pubs.acs.org/Langmuir © 2009 American Chemical Society
Molecular Dynamics Study of the Melt Morphology of Polyethylene Chains with Different Branching Characteristics Adjacent to a Clay Surface Chunli Li, Phillip Choi,* and Michael C. Williams Department of Chemical and Materials Engineering, University of Alberta, Edmonton, Alberta, Canada T6G 2V4 Received September 10, 2009. Revised Manuscript Received November 26, 2009 Conformations of model high-density polyethylene (HDPE) and linear low-density polyethylene (LLDPE) chains with different intramolecular branch distributions adsorbed on a relaxed octahedral surface of kaolinite, a major clay mineral, at 463 K (190 °C) were studied by using classical molecular dynamics (MD) simulation. Prior to the MD simulations, first-principle density functional theory (DFT) calculations were carried out to relax the inorganic surface that was created by cleaving the corresponding kaolinite crystal structure. The high-temperature MD simulation results showed that an ordered polyethylene region with a thickness of about one to three layers of chain segments developed rapidly near the clay surface. On the other hand, chain segments in the far field slowly evolved into another ordered region with a higher degree of order than the one adjacent to the surface. It was observed that the melt morphology in the far field depends on the architecture of the chains. Also, in between the two ordered regions, a region that contained no apparent order formed. The above observation is attributed to the fact that the mobility of chain segments adjacent to the surface was greatly reduced as a result of their strong affinity for the surface, while those in the far field were not. Despite the fact that the results are for the melt state, they suggest that nucleation and lamellar growth of polymer chains nearby an inorganic surface may proceed from the chain segments in the ordered region in the far field rather than from the organic/inorganic interface. This is because chain segments in the three described regions, upon cooling, should not have sufficient thermal energy to reorient themselves drastically to form a single lamella under normal crystallization conditions. However, it should be noted that the above speculation is made based on a rather short equilibration time (∼10 ns) used in the simulations.
1. Introduction Owing to their low cost, light weight, and versatility, different types of clay minerals have been used as fillers to enhance physical and mechanical properties of polymers.1-7 Additionally, the property enhancement is attributed to the property of clay itself as well as to its influence on the crystalline morphology formed in the polymer matrix. As a result, in recent years, considerable research efforts have been focused on studying the conformational characteristics of polymer chains adjacent to clay surfaces to optimize how such surfaces should be chemically modified. Unfortunately, current available analytical techniques are still rather limited in terms of obtaining the desired information.1,5,8,9 At present, molecular simulation may be the only alternative for such a purpose. However, great challenges also exist in this approach. One major reason is that the relaxation time scales involved in such systems range from femtoseconds (i.e., bond vibrations in surface atoms of clay) to as long as microseconds *To whom correspondence should be addressed. E-mail: phillip.choi@ ualberta.ca. (1) Gao, D. G.; Ma, J. Z.; Lu, H.; Chu, Y.; Yang, Z. S. J. Compos. Mater. 2008, 42, 2805. (2) Phuong, N. T.; Gibert, V.; Chuong, B. J. Reinf. Plast. Compos. 2008, 27, 1983. (3) Powell, C. E.; Beall, G. W. Curr. Opin. Solid State Mater. Sci. 2006, 10, 73. (4) Ray, S. S. J. Ind. Eng. Chem. 2006, 12, 811. (5) Ruiz-Hitzky, E.; Aranda, P. In Polymer-Clay Nanocomposites; Beall, G. W., Eds.; John Wiley & Sons: New York, 2000. (6) Zaarei, D.; Sarabi, A. A.; Sharif, F.; Kassiriha, S. M. J. Coat. Technol. Res. 2008, 5, 241. (7) Buggy, M.; Bradley, G.; Sullivan, A. Composites, Part A: Appl. Sci. Manuf. 2005, 36, 437. (8) Wang, Y. L.; Lee, B. S.; Chang, K. C.; Chiu, H. C.; Lin, F. H.; Lin, C. P. Compos. Sci. Technol. 2007, 67, 3409. (9) Zhang, J. P.; Wang, A. Q. React. Funct. Polym. 2007, 67, 737.
Langmuir 2010, 26(6), 4303–4310
(i.e., relaxation of polymer chain segments), which make it infeasible to equilibrate the system completely using computing resources normally available in many simulation laboratories. Many research groups use ab initio methods to study the adsorption behaviors of organic molecules on inorganic surfaces, but they can only take into account several small molecules such as water, H2S, CO2, etc.,10-12 or one single molecule with a few alkyl chains.13 Coarse-grain models have also been employed with these systems by using united atoms to represent repeat units with different chemical compositions.14-16 However, in these systems, very simple adsorption surfaces such as uniform metal surfaces were used; therefore, detailed packing conformations and specific interactions at these surfaces cannot be explored. Recently, full atomic detailed simulations have also been employed to simulate organoclays and polymer nanocomposites.17,18 In particular, Sikdar et al.17 applied molecular dynamics (MD) simulation techniques to investigate the morphology, molecular interactions, and physical properties of organically modified montmorillonite (OMMT) and polymer clay nanocomposites. Their simulation results give useful information on the orientation (10) Alfe, D.; Gillan, M. J. J. Chem. Phys. 2007, 127, 114709. (11) Cosoli, P.; Ferrone, M.; Pricl, S.; Fermeglia, M. Chem. Eng. J. 2008, 145, 86. (12) Qin, Y.; Yang, X. N.; Zhu, Y. F.; Ping, J. L. J. Phys. Chem. C 2008, 112, 12815. (13) Dkhissi, A.; Esteve, A.; Jeloaica, L.; Esteve, D.; Djafari Rouhani, M. J. Am. Chem. Soc. 2005, 127, 9776. (14) Baschnagel, J.; Meyer, H.; Varnik, F.; Metzger, S.; Aichele, M.; Muller, M.; Binder, K. Interface Science 2003, 11, 159. (15) Delle Site, L.; Abrams, C. F.; Alavi, A.; Kremer, K. Phys. Rev. Lett. 2002, 89, 15610-1. (16) Delle Site, L.; Kremer, K. Int. J. Quantum Chem. 2005, 101, 733. (17) Sikdar, D.; Katti, D. R.; Katti, K. S. Langmuir 2006, 22, 7738. (18) Zeng, Q. H.; Yu, A. B.; Lu, G. Q. Int. J. Nanotechnol. 2008, 5, 277.
Published on Web 12/29/2009
DOI: 10.1021/la903425z
4303
Article
Li et al.
and area of coverage of organic modifiers over the interlayer clay surface, as well as interactions between organic modifiers and clay. Their simulations were verified by X-ray diffraction and photoacoustic Fourier transform infrared (FTIR) results. However, it is worth noting that all the simulation time in their research was as short as 200 ps. Additionally, Zeng et al.18 studied the interfacial interactions and layering structure of organoclays and polyurethane nancomposites by using MD simulation techniques. Within about 1 ns of MD simulation time, they made quantitative analyses in atom density distribution, orientations, and mobility of the polymer chains on the clay surface, and their simulation results claimed good agreement with various experimental measurements. On the other hand, in our previous studies,19 we successfully combined ab initio density functional theory (DFT) and classic MD techniques to investigate adsorption behaviors of alkane chains on a relaxed R-alumina surface and presented a solution to this problem. In our approach, DFT calculations were carried out first to relax the R-alumina surface fully, and the positions of all the aluminum and oxygen atoms in the relaxed R-alumina (0001) surface were then fixed during the subsequent classical MD simulation. Our simulation results showed that such an approach not only reveals the conformation of the adsorbed alkane segments but also can reproduce the heat of adsorption of alkanes with different chain lengths reasonably well. Therefore, in the present work, we apply the same approach to study the adsorption behavior of high-density polyethylene (HDPE) and linear low-density polyethylene (LLDPE) chains with different intramolecular branch distributions at a temperature (190 °C) significantly higher than their melting temperatures. The motivation was originated from our previous studies on the effects of branching characteristics on the melt and solid state morphologies of LLDPE chains with different inter- and intramolecular branch distributions.20,21 Nevertheless, in the present work, only the melt state at 190 °C was of interest. This is because, as demonstrated in our previous work, the melt state morphology determines, to a certain extent, the solid state morphology formed at low temperatures, as chains do not have sufficient thermal energy to reorient themselves drastically during the cooling process. As a result of the computational cost, only five chains with the same chain length were used to represent the melt state of each polyethylene model. However, equilibrated structures of HDPE and various types of LLDPE chains surrounded by vacuum and in their bulk melt states were also obtained to illustrate how the chosen clay surface affects the self-arrangement behaviors of those polyethylene chains.
2. Simulation Models and Methods 2.1. Relaxation of the Octahedral Surface of Kaolinite. Since kaolinite is a major mineral found in clay, kaolinite surfaces were used to represent the clay surface in the present work. Kaolinite consists of a silicate Si2O5 tetrahedral sheet bonded to a gibbsite-type (Al(OH)3) octahedral layer. The gibbsite octahedral layer is terminated with three different hydroxyl groups sticking out of the surface and another inner hydroxyl group almost parallel to the layer (Figure 1a). The unit cell parameters of a kaolinite single crystal from the Commercial Software Materials Studio library are a = 5.149 A˚, b = 8. 934 A˚, c = 7.384 A˚, R = 91.93°, β = 105.04°, γ = 89.79°. This structure is consistent with that obtained from neutron powder diffraction experiments.22 (19) (20) (21) (22)
Li, C.; Choi, P. J. Phys. Chem. 2007, 111, 1747. Zhang, M.; Yuen, F.; Choi, P. Macromolecules 2006, 39, 8517. Li, C.; Choi, P. Macromolecules 2008, 41, 7109. Young, R. A.; Hewat, A. W. Clays Clay Miner. 1988, 36, 225.
4304 DOI: 10.1021/la903425z
Figure 1. Structures of the octahedral surface of kaolinite cleaved from the Software Material Studio database, with white spheres representing hydrogen atoms in hydroxyl groups (a) before relaxation and (b) after relaxation.
Figure 2. HDPE and LLDPE chain architectures used in this work.
The (001) octahedral surface of kaolinite is the basal plane that is predominately exposed23,24 and is the surface of primary interest in adsorption studies. In our study, the (001) surface containing three layers of aluminosilicate was first cleaved computationally from kaolinite bulk structure, and then the atomic positions in aluminosilicate layers were relaxed by using the DFT method built into the DMol3 code of the Materials Studio software.25 During the geometry relaxation, only the first layer was fully relaxed, while the other two layers were kept fixed. Exchange and correlation were treated with the generalized gradient approximation GGA-PW91. The k-point set was 4 4 4, and the core treatment was set to all electrons. 2.2. MD Simulation. In this work, a model HDPE and three LLDPE models with different intramolecular branch distributions (end, middle, and random) were used. The architectures of these four types of chains are depicted in Figure 2. Here, the endtype chains contain branches clustered near chain ends, while the middle type chains contain branches clustered around the chain middles. In the case of the random type chains, branches (23) Zbik, M.; Smart, R. Miner. Eng. 2002, 15, 277. (24) Nesse, W. D. Introduction to Mineralogy; Oxford University Press: New York, 1999. (25) Accelrys Software, Inc. Materials Studio User Manual; Accelrys, Inc.: San Diego, CA, 2007.
Langmuir 2010, 26(6), 4303–4310
Li et al.
Article
Figure 3. Density profiles of model HDPE and LLDPEs along the normal direction of the octahedral surface of kaolinite: (a) HDPE; (b) end type; (c) middle type; (d) random type.
distribute randomly along the backbones. All three types of LLDPE model chains possessed the same average branch content of branches per 1000 backbone carbons. Each model chain (for all types of chains) contained 300 methylene units, and each branched chain (for all types of branched chains) contained 3 hexyl branches. For those chains with branches clustered at the chain end or chain middle, the branches were 10 methylene units apart. For the end-type LLDPE, the branch closest to the chain end was connected to the sixth backbone carbon atom, giving a “hexyl end”. Following the approach used in our previous work, all model chains were modeled with implicit hydrogen and the use of the Dreiding force field along with the use of the united atom Lennard-Jones parameters of Rychaert and Bellemans.20,21,26 It is worth noting that no electrostatic interaction was included in the MD calculations. The partial atomic charges of the united carbon atoms were set to zero. All MDs simulations were carried out by using the commercial software Cerius2.25 There are two reasons that we did not and could not include charges in the polyethylene models. First of all, we would like to compare the liquid structures of the model polyethylene chains in the presence of the clay surface to those in vacuum and in the melt state. Since we did not use charges for the polyethylene models in vacuum and in the melt state, we decided not to use charges on the polyethylene models adjacent to the clay surface. It is worth noting that we were able to reproduce the liquid structure of polyethylene in the melt state without charges.20 The other reason is that it is difficult, probably impossible, to assign proper charges to the united atoms (no explicit hydrogen atoms) in the polyethylene models. Given the fact that our main interest is to study the effect of branching characteristics on the liquid structures of various types of (26) Rychaert, J. P.; Bellemans, A. Chem. Phys. Lett. 1975, 30, 123.
Langmuir 2010, 26(6), 4303–4310
Table 1. Self Diffusion Coefficients of the Centers of Mass of Chain Segments in Different Regions in the Layer of Polyethylene Melt above the Kaolinite Surface (10-6 cm2/s)
linear type end type middle type random type
adsorption region
bulk region
vacuum interface
3.98 5.49 1.66 1.77
4.92 6.74 2.94 1.78
6.45 10.53 6.18 2.60
polyethylene chains on the clay surface, we needed to use long chains. This in turn excluded us from using explicit hydrogen atom models as these models would contain three times more atoms. Regarding proper charge assignments, taking the linear polyethylene model as an example, it is obvious that all united atoms (i.e., CH2) should have the same charge, and the total charge of the whole polyethylene molecule should be zero. The most logical choice of the charge on all united atoms should be zero. Kaolinite super cells composed of 5 3 unit cells were employed as the substrate. For each model, five entangled chains were randomly placed onto the (001) octahedral surface. Therefore, the resulting systems contain a kaolinite substrate layer and a polymer layer, both subjected to two-dimensional periodic boundary conditions. The resulting systems (i.e., HDPE or LLDPE chains plus the clay surface) were then energy-minimized with a convergence criterion of 0.1 kcal/mol A˚. After that, MD simulations were carried out at 463 K (190 °C), a common processing temperature for polyethylene, for as long as 10 ns for the systems to equilibrate. For the energy minimization and MD simulations, all the clay atoms in the relaxed (001) octahedral structure were fixed. The rationale for fixing the positions of the clay atoms is that the frequency of the vibration of the clay atoms, on the order of femtoseconds, is by far much faster than that of the relaxations DOI: 10.1021/la903425z
4305
Article
Li et al.
Figure 4. Equilibrated structures in vacuum of model HDPE and LLDPEs (five chains in each model) at 463 K: (a) HDPE; (b) end type; (c) middle type; (d) random type.
of the polyethylene (alkane) chains of interest, on the order of nanoseconds.27,28 The Nose canonical method along with the velocity-Verlet algorithm and a time step of 1 fs was used for generating the MD trajectories. Simulations of models with different initial conformations and branch distributions were carried out to ensure the reproducibility of simulation results obtained in this research (results not shown here). Additionally, for comparison, simulations of both HDPE and LLDPE chains in vacuum were also carried out at 463 K. Five LLDPE chains were also used in such simulations. In addition, our previous simulation results on various types of LLDPE models in the bulk liquid state at 463 K were also used for comparison purposes. However, in the previous work, 10 rather than 5 LLDPE chains were used.
3. Results and Discussion 3.1. Relaxation of the Octahedral Surface of Kaolinite. The structure of the octahedral surface of kaolinite subjected to the DFT relaxation is shown in Figure 1b. Upon relaxation, one hydroxyl group on the surface (group 2 shown in Figure 1) oriented more or less horizontally in the plane of the surface. This result is consistent with the computational observation of Benco and co-workers.29 Considering the fact that most publications on the relaxation of the octahedral surface of kaolinite surface only consider a single kaolinite layer30,31 and that, in our case, a cutoff distance of about 10 A˚ was used in handling the van der Waals interactions, three layers of aluminosilicate with a total (27) Lodziana, Z.; Norskov, J. K. J. Chem. Phys. 2003, 118, 11179. (28) Gao, G. H. Bull. Korean Chem. Soc. 2002, 23, 1595. (29) Benco, L.; Tunega, D.; Hafner, J.; Lischka, H. J. Phys. Chem. B 2001, 105, 10812. (30) Michalkov, A.; Tunega, D. J. Phys. Chem. 2007, 111, 11259. (31) Hu, X. L.; Michaelides, A. Surf. Sci. 2008, 602, 960.
4306 DOI: 10.1021/la903425z
thickness of about 20 A˚ should be thick enough for the subsequent MD simulations. Here, given the fact that all the σ values for various types of atoms involved in the simulations are less than 4 A˚, we chose a 10 A˚ cutoff distance according to the 2.5σ rule. 3.2. Equilibration of Melt Adjacent to the Clay Surface. Unlike melt-state MD simulations in which the end-to-end vector decorrelation function and/or mean square displacement of the centers of mass of the chains can be used to indicate the degree of equilibration, we decided to run the MD simulations as long as we could to achieve “equilibration”. In all cases, we used about 10 ns, a simulation time that is much longer than the equilibration time for comparable systems in the absence of an inorganic surface and that was believed to be long enough for the formation of the melt morphology near the surface.19,20 Figure 3a-d shows plots of the number density distribution of backbone carbon atoms in different polyethylene models versus the vertical distance along the direction perpendicular to the (001) octahedral surface. As can be seen from the figure, the number of chain segments (the first sharp peak in Figure 3a-d) adsorbed on the surface (within 10 A˚ from the surface) becomes quite stable in the early stage of simulation (∼500 ps), while that at distances further away from the surface (10-50 A˚ from the surface) still fluctuates considerably even after long times. Calculated self-diffusion coefficients of chain segments in different regions (see Table 1 and discussion below) show that the mobility of adsorbed chain segments is considerably lower than that of those near the free surface (or in the bulk). And this is one of the main reasons why we did not use the aforementioned equilibration indicators in this work. The difference in the mobility of the polyethylene chains near the surface and further away from the surface led to fairly unique melt morphology of such chains in the presence of the clay surface that will be discussed in details below. One noteworthy point is that, upon energy minimization, all model chains were partly pulled to the Langmuir 2010, 26(6), 4303–4310
Li et al.
Article
Figure 5. Equilibrated structures of model HDPE and LLDPEs at 463 K after about 6 ns of MD simulation in the melt state: (a) HDPE; (b) end type; (c) middle type; (d) random type.
surface, indicating that such chains exhibited a relatively high affinity for the surface. The polymer chains were pulled even closer to the surface during the MD simulations. 3.3. Equilibration of Melt. The equilibrated structures of HDPE and three LLDPE models reported in the previous section but simulated in vacuum (i.e., in the absence of the clay surface) are depicted in Figure 4. It was observed that after about 1 ns of MD simulation at 463 K, they all formed a compacted cluster. Figure 5 shows the corresponding equilibrated structures in their melt state (subjected to three-dimensional periodic boundary conditions). It is obvious that such equilibrated structures contain a substantial amount of local order compared to those shown in Figure 4. It was somewhat expected as Kilian and Piper reported such local order in HDPE melt in their X-ray diffraction experiment more than a decade ago,32 and this issue has also been addressed in our previous work.20 In our MD simulations, such local order was observed to form at about 6 ns. Moreover, close inspection showed that, in the vacuum and melt state simulations, branches on different chains tended to distribute spatially isotropically. 3.4. Melt Morphology Adjacent to the Clay Surface. Figure 6 shows the initial structure of the model HDPE and its structure after 500 ps, 1000 ps, and 10.7 ns of MD simulation, respectively. Figure 6d clearly shows that there exist two layers of chain segments adsorbed on the (001) octahedral surface, which can also be readily identified from the density profile depicted in Figure 3a. In this ordered region, a majority of the adsorbed chain (32) Kilian, H. G.; Pieper, T. Adv. Polym. Sci. 1993, 108, 49.
Langmuir 2010, 26(6), 4303–4310
segments have molecular planes parallel to the surface, and their molecular axes tilt 45° with respect to the edges of the surface. On the other hand, chain segments near the free surface (i.e., chain segments further away from the clay surface) also formed ordered structure. However, such ordered structure contains much longer chain segments (an average of 66 backbone carbon atoms) compared to the one adjacent to the clay surface (15-26 backbone carbon atoms). Although there are four layers of chain segments in the ordered structure formed near the free surface, they do not manifest themselves in the density profile plot, as the structure did not align horizontally with respect to the surface. It is interesting to point out that, in between these two order regions, there exists a region that appears to contain less order. Selfdiffusion coefficients of center of mass of chain segments in different regions are presented in Table 1. In particular, these regions are the adsorption region, bulk region and vacuum interface, respectively. The size of each region is depicted in Figure 3. Table 1 shows that the mobility of chains in the two aforementioned ordered regions differed considerably, leading to a situation in which the two ordered regions formed at different times (the one adjacent to the surface formed first) and did not align with each other. In the case of the end-type LLDPE chains, there exists only one layer of adsorbed chain segments (Figure 3b). Figure 7 depicts their (a) initial structure, (b) structure at 500 ps, (c) structure at 1000 ps, and (d) structure at 12.5 ns. Within the adsorption layer, one interesting observation is that the molecular planes of the adsorbed segments exist alternatively perpendicular or parallel to DOI: 10.1021/la903425z
4307
Article
Li et al.
Figure 6. Structures of HDPE melt on the (001) octahedral surface: (a) initial structure; (b) at 500 ps; (c) at 1,000 ps; (d) final structure at 10.7 ns; (e) top view of the first adsorption layer at 10.7 ns.
Figure 7. Melt structures of the end-type LLDPE on the (001) octahedral surface: (a) initial structure; (b) at 500 ps; (c) at 1000 ps; (d) final structure at 12.5 ns; (e) top view of the first adsorption layer at 12.5 ns.
Figure 8. Melt structures of the middle-type LLDPE on the (001) octahedral surface: (a) initial structure; (b) at 500 ps; (c) at 1000 ps; (d) final structure at 10.2 ns; (e) top view of the first adsorption layer at 10.2 ns.
the (001) octahedral surface. Additionally, the numbers of perpendicular and parallel chain segments are comparable. The number of backbone carbon atoms involved within each adsorbed segment varies from about 24 to 46. And the molecular axes of these adsorbed segments tilted 45° with respect to the edges of the surface. Ordered structure was also formed by the chains in the far field. The length of the chain segments involved in such order consists of more than 36 backbone atoms. Another 4308 DOI: 10.1021/la903425z
noteworthy point is that the branches of different chains aggregated together and formed a less-ordered region that was sandwiched between the aforementioned ordered structures. In addition, simulations of models with different initial conformation and branch distributions show similar final morphologies with two distinguishable ordered regions. LLDPE chains with branches at the chain middles also show only one obvious adsorption layer, as depicted in Figure 8. Langmuir 2010, 26(6), 4303–4310
Li et al.
Article
Figure 9. Melt structures of the random type LLDPE on the (001) octahedral surface: (a) initial structure; (b) at 500 ps; (c) at 1000 ps; (d) final structure at 10.0 ns; (e) top view of the first adsorption layer.
Figure 10. Orientation of the molecular axes of the first melt adsorption layer of all LLDPE models on the substrate kaolinite surface. For each model, only the type 2 hydroxyl groups in Figure 1 are shown for clarification. (a) Linear type; (b) end type; (c) middle type; (d) random type. Images shown correspond to a 2 2 super cell of the original simulation cells.
However, the number of backbone carbon atoms involved in the adsorption layer is significantly lower than that of HDPE and of the other two types of LLDPE chains. Within the adsorption layer, the molecular plane of most chain segments tended to orient parallel to the substrate surface. What is more interesting is that the molecular axes of these adsorbed segments were orthogonal to each other and tilted at about 45° and -45° with respect to the edges of the surface. They did not form hairpin structure like the Langmuir 2010, 26(6), 4303–4310
other models. As a result, most of these chain segments consisted of only about 13 backbone carbon atoms. Figure 9 shows (a) the initial structure, (b) structure at 500 ps, (c) structure at 1,000 ps, (d) final structure, and (e) top view of the first adsorption layer of LLDPE chains with branches distributed randomly along the chains. This type of LLDPE chains shows three obvious adsorption layers adjacent to the surface. However, no obvious ordered structures formed in the far field. Similar to DOI: 10.1021/la903425z
4309
Article
the structures of HDPE and the end-type LLDPE, the orientation of the adsorbed chain segments tilted about 45° with respect to the edges of the surface. However, molecular planes of the adsorbed segments oriented alternatively perpendicular or parallel to the (001) octahedral surface. The lengths of those segments were also about 24 backbone carbon atoms. 3.5. Structure of the First Adsorption Layer. Compared to the equilibrated structures of model HDPE and LLDPE chains in vacuum and in the melt state, the octahedral surface of kaolinite definitely altered the morphology of such chains. The surface also imposed a certain orientation to the first couple of layers of adsorbed chain segments. Except for the middle-type LLDPE model, the molecular axes of the adsorbed chain segments of the other models tilted about 45° with respect to the edges of the surface, and a hairpin structure was formed (Figure 10a,b,d). In the case of the middle-type LLDPE, no hairpin structure was found, and short adsorbed chain segments oriented in an orthogonal pattern (Figure 10c). The observed orientation of the adsorbed chain segments may be due to the commensuration of such segments with the structure of the surface atoms in the octahedral surface. Figure 10 shows the orientation of the molecular axes of the first adsorption layer of all models with respect to the group 2 hydroxyl moiety depicted in Figure 1. It seems that orientation of those adsorbed segments prefer to arrange in a direction similar to those of the group 2 hydroxyl moiety, although there exist some differences among different models. As a result, the ordered regions formed in these models in the far field also differ as discussed earlier. Another observation was that, regardless of the types of LLDPE chains, their branches were not able to distribute uniformly, at least over the MD simulation times used. This is once again related to the reduced mobility of the chains. Given the fact that chains would not have sufficient thermal energy to reorient themselves drastically, especially the first few layers of adsorbed chain segments (i.e., lower diffusivity of adsorbed chain segments compared with those in bulk and at vacuum interface presented in Table 1), when they are subjected to cooling, it is speculated that lamellae would form from the ordered region that is further away from the surface rather than at the surface. Furthermore, compared to HDPE and other types of LLDPEs, the middle-type LLDPE may not adhere strongly to the
4310 DOI: 10.1021/la903425z
Li et al.
surface as there exist a considerably smaller number of adsorbed carbons and no hairpin structure. We present these speculations for experimental verification.
4. Conclusions Melt morphology of chains of model HDPE and model LLDPE with different intramolecular branch distributions nearby an octahedral surface of kaolinite, a common clay mineral, was studied by combining ab initio DFT and classical MD simulation techniques. Comparing with equilibrated structures of the corresponding polyethylene models in the vacuum and bulk melt state, it seems that the clay surface facilitates the formation of two regions of ordered structures with specific orientation. And in between them, there exists a region with less apparent order. Furthermore, the characteristics of such ordered structures are distinct in different models. The formation of multiregion orders is attributed to the fact that chains in the interaction range (∼1 nm) of the surface tended to adsorb on the surface rapidly in the early stage of the simulations and lost their mobility. In addition, such an ordered region tended to align according to the orientation of the atoms in the kaolinite surface. Since chains in the far field tended to have higher mobility than those next to the surface, they evolved slowly to become more ordered structures with different orientations than the one near the surface. The results suggest that nucleation and lamellar growth of such chains may not start at the organic/inorganic interface. Rather, lamellae may only nucleate from the ordered region in the far field. However, it should be noted that the above speculation is made based on a rather short equilibration time (∼10 ns) used in the simulations. Acknowledgment. The authors would like to thank Mr. Xiao Ni for doing the simulation work on the relaxation of the octahedral surface of kaolinite. Funding from the Natural Science and Engineering Research Council of Canada is also gratefully acknowledged. This research has been enabled by the use of WestGrid computing resources, which are funded in part by the Canada Foundation for Innovation, Alberta Innovation and Science, BC Advanced Education, and the participating research institutions. WestGrid equipment was provided by IBM, HewlettPackard, and SGI.
Langmuir 2010, 26(6), 4303–4310
