrations of copolymer beads in the pore and the bulk, respectively;
and
represent the numbers of copolymer chains in the pore and the bulk, respectively; and the total chain length
. If the association of copolymers does not occur in either phase, apparently,
, where
stands for the partition coefficient of non-associated chains (see Equation (
2)). For brevity, in this article, we often use the short term “partition coefficient” for the effective partition coefficient,
. The majority of the simulations has been performed at the temperature
. At end of the paper, when we show and discuss the results of simulations in the temperature range
to 1.8; the actual value is always given at the pertinent place.
3. Results and Discussion
First, we studied the partitioning of symmetric A32B32 diblock copolymers in moderately narrow pores with inert walls, of which the diameters,
, range from 15 to 30. The parameters were used emulate the conditions of the SEC regime under mild confinements because the coil-to-pore size ratio,
, ranges from 0.36 to 0.72 (the simulated radius of gyration,
, of the single chain in bulk is 5.42). The simulations started with all the chains in the bulk corresponding to the injection of the polymer solution into the column in practical chromatography.
Figure 2a depicts the variation of the effective partition coefficient,
, with the total concentration of beads,
, ranging from
to
. We first describe the data for
. In the dilute regime for
,
is approximately 0.2 and almost independent of
(decreases only negligibly with
), indicating that the effective solvent quality for the whole copolymer is slightly worse than the
-solvent condition [
34]. However, when the concentration is higher than
,
drops steeply to 0.07 and then decreases slowly with increasing
. The trends of
vs.
are similar for all studied
. The values of
increase with increasing
for all concentrations in accordance with the basic feature of SEC [
68]. For the widest pore of
,
starts at 0.68 (for
) and decreases slightly with increasing
. As in other pore diameters, a sudden abrupt drop to 0.24 occurs when
reaches approximately
, and afterwards,
decreases slightly with increasing
. In conclusion, the concentration dependence of associating diblock copolymers exhibits a pronounced decreasing sigmoidal shape and differs from the slightly increasing or decreasing (almost linear) curves observed for non-association polymers in current solvents differing in thermodynamic quality until the saturation of pores [
34,
35,
36,
37]. We performed another set of simulations starting with all polymer chains in the pore and obtained the same results (see
Figure S3 in the Supplementary Material), which confirms (i) the ergodicity of the simulation procedure used and (ii) the data describing the equilibrium behavior.
The snapshots in
Figure 2a corroborate our hypothesis that the observed steeply changing behavior with the sudden drop in the
vs.
curves is caused by the association of A32B32 copolymers in the bulk phase. At low concentrations, all chains are well dispersed in bulk, and the bulk-pore partitioning of unimers (Equation (
2)) controls the phase equilibria. This behavior is analogous to that of non-associating systems. When the concentration exceeds the CMC, the unimers start to associate and the aggregates, i.e., the micelles with solvophobic cores and solvophilic shells, form in the mobile phase and coexist in equilibrium with unimers. The concentration of micelles increases fast with the increase in the total concentration, and at
, the phase equilibria and consequently the chromatographic behavior are strongly influenced by the micellization of copolymers in the mobile phase. The CMC value can be estimated from the intersection point of the extrapolated dotted curves: (i) the relatively flat part of the curve for low
(green) corresponding to dilute solutions containing only unimers; and (ii) the decreasing part (brown). The intersection at
in
Figure 2a indicates the first appearance of micelles, and the narrow region close to CMC corresponds to the metastable regime, in which the micelles start to form. In this region, the fluctuations grow exponentially for
[
69], and thus the error bars in
vs.
are remarkably large. The steeply decreasing part reflects the initial rise in micelle concentration, while the ratio of unimer-to-micelle concentrations is still important, and the micelles do not yet dominate the partitioning. The dominance of micelles in bulk translates in the third almost flat (only slightly decreasing) low
part at high
. Note that the value of the directly estimated critical concentration,
, differs from the CMC of the bulk solution because the
axis measures the total concentration of copolymers in two boxes (two phases) and the association takes place only in bulk. Accordingly, the true CMC of the bulk solution can be estimated as
. As the relationship between CMC and
is straightforward, for simplicity, we will use only the generic term CMC in the remaining part of the paper when discussing the features and trends of the partitioning and their impacts on chromatographic behavior. Furthermore, the snapshots of copolymer chains inside the pores of
and 30 corresponding to the partitioning at the highest total concentration, i.e.,
, are shown in
Figure 2b. No micelle is observed to form in both pores and this will be discussed later. A slightly surprising small drop on the curves at the concentration ca.
in
Figure 2a will be explained in the next part.
Using the criterion for discerning aggregates from unimers (described in the methodology section), we subsequently measured the number of free (non-associated) chains,
, in the bulk and the pore. As depicted in
Figure 3a,
in the bulk phase first increases (at the low concentrations), peaks at the point representing the CMC, then decreases slightly with increasing
. This behavior, which occurs in all pores differing in
, is a consequence of the closed association mechanism [
16] and is consistent with the light scattering data [
71,
72] and with observations concerning the “anomalous micellization” [
73]. Just below and immediately above the CMC, various temporary (irregular and strongly fluctuating) diffuse aggregates form, but with the increasing concentration, the regular micellization soon prevails over the formation of metastable aggregates. The concentration of free chains also fluctuates, and we observe an increase and peaking of unimer concentration in this metastable region, presumably as a result of (i) the chaotic and strongly fluctuating micellization dynamics and (ii) the fact that the distances between beads of different chains in loose metastable aggregates are larger than those in regular micelles, and some chains (which change their conformations rapidly) could not have been identified as parts of loose aggregates by the criterion based on a fixed distance between two pairs of insoluble beads from different chains. The dependence of
on
in the pore is analogous to that in the bulk but exhibits a strong pore size effect, i.e.,
increases with increasing pore diameter,
, for a given
because of the weakening constraint exerted by the pore (
Figure 3b).
To compare the information of bulk CMC provided by the bulk-pore partitioning simulations (or characterized by LC) with that obtained from bulk solutions only, i.e., without the bulk-pore partitioning, we plot
against the
for both the bulk solutions and the bulk phases of partitioning simulations in
Figure S6a in the Supplementary Material. Note that the concentration of beads in the bulk phase is derived from
. We also plot the partition coefficient,
, against
in
Figure S6b in the Supplementary Material. The curves in
Figure S6a,b indicate that the CMCs provided by the bulk-pore partitioning simulations are consistent with that computed from the bulk solutions, and that the properties of associating polymer solutions can be chromatographically characterized.
We believe that the second drop on the curves in
Figure 2a at the concentration ca.
and the peaking of unimer concentrations above the CMC in
Figure 3 are results of fluctuations in the metastable region where the micelles start to form. When discussing the strong effect of fluctuations in the metastable region close to the CMC on the data presented in
Figure 2 and
Figure 3, it is worth mentioning that the shapes of
vs.
curves obtained in our study are very similar to those of
vs.
curves measured by static light scattering [
71,
72], but our data, which are much more sensitive to fluctuations than the SLS results, show the second (small and relatively gradual) drop, which indicates the termination of the chaotic metastable regime. Note that the SLS data, which show the reciprocal values of the weight-average molar masses, are strongly affected by high masses of associates and almost ignore the fluctuating contribution of unimer chains. Hence, our study reveals that the transient regime spans from the CMC to ca. 2 × CMC and cannot be detected by experimental methods that monitor the number-average, weight-average, or z-average molar masses or sizes.
Further, we focus on the bulk-pore partitioning of free (non-associated) unimer chains.
Figure 4a depicts the dependences of the partition coefficient of free unimers,
, on the total concentration,
, for the systems with non-adsorbing walls, where
and
represent the concentrations of free unimers in the pore and in the bulk, respectively. The values of
are not constant even in the range of the lowest concentrations, and the positive slopes of curves grow with the increase in the pore diameter,
. The observed trend is opposite to that of a common non-associating species partitioning between the bulk and pore, of which
is almost constant at low concentrations; then, it slightly increases or decreases with
depending on the solvent quality [
35] and drops finally quite fast due to the saturation of the pore. In the studied system, the unimer concentrations in both the bulk and the pore are low and do not exceed the CMC in the whole
region, and hence the saturation effect does not come in account. The fact that
grows non-negligibly with increasing
, which means that
increases in spite of the fact that
remains constant, is slightly surprising. Nevertheless, we believe that this unexpected behavior can be explained by the following arguments: in the pores with
ranging from 15 to 30, which are relatively narrow for micelles, the appreciable confinement effect hinders the formation of multimolecular micelles. The unimer chains in pores do not associate, and
can thus exceed the bulk CMC. Because the interaction of the insoluble block with the inert pore wall is more convenient than that with the solvent, the unimer chains are slightly energetically driven into pores where the insoluble blocks concentrate close to the walls. This process (pseudo-adsorption of hydrophobic blocks on inert pore walls) is reminiscent of the formation of the unimolecular layer of surfactant molecules at the water–air interface, with the hydrophobic tails stretched towards the air to avoid the hostile aqueous medium [
74,
75].
Intuitively, one expects that the low unimer concentration in bulk should play the role of the stop-factor, preventing the penetration of chains into pores, but the micelles in bulk, which are in a reversible equilibrium with unimers, can dissociate and serve as a reservoir, providing the unimer chains for the process outlined above. Even though the surface-to-volume ratio decreases with , the behavior at the low concentrations is reminiscent of the behavior of surfactants and depends only on the surface, i.e., it increases linearly with . At the low concentrations, the copolymer chains bind to the surface analogously to surfactants that lower the water–air (or water-oil) interface tension. The maximum number of the surface-adsorbed chains is proportional to the surface and does not depend on the volume. Therefore, the slope of vs. increases with . The situation is different at high concentrations when the behavior depends on the surface-to-volume ratio, but here we discuss the values of at the extremely low concentrations.
Nonetheless, we still need to confirm whether the A32B32 diblock copolymers can aggregate in the pore at high concentrations or not. To do so, we plot the fraction of free chains,
, in both the bulk and the pore against the total concentration
, where
represents the total number of beads in the bulk or in the pore. As expected, the curves of
vs.
for the bulk phase shown in
Figure 4b almost overlap, indicating that the self-assembly in bulk is only slightly affected by the phase equilibrium with pores and by pore size. For a given
, the
of the bulk copolymers starts at 1 and slightly decreases (to ca. 0.84) with increasing
in the dilute regime, which implies that at least 84% of the bulk copolymers dissolve as single chains before the sudden drop at the CMC. At
,
remains constant, but the
of bulk copolymers decreases steadily and gradually with increasing
(see
Figure 4b) because more and more micelles are formed in the bulk. The
vs.
plot thus confirms the conclusion on self-assembly drawn from
Figure 2.
Figure 4b shows that the fractions,
, in the pores of all diameters are almost constant (close to 1) in the whole concentration region but the values for
are lower than those for
. This provides unambiguous proof that (i) the micelles formed in the bulk do not enter the pores; and (ii) the association of A32B32 copolymers in the narrowest pore is strongly sterically prohibited, but small aggregates with low association numbers can form under the moderate confinement, i.e., in the pore of
.
Using the criterion for discerning the associated from the non-associated chains, we assessed the number-average association number,
. We plot
against the bulk concentration,
, in
Figure S7 in the Supplementary Material. The shape of the plot agrees with the conclusions drawn from the concentration dependences of the partition coefficient and from the plots of
vs.
. Additionally, the plots of the number distribution,
, and weight distribution,
, of the association numbers are shown in
Figures S8 and S9, respectively, in the Supplementary Material. Here, we would like to note that the closed association scheme is a simplification of the behavior of real systems, and the same changes in
and in
for
were also reported by other experimentalists and computational scientists who studied the micellization in the bulk [
59,
76,
77,
78,
79].
All data presented so far indicate that A32B32 copolymers can hardy associate inside the narrow and medium narrow cylindrical pores (
up to 30) with inert walls. The data presented in
Figure S3 in the Supplementary Material corroborate this conclusion. They show that A32B32 do not associate even if they were initially inserted into the pores. The chains escape fast from the pore and associate in the bulk phase. Because the strong confinement effect obviously prevents the formation of micelles in narrow pores, we were curious about the behavior in wide pores. Hence, we enlarged the pore to
and performed two sets of simulations at high concentrations, starting from both the bulk-located and the pore-located chain configurations. As shown in
Figure 4c, the results of these two sets of simulations differ considerably, indicating the frozen non-equilibrium behavior in the case that all chains were initially in the pore. The unacceptably high
for SEC is apparently due to the fact that the chains associate in wide pores at concentrations above the CMC and the confined micelles are trapped there due to several contributing effects. Steric obstacles not only restrict the motion of crowded bulky micelles and prevent their escape into the bulk phase but also hinder their reorganization and dissociation. The escape of free chains from pores, which is expected to shift the equilibria in favor of the associates in bulk (see Equations (
1)–(
3)), is also inefficient because the transfer of unimers from pores into bulk controlled by low unimer concentrations and by tiny concentration gradient is extremely slow.
The simulations starting with all chains in the bulk faithfully emulate the equilibration process in SEC and IC: during the gradual passage through the column, the chains at the front edge of the analyte zone first enter the empty pores from the mobile phase, and hence the accumulation and stacking of chains in pores do not occur. In the real systems studied by experimental chromatography, as well as in model systems studied by our simulations, the associates in the bulk are relatively far from each other and the free chains can move and enter the pores without major obstacles. Therefore, we are persuaded that the simulations starting with chains in the bulk provide the equilibrium data.
To gain insight into the association of A32B32 diblock copolymers inside the pores, we performed the separate CBMC simulations in the pores of different diameters,
, excluding the possibility of the bulk-pore partitioning. In
Figure 5a, we plot the fraction of free chains in the pore,
, against
for the constant concentration
, i.e., for 324 chains in the pore. In this case, when the concentration is high and the chains cannot escape from the pores, the association takes place even in the narrowest studied pore of
(
is merely 0.32). The corresponding snapshot in
Figure 5a shows that the associating copolymers under confinement do not form typical micelles like those in bulk, which agrees with extensive studies of other authors [
40,
46,
48]. Therefore, we do not show the simulation data here and concentrate on the findings important for chromatography.
Figure 5a shows that
decreases rapidly with increasing
and reaches a plateau (ca. 0.05) for
. The extremely low
indicates significant association due to high
(see Equation (
3)). The snapshot for
in
Figure 5a reveals the micelle-like clusters in wide pores with solvophobic cores concentrated close to the pore center. In contrast, the snapshot for
shows that the blocks (B) gather near the pore surface and the solvophilic ones concentrate near the pore center. To confirm these observations, we plot the concentration of beads,
, against the distance from the pore center perpendicular to the pore axis,
, for the low and high concentrations in
Figure 5b,c. For all
, we see that at the extremely low concentration (
), i.e., only 5 single chains in the pore,
is higher than
in the central region, and lower than
near the pore wall. At the high concentration (
),
is considerably higher than
in the pore center and lower than
near the wall for
and 60 as a result of the micelle formation. Nevertheless, for
, both curves of
and
vs.
are analogous to those at the low concentrations, which can be attributed to the strong confinement effect. Hence, the conclusions drawn from the concentration profiles agree with those drawn from the snapshots in
Figure 5a.
From the simulations performed separately in the bulk solution and in the pores of various
, i.e., without taking the partitioning equilibrium in account, we evaluated the excess chemical potential,
, of A32B32 diblock copolymers using the Rosenbluth method as described in the methodology section [
80,
81]. In
Figure 5d, we plot
against the concentration of beads in the simulation box,
, for the copolymers in the bulk and in the pores of which the diameter,
, ranges from 20 to 60. The comparison of curves for the bulk phase and the pores differing in diameter is noteworthy. In the dilute regime, i.e.,
(see the inset of
Figure 5d), the copolymers in the bulk acquire the highest values of
, and the
of the confined chains decreases with decreasing
at a given concentration,
. This means that the A32B32 diblock copolymers in the narrowest pore of
have the lowest
. Nonetheless, the difference in
between the bulk and the pore, which controls the bulk-pore partitioning in the SEC mode at rather low concentrations (or at
), is small regardless of the pore diameter,
. Note that the shape of
vs.
curves at higher concentrations is particularly interesting. The chemical potential,
, for the narrow pore of
decreases smoothly with
in the whole concentration region, which precludes major structural changes and indicates that the unimer present at low concentrations also persists as a unique species at concentrations exceeding the bulk CMC. The
of the A32B32 diblock copolymers submitted to the pore of
monotonically decreases with
as well but becomes considerably lower than that of the copolymers in the pore of
when
because of the formation of aggregates with low association numbers inside the pore, demonstrated by the
vs.
and
vs.
curves for
in
Figure 3b and
Figure 4b, respectively. As the aggregates are small and the association/dissociation is reversible, we can still obtain the equilibrium data such as the partition coefficients,
, from the pore of
(the orange curve in
Figure 2a). However, the shapes of curves for the bulk and for the wide pores of
, and their
values, significantly differ from those for the relatively narrow pores of
and 30. The initial parts of
vs.
curves for the bulk and the wide pores of
at low
are similar to those for
, but later the curves exhibit distinct break points, indicating a significant structural change. The breaks occur approximately at the bulk CMC, and then all
decrease appreciably. In spite of the trivial differences in
between the bulk and the pores, all the values of
at
are considerably low and similar, indicating that the highly stable micelles do not form only in the bulk but can also form in the pores at
under certain conditions, i.e., if they cannot escape in bulk. Moreover, the strong association of copolymers in the wide pores results in the non-equilibrium frozen states, and consequently the partition coefficients obtained from the simulations starting from the pore are unreliably high as shown in
Figure 4c. In summary, we carefully investigated the phase equilibria of amphiphilic A32B32 diblock copolymers partitioning between the bulk and the pores with inert walls and have shown the impact of the association of copolymers on the partition coefficient,
.
In the next part, we investigate the potential of the SEC characterization of AB diblock copolymers differing in the lengths of two blocks (the total length is constant, i.e.,
). We simulated their partitioning between the bulk solvent and the pore of
and varied the temperature,
, from 1.5 to 1.8. The dependences of
on
obtained from the simulations are shown in
Figure 6. They are qualitatively similar to those shown in the previous figures and agree with the generally known effect of the solubility of block copolymers on their associating behavior. The solubility of copolymers decreases with the increasing relative length of the insoluble block and with decreasing temperature, and consequently their association tendency increases, which translates into variations of simulated
vs.
curves. The CMC values estimated from the simulations shift to lower
with the increase in length of the insoluble block, and simultaneously the
-drop becomes steeper and deeper (
approaches almost zero) as a result of the increasing association tendency. As the solubility of copolymers with long soluble blocks A is sufficiently high, these copolymers do not associate at all, which is shown by more or less constant or slightly increasing
vs.
plots, i.e., at
and 1.8, indicating that the non-associated chains behave as effectively at elevated temperatures as polymers in good solvents [
36,
37].
At first glance, the effect of the relative length of blocks on
in the region of low concentrations can be surprising.
increases appreciably with the length of the insoluble block and with decreasing temperature. At
(
Figure 6a), the partition coefficients of A24B40 (green crosses) and A28B36 (orange squares) are even higher than 1 when
. This is not the simulation artifact, even though such values are not achievable in real SEC experiments. The increase of the partition coefficient,
, with the decreasing solubility of the copolymers and the extraordinary
for copolymers with long solvophobic blocks can be explained relatively easily: they stem from an extremely strong unfavorable interaction between the solvent and block B. As the chains do not associate below the CMC and the collapse of the insoluble block does not sufficiently prevent the contacts of B segments with the solvent, the high-energy system of dissolved chains exploits another possibility to minimize the Gibbs function. The behavior is reminiscent of that of surface-active compounds that accumulate at interfaces [
82,
83]. The minimization of the number of unfavorable interactions is achieved when blocks B concentrate close to the inert pore wall. To reach this goal, the enthalpic driving force pushes the copolymer chains into the pore, where they acquire suitable conformations close to the wall.
In summary,
Figure 6 depicts the changes in phase equilibria of dissolved associating systems during their transition from a selective solvent to a poor common solvent, later to the
-solvent, and finally to a good common solvent with the increase in the ratio of solvophilic–to-solvophobic blocks and with temperature. We believe that the interesting outlined trends can be efficiently studied experimentally by temperature-modulated chromatographic techniques, e.g., by temperature–gradient interaction chromatography (TGIC) [
3].
At the end of this article, we focus on the partitioning of A32B32 copolymers between the bulk phase and pores with adsorbing walls, which relates to the interaction chromatography (IC) mode. We assume that the pore wall is inert for the solvophilic block (A), i.e.,
, and attracts the solvophobic block (B), i.e.,
. One can imagine that if the attraction is very strong, the micelles in bulk will not form because all the chains will tenaciously adsorb on the pore wall, and
will be very high. However, this extreme case is unsuitable for real chromatography. Hence, the
used is up to
and the temperature
, which is relevant for practical LC experiments. We performed two sets of simulations, which started with all chains either in the bulk or in the pore. In
Figure 7a,b, we show the
vs.
curves obtained by simulations starting from the bulk phase for
and
, respectively. For the reader’s convenience, the curves for
(SEC mode) are also included. The comparisons between two groups of simulations with different initial conditions for
and 60 are presented in
Figures S4 and S5 in the Supplementary Material. As shown in
Figure S4, the values of
for
are identical because very few chains can aggregate in the narrow pore. However, for the wide pore of
, the data from the two sets of simulations differ significantly when
(
Figure S5). The extremely high values of
provided by the simulations starting with all chains in the pore imply that the micellization of A32B32 copolymers takes place inside the pore and combines with the adsorption of chains on the pore wall. The micelles become trapped in pores similarly to the previously studied SEC systems (see
Figure 4c). The reasons are similar in both cases, but in the IC regime, the escape of polymer chains from pores is even less probable because
and the concentration of free chains in bulk is lower than that in pores, i.e., in this particular case lower than CMC, which precludes the association of chains in the bulk and thus eliminates the process that could drive the chains from the pores into the bulk phase. Even though
does not describe the equilibrium situation and the data in
Figure S5 are irrelevant for common IC chromatography, the obtained pieces of information are still useful for analyzing specific LC experiments, e.g., for the gradient or barrier chromatography [
7]. In agreement with common LC processes, here we focus only on the case that all copolymers are initially in the bulk mobile phase (
Figure 7a,b). Generally,
increases with increasing
in the full range of studied concentrations for both
and 30, and the increase in
is more pronounced in the dilute regime because the self-assembly in bulk dominates the partitioning at high concentrations, i.e., above the CMC. Nonetheless,
Figure 7a demonstrates that the copolymers still undergo SEC separations in narrow pores, i.e., the maximum
for
is merely 0.28 (considerably lower than 1) because the attractive interactions of insoluble blocks with the pore wall are insufficient to offset the entropy loss caused by the severe confinement and are undoubtedly incomparable with the strong association in bulk. However, for the wide pore of
with the strongest adsorption of
(see the magenta curve with inverted triangles in
Figure 7b),
is approximately 1.8 when
, first increases with
, and reaches a maximum of
, indicating that the separation in the dilute regime obeys the IC mechanism and the adsorptive interaction prevails over the entropy loss.
We assume that the pronounced maximum and the shift of the fairly steeply decreasing part of vs. on the magenta curve (for the strong attractive interaction of insoluble beads with the walls of the wide pore) to higher concentrations followed by a slower but still appreciable -decrease reflect the adsorption of insoluble blocks and consequent gradual saturation of the pore surface, followed by potential changes in the self-organization of chains due to the increase in , e.g., by the formation of core-shell associates inside the pore. To prove this assumption, we plot further the concentration profiles of beads and present typical simulation snapshots.
Using the simulation data on the copolymer partitioning, we plot the radial concentration profiles of beads A and B, i.e.,
vs.
and
vs.
, in the pore of
at the high total concentration
for various
in
Figure 7c,d, respectively. Note that in contrast to the radial concentration profiles plotted in
Figure 5b,c, the total number of beads in the pore (at constant total concentration,
) is not constant, and it increases with
, which slightly affects the absolute values of
and
but does not change the general trends. As expected, the concentration of solvophobic segments (B) near the inner surface of the pore is higher than that of solvophilic beads (A), and the difference increases with increasing
. The concentration profiles of beads A and B for
reveal that the solvophobic blocks are enriched in the layer near the pore wall, while the solvophilic ones favor the central region, which is also confirmed by the snapshot of the stationary phase in
Figure 7e.
As the micellization of block copolymers and surfactants obeys the same principles, our computer study (particularly the simulations for the wide adsorptive pore, i.e.,
and
) represents the first step towards the understanding of complex phase equilibria of surfactants in MLC. The structure of the micelles composed of dense solvophobic cores and less dense solvophilic shells is suitable for the solubilization or for the specific non-covalent binding of various molecules: (i) both the shells of surfactant micelles in bulk (mobile phase) and the soluble parts of unimers firmly adsorbed on pore walls can specifically interact with various soluble compounds, facilitate their penetration into pores, and control their chromatographic separation from non-interacting species [
8]; and (ii) the cores of micelles in bulk can solubilize the otherwise insoluble compounds, and the insoluble parts of adsorbed unimers in pores can interact favorably with them, which can intermediate their partitioning, transport, and separation in the column.
We envisage that our study of the partitioning of copolymers and surfactants between the bulk and the wide pore, of which the inner surface strongly adsorbs solvophobic units (e.g., and ), demarks the correct approach towards the investigation of the MLC separation mechanism. The synergy of the micellization in the bulk, the adsorption of surfactants on the pore wall, and the attractive interaction with analytes enables the versatile tuning of MLC elution, and the present study already reveals some principles of this complex molecular mechanism. The computational investigation of the MLC process based on this model will be presented in our future work.