Coil–Globule Collapse of Polystyrene Chains in ... - ACS Publications

Coil–Globule Collapse of Polystyrene Chains in Tetrahydrofuran–Water Mixtures. Tatiana I. ... View: ACS ActiveView PDF | PDF | PDF w/ Links | Full...
0 downloads 0 Views 2MB Size
Article Cite This: J. Phys. Chem. B 2018, 122, 2130−2137

pubs.acs.org/JPCB

Coil−Globule Collapse of Polystyrene Chains in Tetrahydrofuran− Water Mixtures Tatiana I. Morozova and Arash Nikoubashman* Institute of Physics, Johannes Gutenberg University Mainz, Staudingerweg 7, 55128 Mainz, Germany

Downloaded via KAOHSIUNG MEDICAL UNIV on June 30, 2018 at 16:39:02 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

S Supporting Information *

ABSTRACT: We study the coil and globule states of a single polymer chain in solution by performing molecular dynamics simulations with a united atom model. Specifically, we characterize the structural properties of atactic polystyrene chains with N = 20−150 monomers in tetrahydrofuran−water mixtures at varying mixing ratios. We find that the hydrophobic polymers form rather open coils when the mole fraction of water, XW, is roughly below 0.25, whereas the chains collapse into globules when XW ≳ 0.75. We confirm the theoretically expected scaling laws for the radius of gyration, Rg, in these regimes, i.e., Rg ∝ N3/5 and Rg ∝ N1/3 for good and poor solvent conditions, respectively. For poor solvent conditions with XW = 0.75, we find a sizable fraction of residual tetrahydrofuran trapped inside the collapsed polymer chains with an excess amount located at the globule surface, acting as a protective layer between the hydrophobic polystyrene and the surrounding water-rich mixture. These findings have important implications for nanoparticle fabrication techniques where solvent exchange is exploited to drive polymer aggregation, since residual solvent can significantly influence the physical properties of the precipitated nanoparticles.

1. INTRODUCTION Polymers in solution are ubiquitous in technology and nature, where they serve a wide range of purposes, for example, as viscosity modifiers in the food and cosmetic industry1 or for storing the genetic makeup in the form of DNA. The physical properties and functions of these macromolecules depend to a high degree on their conformation. For example, proteins in biological cells fold into specific structures based on the sequence of the constituent amino acids and the local environment,2,3 and it has been hypothesized that misfoldings can give rise to numerous pathological conditions.4 Thus, it is crucial to understand and control the numerous factors which can influence the self-assembly of macromolecules. One important aspect that determines the conformation of a polymer dispersed in solution is the quality of the surrounding solvent. Under good solvent conditions, the chains attain a swollen coil-like configuration with a sizable amount of solvent located within the enclosed volume. Under poor solvent conditions, however, the chains collapse into a compact globule to minimize the interface area with the surrounding liquid. This transition between extended and collapsed states has been the focus of numerous theoretical,5−7 computational,8−10 and experimental studies.7,11−13 Here, the main goal usually was to elucidate the details of the coil−globule transition of a single chain, e.g., whether it is a first- or second-order phase transition. At sufficiently high polymer concentrations, the coil−globule transition of the individual chains is usually accompanied by the aggregation of neighboring macromolecules. This behavior can be exploited for the controlled fabrication of polymeric colloids,10,14−18 where the structure and composition of the aggregates can be tuned through, e.g., the employed polymers, © 2018 American Chemical Society

additives, and mixing conditions. However, the underlying microscopic mechanisms responsible for the particle formation are still unclear, as the precipitation process typically takes place on nanometer length scales and millisecond time scales, making experimental characterization of this process difficult. Recent computational efforts have shed some light on the precipitation process10,19−21 but relied on a rather coarse-grained description of the polymer solution to render the simulations computationally tractable. Polymers were modeled as generic bead− spring chains, and mixing of the good and poor solvent was mimicked by a gradual change of the effective monomer− solvent interactions. Although such simplified models can capture a substantial amount of the relevant physics, they suffer from a certain ambiguity of the interaction parameters, and hence inevitably miss some effects which are related to the specific chemistry of the materials. Further, the grouping of good and poor solvent molecules into effective particles precludes situations where the distribution of solvent molecules becomes heterogeneous, which can be, for instance, the case when good solvent is enclosed in the precipitated nanoparticles. To better understand the coil−globule transition and the effect of solvent trapping, we performed computer simulations of a single polystyrene (PS) chain in tetrahydrofuran−water mixture at various mixing ratios. We employed a united atom (UA) model to reproduce the experimental system as realistically as computationally feasible. We show that the coil−globule transition occurs when the mole fraction of water, Received: October 26, 2017 Revised: January 29, 2018 Published: January 29, 2018 2130

DOI: 10.1021/acs.jpcb.7b10603 J. Phys. Chem. B 2018, 122, 2130−2137

Article

The Journal of Physical Chemistry B

longest polymers (N = 150). The simulation time, τsim, was chosen to be several Zimm relaxation times of the dispersed polymer chain,27 τZimm = ηRg3/kBT. As a reference, we chose τsim = 10 ns for the shortest chains in pure water (τZimm = 0.22 ns) and τsim = 40 ns for the longest chains in THF−water mixtures with XW = 0.75 (τZimm = 14.3 ns). All systems were studied at room temperature (T = 298 K) and atmospheric pressure (P = 101,325 Pa), as in recent flash nanoprecipitation experiments involving PS homopolymers.10,28 Molecular dynamics (MD) simulations were carried out on Graphical Processing Units (GPUs) using the HOOMD-blue software package.29−31 The equations of motion were integrated using the velocity Verlet algorithm32,33 with a time step of Δt = 1.0 fs. A Nosé−Hoover thermostat was employed for simulations under NVT conditions,34,35 while the Martyna−Tobias−Klein equations of motion were employed to achieve NPT conditions.36

XW, is roughly above 0.75. Under such poor solvent conditions, we find a sizable fraction of residual tetrahydrofuran (THF) trapped inside the collapsed polymer chains. Further, an excess amount of THF is located at the surface of the globule, acting as a protective layer between the hydrophobic polystyrene and the surrounding water-rich mixture. The rest of this manuscript is organized as follows. In section 2, we describe in detail the employed simulation model and method, while we present and discuss our findings in section 3. In section 4, we provide our conclusions and outlook.

2. SIMULATION DETAILS AND METHODS In order to model PS chains in solution as accurately as computationally possible, we opted for a UA model where a small group of adjacent atoms is represented by one effective bead. For instance, hydrogen and carbon atoms in a methyl group are treated as a single interaction site, which is a reasonable approximation for molecular systems where the intermolecular motion is much more important than the intramolecular one. Such an approach can save valuable computation time compared to fully atomistic models, while preserving molecular detail to a greater extent than more coarse-grained models. In particular, we decided to use the transferable potentials for phase equilibria-united atom (TraPPE-UA) force fields for describing the THF molecules22 and atactic PS chains.23,24 We modeled the water molecules using the three-site SPC/Fw model with flexible bonds and point charges.25 In what follows, we will provide a detailed description of the simulation model and method. We have included the employed simulation parameters as Supporting Information, for the sake of brevity. Figure 1 shows a schematic representation of a THF molecule and a PS monomer, together with the employed

3. RESULTS AND DISCUSSION 3.1. Structural and Dynamic Properties of THF−Water Mixtures. We first performed MD simulations of THF−water mixtures at different compositions without PS chains, to test our model and to characterize the properties of the liquid. In particular, we studied the following mole fractions of water, XW = 0, 0.3, 0.63, 0.9, and 1. In order to determine the dynamic and static properties of pure THF (XW = 0) and pure water (XW = 1), we conducted simulations with 1200 molecules in the simulation box. For the THF−water mixtures, the simulation box typically contained ∼4000 solvent molecules, distributed between the water and THF molecules according to XW. We conducted three independent runs at each state point to compute averages and standard deviations. We determined the mass density ρ at room temperature and ambient pressure by conducting NPT simulations for 30 ns. The dynamic viscosity was then calculated in the NVT ensemble using start configurations obtained from the previous NPT simulations. Here, we employed the Green−Kubo relation for computing the shear viscosity η=

V kBT

∫0



dt ⟨Pαβ(t )Pαβ(0)⟩

(1)

where Pαβ are the Cartesian components of the stress tensor. We improved the sampling by using all five independent components of the traceless stress tensor,37 i.e., Pxy, Pxz, Pyz, (Pxx − Pyy)/2, and (Pyy − Pzz)/2, and computed η from the averaged stress autocorrelation function of each trajectory. We have plotted our simulation results for all investigated XW in Figure 2, where we also included experimental measurements38,39 as a reference. The mass densities determined from the simulations are always within 1% of the experimental value, whereas the viscosities deviate by up to 20% with respect to the experimental measurements. Note that the discrepancy between η from simulations and experiments is almost negligible for XW ≥ 0.9, since the employed SPC/Fw water model has been parametrized to faithfully reproduce the dynamic properties of water.25 We also verified that the THF and water molecules remained mixed during the simulation by visual inspection and by computing the intermolecular radial pair distribution function g(r) between the centers of mass of THF and water molecules [g(r) for THF−THF and water−water can be found in the Supporting Information]. The g(r) data shown in Figure 3

Figure 1. Schematic representation of (a) a THF molecule and (b) a PS monomer, highlighting the employed grouping into pseudoatoms. Partial charges for the THF molecule are also shown, where e is the elementary charge.

grouping into pseudoatoms. Specifically, the CHx groups are represented by one interaction site: for a THF molecule, each CH2 group and each oxygen atom is modeled as a single bead (five in total per molecule), while, for a PS monomer, each CH3, CH2, CH, Caro, and CHaro group is represented by a single bead (eight in total per monomer). For the simulations containing polymer chains, starting configurations were generated using a modified version of the program Assemble!,26 and we performed three independent runs for each polymer length and mixture composition. The edge length of the cubic simulation box was chosen to be at least 2.5 times the radius of gyration of the dispersed polymer, Rg, to avoid self-interactions. Typically, the simulation box contained approximately 2000 solvent molecules for the shortest PS chains (N = 20) and up to 32,000 solvent molecules for the 2131

DOI: 10.1021/acs.jpcb.7b10603 J. Phys. Chem. B 2018, 122, 2130−2137

Article

The Journal of Physical Chemistry B

XW = 0 (pure THF), XW = 0.25, XW = 0.50, XW = 0.75, and XW = 1 (pure water). The length of the PS chain was set to N = 20, 50, 100, and 150 monomers, which corresponds to a molecular weight of M = 2.1, 5.2, 10.4, and 15.5 kg/mol, respectively. Figure 4 shows simulation snapshots of a PS chain with 150

Figure 2. (a) Mass density ρ and (b) dynamic viscosity η from experiments38,39 and simulations for THF−water mixtures as a function of water mole fraction XW. All data were reported for systems at room temperature (T = 298 K) and under atmospheric pressure (P = 101,325 Pa).

Figure 4. Simulation snapshots of a PS chain with 150 monomers (M = 15.5 kg/mol) dispersed in (a) pure THF and (b) pure water. The scale bar indicates 10 Å. Snapshots rendered using Visual Molecular Dynamics 1.9.2.43

monomers dispersed in pure THF and water, respectively. In THF, the polymer is in an extended state, whereas, in water, the hydrophobic chain collapses to a compact globule to minimize the interface with the surrounding poor solvent. In order to quantify the size and shape of the dispersed polymers, we computed the radius of gyration tensor Gαβ = Figure 3. Intermolecular radial pair distribution function, gCM(r)THF−W, between the centers of mass of THF and water molecules at different water mole fractions XW.

1 M

NUA

∑ miΔri ,αΔri ,β i=1

(2)

where NUA is the number of united atoms in a polymer chain, M is the total polymer mass, and Δri,α is the position of united atom i relative to the polymer center of mass rcm, while α and β are the components in the Cartesian x, y, and z directions. The polymer radius of gyration is then given by Rg2 = Gxx + Gyy + Gzz. We have plotted in Figure 5a Rg as a function of N for various solvent compositions, and it is clear that the polymers are collapsed when they are dispersed in solutions with a high water content (XW ≥ 0.75). In contrast, the chains are rather swollen when the mixtures consist of mostly THF (XW ≤ 0.25). We found that the radius of gyration scaled as Rg ∝ Nν, with νpoor = 0.27 ± 0.06 and νgood = 0.67 ± 0.12 for our simulations in pure water and pure THF, respectively. These scaling exponents are in good agreement with theoretical considerations, which predict a value of νpoor = 1/3 for a densely

reveals the presence of two pronounced solvation shells of water around THF molecules at r ≈ 3 Å and r ≈ 5 Å, respectively. For larger distances r ≳ 10 Å, we find g(r) ≈ 1, which indicates homogeneous mixing of the two liquids. Similar results were obtained in previous united atom40 and fully atomistic41,42 simulations of THF−water systems. Note that the radial pair distribution function between the centers of mass of THF and water molecules did not change significantly when a polymer was dispersed in the mixture (graphs not shown for the sake of brevity). 3.2. Polystyrene in THF−Water Mixtures. We studied the properties of a single PS chain in THF−water mixtures at 2132

DOI: 10.1021/acs.jpcb.7b10603 J. Phys. Chem. B 2018, 122, 2130−2137

Article

The Journal of Physical Chemistry B κ2 = 1 − 3

λ1λ 2 + λ1λ3 + λ 2λ3 (λ1 + λ 2 + λ3)2

(3)

where λ1, λ2, and λ3 are the three eigenvalues of the instantaneous gyration tensor G. The shape anisotropy parameter attains a value of κ2 = 0 for a fully isotropic shape (sphere) and κ2 = 1 for an infinitely thin rod. For a randomly coiling chain in three dimensions, a value of κ2 ≈ 0.38 is expected in the limit of N → ∞ due to conformational fluctuations of the macromolecule. (Note, however, that the eigenvalues of the average gyration tensor ⟨G⟩ become identical for sufficiently long sampling and thus exhibit no anisotropy.) Figure 5b shows κ2 vs N for various solvent qualities, and we can see that κ2 is close to zero for collapsed chains in pure water and XW = 0.75, as expected. For good and Θ-solvent conditions, we find κ2 ≈ 0.4, which is in line with the theoretical prediction for randomly coiling chains.51,52 To quantify the shape of polymers in solution, it is also instructive to consider the single chain structure factor S(q) S(q) =

1 N

∑ exp(−iq·rij) (4)

i,j

where q is the scattering wave vector. For isotropic systems, the structure factor depends only on the magnitude of q, so that we can simplify eq 4 to Figure 5. (a) Radius of gyration Rg and (b) relative shape anisotropy parameter κ2 of a PS chain as a function of chain length N for various water mole fractions XW. The solid lines in part a are power law fits for pure water (poor solvent) and pure THF (good solvent) with exponents νpoor = 0.27 ± 0.06 and νgood = 0.67 ± 0.12, respectively. Open symbols correspond to experimental Rg measurements of PS in toluene.46,47 The solid line in part b shows the value of κ2 for an ideal chain.51,52 Data in part a shown in log-log scale. Data in part b shown in lin-log scale.

S(q) =

1 N

∑ i,j

sin(qrij) qrij

(5)

The number of monomers in the scattering volume 1/q3 depends on the size and conformation of the polymer. Theoretical considerations suggest S(q) ∝ q−1/ν for self-similar objects such as polymers at intermediate wave vectors 1/Rg < q < 1/b (monomer size b), whereas one would expect S(q) ∝ q−4 for a compact spherical particle.44 We have plotted in Figure 6 the structure factor S(q) of a PS chain with N = 150 monomers in pure THF as well as in pure water. (The curves for N = 100 are qualitatively similar and have been omitted for the sake of brevity.) For small wave vectors q ≪ 1/Rg, we find S(q) ≈ N as expected for dilute systems. At intermediate q, the structure factor decays with a power law. For good solvent conditions, we find S(q) ∝ q−5/3, which corroborates our measurements of the Flory exponent νgood = 0.67 ± 0.12 (cf. Figure 5a). In contrast, the scattering function decays as S(q) ∝ q−4 under poor solvent conditions. To facilitate the interpretation of the simulation data, we have included in Figure 6 also the analytical solution of S(q) for a solid spherical particle with the same geometric radius as the collapsed globule,44 i.e., a = Rg(5/3)1/2 ≈ 20.3 Å. Indeed, for q ≲ 0.2 Å−1, the simulation data precisely follow the theoretical S(q) prediction, suggesting a fully collapsed poor solvent conformation of the PS chain. As q increases, the simulation data start to deviate from the analytical solution, due to the internal structuring of the globule (multiple scatterers vs single scatterer). At high q, we can identify the microscopic structures of the PS chain. The peak of S(q) at q ≈ 2 Å−1 corresponds to the nearest neighbor distance of bonded monomers, and is therefore visible under both good and poor solvent conditions. In pure water, we can identify an additional maximum at q ≈ 0.9 Å−1, which stems from the packing of monomers in the interior of the globule. Next, we investigated the localization of the THF and water molecules with respect to the dispersed PS chain. To this end,

packed globular configuration in a poor solvent and νgood = 3/5 for good solvent conditions set by the balance between excluded volume and conformational entropy.44 Our simulation results for PS in pure THF also agree well with previous experiments, where a scaling exponent of νgood = 0.56 ± 0.02 was found.45 Furthermore, the computed Rg values in pure THF are in quantitative agreement with experimental measurements of low M PS in toluene (open symbols in Figure 5a), which is also a good solvent for PS.46,47 At intermediate solvent qualities, the intermolecular interactions between polymer chain segments and coordinated solvent molecules become similar (Θ-conditions), and the dispersed chain effectively acts like an ideal chain with νΘ = 1/2. From the simulation data shown in Figure 5a, we expect that Θconditions should be achieved roughly in the range 0.25 ≲ XW ≲ 0.50. Experimental measurements of PS in THF−water mixtures revealed that the Θ-composition for PS in THF− water mixtures is XW ≈ 0.3,48 which lies in the range estimated from our simulations. Note that the effect of solvent quality becomes considerably less pronounced for short chains, since the inherent stiffness of PS impedes the full collapse of neighboring segments. (The chains with N = 20 consist of approximately three Kuhn segments.44,49) A similar trend has also been observed in experiments for low M PS using a wide range of good solvents.45,50 The shape anisotropy of the dispersed polymer can be quantified through 2133

DOI: 10.1021/acs.jpcb.7b10603 J. Phys. Chem. B 2018, 122, 2130−2137

Article

The Journal of Physical Chemistry B

we computed the radial distribution function between the center of mass of the dispersed polymer chain and the centers of mass of the solvent molecules, gCM(r)PS−THF/W. Here, we focus on the cases with N = 100 and N = 150 monomers, as these chain lengths exhibit the strongest dependence on solvent quality (cf. Figure 5). It is clear from the data shown in Figure 7a and c that the THF molecules are almost homogeneously distributed in the system, irrespective of solvent composition XW and chain length N; for the cases with XW = 0.75, we can identify a slight depletion of THF molecules in the polymer core (r ≈ 0) and a weak accumulation in the polymer shell (r ≈ Rg). In contrast, the distribution of water molecules depends much more strongly on the solvent composition. For XW = 0.25 and 0.5, we find a sizable fraction of water inside the PS coils, which increases monotonically with r until it reaches its bulk value at r ≈ 1.5 Rg (see Figure 7b and d). At first, this behavior might seem implausible because of the hydrophobic nature of PS, but it can be rationalized when one considers the open and coil-like conformation of the dispersed PS chains under these conditions; although some water molecules are located near the polymer’s center of mass, they do not necessarily need to be in close contact with the contour of the polymer (see discussion below). For XW = 0.75 and 1, there is almost no water inside the collapsed polymer globules [gCM(r)PS−W ≤ 0.5 for r ≤ Rg], and the bulk value is reached at r ≈ Rg after a relatively sharp transition. (Note that the geometric radius of a spherical globule is slightly larger than its radius of gyration.44) To study in more detail the distribution of THF and water molecules in the immediate vicinity of the dispersed PS chain, we also computed the radial distribution function between

Figure 6. Structure factor S(q) of a single PS chain with N = 150 monomers in pure THF and pure water. The solid lines indicate the theoretically expected behavior for good [S(q) ∝ q−5/3] and poor solvent conditions [S(q) ∝ q−4], respectively. The dashed line shows the analytic S(q) for a spherical particle. Data shown in log-log scale.

Figure 7. Radial distribution functions gCM(r) between the centers of mass of the dispersed PS chain and THF/water molecules for various solvent mixtures. Data for N = 100 are shown in parts a and b, while data for N = 150 are shown in parts c and d. The shaded regions correspond to the measurement uncertainty determined from the standard error of the independent simulations. 2134

DOI: 10.1021/acs.jpcb.7b10603 J. Phys. Chem. B 2018, 122, 2130−2137

Article

The Journal of Physical Chemistry B monomers and solvent molecules, gCM(r)S−THF/W. Figure 8a shows the distribution of THF molecules, and we can identify a

Figure 9. Number densities of (a) THF and (b) water molecules around a polymer chain with N monomers, normalized by the bulk value.

Figure 8. Radial distribution function gCM(r) between the centers of mass of the monomers of a PS chain (N = 150) and THF/water molecules for various solvent mixtures. The shaded regions correspond to the measurement uncertainty determined from the standard error of the independent simulations.

chain, which is consistent with the g(r) data shown in Figures 7 and 8. Figure 9b shows the normalized local number density of water molecules along the PS contour, from which we can 0 identify several key trends. First, ρW/ρW decreases with increasing chain length N for all investigated XW, because the water molecules are primarily located at the shell of the polymer coil (cf. Figure 7b and d) and the surface to volume ratio of the polymer decreases with increasing polymerization. Second, the relative fraction of adsorbed water molecules decreases with increasing XW for the THF−water mixtures, which seems counterintuitive at first sight. This behavior can be explained, however, when we consider that the polymer is much more collapsed at high water concentrations, and thus fewer water molecules can penetrate the core of the polymer (see Figure 5a). Finally, we briefly investigated the dynamics of the coil− globule transition, by conducting a simulation in pure water where the dispersed polymer (N = 100) was initialized in an extended coil configuration. Once we started the simulation, the hydrophobic macromolecule tightened into a compact globule (cf. Figure 4b), and we estimated the characteristic collapse time τc by fitting the time evolution of Rg through a sigmoid function.10 Here, we found τc = 3.0 ± 0.1 ns, which is comparable to the Zimm relaxation time of the entire polymer (τZimm ≈ 5 ns). However, this finding should be taken with a grain of salt, because of the limited sample size here; there are numerous pathways from the coil state to the collapsed one, and studying the coil−globule collapse dynamics in detail requires advanced sampling methods, such as Markov state modeling,53 so that all relevant transitions can be captured efficiently. This aspect will be the focus of a future publication.

rather sharp peak at r ≈ 6 Å which becomes more distinct with increasing XW. This peak is followed by a sequence of less pronounced minima and maxima, reflecting the layering of solvent particles. From Figure 8b, we can further see that g(r) between PS monomers and water molecules reaches unity only at distances of r ≳ Rg (cf. Figure 5a), due to the hydrophobic nature of the PS chain. Interestingly, for the THF−water mixtures, the relative amount of water around a PS monomer decreases slightly with increasing water content, while the local amount of THF increases. These findings suggest the formation of a (thin) protective layer between the hydrophobic polymer and the aqueous surrounding (cf. Figure 7a and c). In order to quantify the enrichment/depletion of THF and water molecules around the PS chain, we computed the local density along the polymer contour. To this end, we first constructed a fictitious worm-like chain along the PS backbone, with the thickness set to the interaction range of the dispersion force (rc = 16 Å for THF and rc = 10 Å for water). Then, we identified all THF and water molecules lying in this region, and computed the size of the search volume through Monte Carlo integration. Figure 9 shows the local number densities of THF (ρTHF) and water (ρW) molecules normalized by their bulk value, ρ0THF and ρ0W, respectively. For THF, we find that ρTHF is slightly below ρ0THF for most mixing fractions irrespective of chain length, which we attribute to the higher solubility of THF in THF than PS in THF. Interestingly, the simulations with the largest water fraction (XW = 0.75) indicate a small but significant enrichment of THF along the contour of the PS 2135

DOI: 10.1021/acs.jpcb.7b10603 J. Phys. Chem. B 2018, 122, 2130−2137

Article

The Journal of Physical Chemistry B

4. CONCLUSIONS We studied the conformation of single polystyrene chains with various degrees of polymerization, N, dispersed in a tetrahydrofuran−water mixture using molecular dynamics simulations of a united atom model. We systematically varied the mixing ratio of the two liquids to tune the solvent conditions from good (pure tetrahydrofuran) to poor (pure water). In good solvents, the macromolecules exhibited an open coil-like conformation, whereas the chains collapsed to compact globules in the poor solvents. We measured the radius of gyration, Rg, and found that the simulation data followed the theoretically expected trends, i.e., Rg ∝ N3/5 and Rg ∝ N1/3 for good and poor solvent conditions, respectively. Further, we investigated the spatial distribution of the tetrahydrofuran and water molecules with respect to the polystyrene chains. We discovered that the location of water molecules depended strongly on the specific solvent composition, whereas the tetrahydrofuran molecules were almost homogeneously distributed in the system, irrespective of solvent composition and chain length. Interestingly, we identified a small mixing regime (water mole fraction XW = 0.75), where a sizable fraction of residual tetrahydrofuran molecules was trapped inside the collapsed polymers, with an excess amount located at the globule−solvent interface, serving as a protective layer between the hydrophobic polystyrene and the surrounding water-rich mixture. Further, our simulations provide important insights for developing coarse-grained models for polymers in solvent mixtures: when the distribution of the good and poor solvent can become heterogeneous (as is the case in our investigated systems), then a coarse-grained description where multiple solvent molecules of different species are lumped into a single effective solvent bead does not capture the entire physical picture anymore, since effects such as solvent trapping cannot be replicated.54 In addition, our simulations lay the foundation for accurately studying the dynamic pathways of the coil− globule transition for polymers in poor solvents. On the basis of these atomistic simulations, we plan to develop enhanced coarse-grained polymer models, which reproduce not only the structural polymer properties but also the internal polymer dynamics and its hierarchies.



granted on the supercomputer Mogon at Johannes Gutenberg University Mainz (hpc.unimainz.de).



(1) Schulz, D. N., Glass, J. E., Eds. Polymers as rheology modifiers; ACS Symposium Series; American Chemical Society: Washington, DC, 1991; Vol. 462. (2) Gething, M.-J.; Sambrook, J. Protein folding in the cell. Nature 1992, 355, 33−45. (3) Dobson, C. M. Protein folding and misfolding. Nature 2003, 426, 884−890. (4) Chiti, F.; Dobson, C. M. Protein misfolding, functional amyloid, and human disease. Annu. Rev. Biochem. 2006, 75, 333−366. (5) de Gennes, P. G. Scaling concepts in polymer physics; Cornell University Press: Ithaca, NY, 1979. (6) de Gennes, P. G. Kinetics of collapse for a flexible coil. J. Phys., Lett. 1985, 46, 639−642. (7) Vasilevskaya, V. V.; Khokhlov, A. R.; Matsuzawa, Y.; Yoshikawa, K. Collapse of single DNA molecule in poly(ethylene glycol) solutions. J. Chem. Phys. 1995, 102, 6595. (8) Ivanov, V. A.; Paul, W.; Binder, K. Finite chain length effects on the coil-globule transition of stiff-chain macromolecules: A Monte Carlo simulation. J. Chem. Phys. 1998, 109, 5659. (9) Polson, J. M.; Moore, N. E. Simulation study of the coil-globule transition of a polymer in solvent. J. Chem. Phys. 2005, 122, 024905. (10) Nikoubashman, A.; Lee, V. E.; Sosa, C.; Prud’homme, R. K.; Priestley, R. D.; Panagiotopoulos, A. Z. Directed assembly of soft colloids through rapid solvent exchange. ACS Nano 2016, 10, 1425. (11) Swislow, G.; Sun, S.-T.; Nishio, I.; Tanaka, T. Coil-Globule Phase Transition in a Single Polystyrene Chain in Cyclohexane. Phys. Rev. Lett. 1980, 44, 796. (12) Kiriy, A.; Gorodyska, G.; Minko, S.; Jaeger, W.; Štěpánek, P.; Stamm, M. Cascade of coil-globule conformational transitions of single flexible polyelectrolyte molecules in poor solvent. J. Am. Chem. Soc. 2002, 124, 13454−13462. (13) Xu, J.; Zhu, Z.; Luo, S.; Wu, C.; Liu, S. First observation of twostage collapsing kinetics of a single synthetic polymer chain. Phys. Rev. Lett. 2006, 96, 027802. (14) Fessi, H.; Puisieux, F.; Devissaguet, J. P.; Ammoury, N.; Benita, S. Nanocapsule formation by interfacial polymer deposition following solvent displacement. Int. J. Pharm. 1989, 55, R1−R4. (15) Johnson, B. K.; Prud’homme, R. K. Chemical processing and micromixing in confined impinging jets. AIChE J. 2003, 49, 2264− 2282. (16) Bilati, U.; Allémann, E.; Doelker, E. Development of a nanoprecipitation method intended for the entrapment of hydrophilic drugs into nanoparticles. Eur. J. Pharm. Sci. 2005, 24, 67−75. (17) Higuchi, T.; Tajima, A.; Yabu, H.; Shimomura, M. Spontaneous formation of polymer nanoparticles with inner micro-phase separation structures. Soft Matter 2008, 4, 1302−1305. (18) Lee, V. E.; Sosa, C.; Liu, R.; Prud’homme, R. K.; Priestley, R. D. Scalable platform for structured and hybrid soft nanocolloids by continuous precipitation in a confined environment. Langmuir 2017, 33, 3444. (19) Spaeth, J. R.; Kevrekidis, I. G.; Panagiotopoulos, A. Z. A comparison of implicit- and explicit-solvent simulations of selfassembly in block copolymer and solute systems. J. Chem. Phys. 2011, 134, 164902. (20) Spaeth, J. R.; Kevrekidis, I. G.; Panagiotopoulos, A. Z. Dissipative particle dynamics simulations of polymer-protected nanoparticle self-assembly. J. Chem. Phys. 2011, 135, 184903. (21) Li, N.; Panagiotopoulos, A. Z.; Nikoubashman, A. Structured nanoparticles from the self-assembly of polymer blends through rapid solvent exchange. Langmuir 2017, 33, 6021. (22) Keasler, S. J.; Charan, S. M.; Wick, C. D.; Economou, I. G.; Siepmann, J. I. Transferable potentials for phase equilibria-United atom description of five-and six-membered cyclic alkanes and ethers. J. Phys. Chem. B 2012, 116, 11234−11246.

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jpcb.7b10603. Simulation details and force field parameters (PDF)



REFERENCES

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. ORCID

Arash Nikoubashman: 0000-0003-0563-825X Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS We thank Prof. F. Müller-Plathe for helpful discussions. We further acknowledge funding from the German Research Foundation (DFG) under the project number NI 1487/2-1 and SFB TRR146, and also acknowledge the computing time 2136

DOI: 10.1021/acs.jpcb.7b10603 J. Phys. Chem. B 2018, 122, 2130−2137

Article

The Journal of Physical Chemistry B

(46) Ragnetti, M.; Geiser, D.; Höcker, H.; Oberthür, R. C. Small angle neutron scattering (SANS) of cyclic and linear polystyrene in toluene. Makromol. Chem. 1985, 186, 1701−1709. (47) Abe, F.; Einaga, Y.; Yoshizaki, T.; Yamakawa, H. Excludedvolume effects on the mean-square radius of gyration of oligo- and polystyrenes in dilute solutions. Macromolecules 1993, 26, 1884−1890. (48) Spychal, T.; Lath, D.; Berek, D. Thermodynamic and hydrodynamic properties of the the systems polymer tetrahydrofuran water: 1. Solution properties of polystyrene. Polymer 1979, 20, 437− 442. (49) Rawiso, M.; Duplessix, R.; Picot, C. Scattering function of polystyrene. Macromolecules 1987, 20, 630−648. (50) Huber, K.; Bantle, S.; Lutz, P.; Burchard, W. Hydrodynamic and thermodynamic behavior of short-chain polystyrene in toluene and cyclohexane at 34.5° C. Macromolecules 1985, 18, 1461−1467. (51) Šolc, K. Shape of a random-flight chain. J. Chem. Phys. 1971, 55, 335−344. (52) Rudnick, J.; Gaspari, G. The shapes of random walks. Science 1987, 237, 384−389. (53) Knoch, F.; Speck, T. Nonequilibrium Markov state modeling of the globule-stretch transition. Phys. Rev. E: Stat. Phys., Plasmas, Fluids, Relat. Interdiscip. Top. 2017, 95, 012503. (54) Schneider, J.; Panagiotopoulos, A. Z.; Müller-Plathe, F. Polymer chain collapse upon rapid solvent exchange: Slip-spring dissipative particle dynamics simulations with an explicit-solvent model. J. Phys. Chem. C 2017, 121, 27664−27673.

(23) Wick, C. D.; Martin, M. G.; Siepmann, J. I. Transferable potentials for phase equilibria. 4. United-atom description of linear and branched alkenes and alkylbenzenes. J. Phys. Chem. B 2000, 104, 8008−8016. (24) Harmandaris, V.; Adhikari, N.; van der Vegt, N. F.; Kremer, K. Hierarchical modeling of polystyrene: From atomistic to coarsegrained simulations. Macromolecules 2006, 39, 6708−6719. (25) Wu, Y.; Tepper, H. L.; Voth, G. A. Flexible simple point-charge water model with improved liquid-state properties. J. Chem. Phys. 2006, 124, 024503. (26) Degiacomi, M. T.; Erastova, V.; Wilson, R. M Easy creation of polymeric systems for molecular dynamics with Assemble! Comput. Phys. Commun. 2016, 202, 304−309. (27) Zimm, B. H. Dynamics of polymer molecules in dilute solution: viscoelasticity, flow birefringence and dielectric loss. J. Chem. Phys. 1956, 24, 269−278. (28) Zhang, C.; Pansare, V. J.; Prud’homme, R. K.; Priestley, R. D. Flash nanoprecipitation of polystyrene nanoparticles. Soft Matter 2012, 8, 86−93. (29) Anderson, J. A.; Lorenz, C. D.; Travesset, A. General purpose molecular dynamics simulations fully implemented on graphics processing units. J. Comput. Phys. 2008, 227, 5342−5359. (30) LeBard, D. N.; Levine, B. G.; Mertmann, P.; Barr, S. A.; Jusufi, A.; Sanders, S.; Klein, M. L.; Panagiotopoulos, A. Z. Self-assembly of coarse-grained ionic surfactants accelerated by graphics processing units. Soft Matter 2012, 8, 2385−2397. (31) Glaser, J.; Nguyen, T. D.; Anderson, J. A.; Lui, P.; Spiga, F.; Millan, J. A.; Morse, D. C.; Glotzer, S. C. Strong scaling of generalpurpose molecular dynamics simulations on GPUs. Comput. Phys. Commun. 2015, 192, 97−107. (32) Allen, M. P.; Tildesley, D. J. Computer simulation of liquids; Clarendon Press: Oxford, U.K., 1989. (33) Frenkel, D.; Smit, B. Understanding molecular simulation: From algorithms to applications; Academic Press: San Diego, 2001. (34) Nosé, S. A unified formulation of the constant temperature molecular dynamics method. J. Chem. Phys. 1984, 81, 511−519. (35) Hoover, W. G. Canonical dynamics: equilibrium phase-space distributions. Phys. Rev. A: At., Mol., Opt. Phys. 1985, 31, 1695−1697. (36) Martyna, G. J.; Tobias, D. J.; Klein, M. L. Constant pressure molecular dynamics algorithms. J. Chem. Phys. 1994, 101, 4177−4189. (37) Alfe, D.; Gillan, M. J. First-principles calculation of transport coefficients. Phys. Rev. Lett. 1998, 81, 5161−5164. (38) Das, B.; Roy, M. N.; Hazra, D. K. Densities and viscosities of the binary aqueous mixtures of tetrahydrofuran and 1, 2-dimethoxyethane at 298, 308 and 318 K. Indian J. Chem. Technol. 1994, 93−97. (39) Aminabhavi, T. M.; Gopalakrishna, B. Density, viscosity, refractive index, and speed of sound in aqueous mixtures of N, Ndimethylformamide, dimethyl sulfoxide, N, N-dimethylacetamide, acetonitrile, ethylene glycol, diethylene glycol, 1, 4-dioxane, tetrahydrofuran, 2-methoxyethanol, and 2-ethoxyethanol at 298.15 K. J. Chem. Eng. Data 1995, 40, 856−861. (40) Sedov, I.; Magsumov, T. Thermodynamic functions of solvation of hydrocarbons, noble gases, and hard spheres in tetrahydrofuranwater mixtures. J. Phys. Chem. B 2015, 119, 8773−8780. (41) Vasudevan, V.; Mushrif, S. H. Insights into the solvation of glucose in water, dimethyl sulfoxide (DMSO), tetrahydrofuran (THF) and N, N-dimethylformamide (DMF) and its possible implications on the conversion of glucose to platform chemicals. RSC Adv. 2015, 5, 20756−20763. (42) Smith, M. D.; Mostofian, B.; Petridis, L.; Cheng, X.; Smith, J. C. Molecular driving forces behind the tetrahydrofuran-water miscibility gap. J. Phys. Chem. B 2016, 120, 740−747. (43) Humphrey, W.; Dalke, A.; Schulten, K. VMD - Visual Molecular Dynamics. J. Mol. Graphics 1996, 14, 33−38. (44) Rubinstein, M.; Colby, R. H. Polymer physics; Oxford University Press: Oxford, U.K., 2003. (45) Venkataswamy, K.; Jamieson, A. M.; Petschek, R. G. Static and dynamic properties of polystyrene in good solvents: ethylbenzene and tetrahydrofuran. Macromolecules 1986, 19, 124−133. 2137

DOI: 10.1021/acs.jpcb.7b10603 J. Phys. Chem. B 2018, 122, 2130−2137