Ultrathin Molecular-Layer-by-Layer Polyamide Membranes: Insights

Aug 25, 2016 - Physicochemical properties such as density, free volume, thickness, the degree of cross-linking, atomic compositions, and average molec...
1 downloads 11 Views 3MB Size
Download PDF
Article pubs.acs.org/JPCB

Ultrathin Molecular-Layer-by-Layer Polyamide Membranes: Insights from Atomistic Molecular Simulations Thilanga P. Liyana-Arachchi,†,§ James F. Sturnfield,‡ and Coray M. Colina*,†,§ †

Department of Materials Science and Engineering, The Pennsylvania State University, University Park, Pennsylvania 16802, United States ‡ Engineering and Process Sciences, Process Optimization, The Dow Chemical Company, North Brazosport Boulevard, Freeport, Texas 77541, United States S Supporting Information *

ABSTRACT: In this study, we present an atomistic simulation study of several physicochemical properties of polyamide (PA) membranes formed from interfacial polymerization or from a molecular-layer-by-layer (mLbL) on a silicon wafer. These membranes are composed of meta-phenylenediamine (MPD) and benzene-1,3,5-tricarboxylic acid chloride (TMC) for potential reverse osmosis (RO) applications. The mLbL membrane generation procedure and the force field models were validated, by comparison with available experimental data, for hydrated density, membrane swelling, and pore size distributions of PA membranes formed by interfacial polymerization. Physicochemical properties such as density, free volume, thickness, the degree of cross-linking, atomic compositions, and average molecular orientation (which is relevant for the mLbL membranes) are compared for these different processes. The mLbL membranes are investigated systematically with respect to TMC monomer growth rate per substrate surface area, MPD/TMC ratio, and the number of mLbL deposition cycles. Atomistic simulations show that the mLbL deposition generates membranes with a constant film growth if both the TMC monomer growth rate and MPD/TMC monomer ratio are kept constant. The film growth rate increases with TMC monomer growth rate or MPD/TMC ratio. Furthermore, it was found on one hand that the mLbL membrane density and free volume varies significantly with respect to the TMC monomer growth rate, while on the other hand the degree of cross-linking and the atomic composition varies considerably with the MPD/TMC ratio. Additionally, it was found that both TMC and MPD orient at a tilted angle with respect to the substrate surface, where their angular distribution and average angle orientation depend on both the TMC growth rate and the number of deposition cycles. This study illustrates that molecular simulations can play a crucial role in the understanding of structural properties that can empower the design of the next generation RO membranes created from molecular-layer-by-layer (mLbL) on a silicon wafer.

1. INTRODUCTION Although current membrane processes are highly effective and efficient in producing clean water, the demands for meeting the needs of human consumption, agricultural usage, and industrial requirements worldwide require even better performance of membranes that can be used for a wider range of challenging situations.1 There is a range of advanced materials and separation processes that are being explored to meet these needs, but there are numerous challenges to utilize these approaches in commercial production.2 Part of the difficulty in moving forward with many of the promising proposals is the inability to predict the complete behavior of these new (or modified) materials in production scale application.1 The reverse osmosis (RO) membranes based on polyamide (PA) thin film composites represent one of the most energy efficient membrane technologies used today for producing clean water from seawater and brackish water.3−6 These thin film cross-linked PA-RO membranes are synthesized via interfacial polymerization (IP) of benzene-1,3,5-tricarboxylic acid chloride (TMC) and m-phenylenediamine (MPD) as a coating on a microporous polymeric support. The thin film © XXXX American Chemical Society

coating is formed on the support layer which has been soaked in a water solution of one monomer with an organic solvent containing the other monomer applied on top of it.3,6 The TMC and MPD monomers polymerize at the organic/aqueous interface7 that typically creates PA films with a large amount of variation in thickness.8 The typical RO membranes synthesized via IP can have large variations in many of the typical structural characterizations. In particular, IP membranes vary greatly over (1) film thickness, (2) surface roughness, (3) degree of cross-linking, and (4) local chemical composition which influence membrane performance.9,10 These variations present great challenges in characterizing the membrane, in producing a realistic model, and in the understanding of the role that various structural elements play in the membrane performance. 11 To overcome these challenges, ultrathin membranes synthesized via molecularlayer-by-layer (mLbL) deposition have been proposed.11−20 Received: March 17, 2016 Revised: July 30, 2016

A

DOI: 10.1021/acs.jpcb.6b02801 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B

understanding the impact of modifications of interfacially produced membranes for advances in industrial applications. Virtual synthesis and characterization of a realistic molecular polymeric membrane model that can reproduce specific experimental properties is a challenging task. In reality, polymerization takes place over an extended period of time (i.e., seconds, minutes), and thus, “brute force” atomistic molecular dynamic simulations are not a viable route to use with the current computational resources to mimic polymerization. However, atomistic detail is still necessary to understand several challenges of fundamental and industrial interest, and thus, it is necessary to use alternative methods and molecular simulation techniques to mimic the formation of a polymeric membrane. Previous molecular simulation studies have investigated IP-PA membranes composed of MPD and TMC.21−31 The techniques employed in these studies for structure generation include the following: (1) solvating polymer chains in bulk water, (2) growth of a single polymer chain in a periodic simulation box using a lattice distribution technique, (3) packing equal length polymeric chains in a periodic simulation box using a self-avoiding random walk approach, or using the configuration bias Monte Carlo technique, and (4) inserting monomer repeat units inside a periodic simulation box matching the experimental density and bonding reactive atoms within a cutoff distance. Most of these studies focused on water/salt dynamics in polymer structures (water/polymer interfacial systems) where periodic boundary conditions are applied in two dimensions with a finite thickness less than 17 nm.21−23,25−31 In previous work, polymer structures were either hydrated by inserting water and salt ions matching experimental data or combined with bulk salt/ water slabs to construct polymer/water interfacial systems. Furthermore, Kotelyanskii et al.24 investigated water/salt dynamics in polymeric structures with periodic boundary conditions applied in all three dimensions. It is important to note that, in this work, IP-PA membranes were investigated in order to validate the molecular models used for mLbL membranes. The main focus of this study is mLbL structure generation and investigating key membrane structural properties with respect to membrane variable. In this work, we first applied the Polymatic simulated polymerization algorithm32,33 developed in our group to virtually synthesize IP-PA “bulk” membranes composed of MPD and TMC. Polymatic has been validated for virtually synthesizing commodity polymers such as polystyrene and poly methyl methacrylate, as well as amorphous nanoporous polymers such as polymers of intrinsic microporosity, pyrenebased conjugated microporous polymers, and thermally crosslinked octene-styrene-divinylbenzene polymers among others.34−39 In previous studies, it has been demonstrated that the virtual polymeric samples are capable of reproducing experimental data for density (without setting a target density value), structure factors, BET surface area, permeability, fractional free volume, pore size distributions, adsorption isotherms, and enthalpies of adsorption in a predictive fashion.34−39 After IP-PA “bulk” membranes were synthesized, we compared key structural properties such as density, pore size distribution, and swelling against available experimental and simulation data. After this model validation, we applied the same model and the simulated polymerization algorithm to synthesize membranes, via mLbL deposition on silicon substrates, mimicking the experimental mechanisms. It is

mLbL membranes are built one molecular layer at a time using alternative monomers via deposition on a substrate, reaction, and removal of excess solvated monomers. During each cycle, the substrate with the growing polymer is exposed to an excess of monomer that pushes the reactivity toward completion during each cycle. This leads to (1) control over the film thickness, (2) reduction in surface roughness, (3) control over local chemical composition, and (4) homogeneous membrane structures, which is essential in understanding the relationship of physicochemical properties of the membrane to its performance. Previous works have established that mLbL membranes can be used to synthesize ultrathin films via alternative reactions using monomers employed in IP such as m-phenylenediamine/benzene-1,3,5-tricarboxylic acid chloride11−16 and p-phenylenediamine/terephtaloyl chloride.20 Furthermore, it has been shown that PA-mLbL membranes with 15 cycles of the two monomers can achieve competitive flux and salt rejection compared to interfacially produced RO commercial membranes.16 In short, PA-mLbL membranes can provide the performance of a commercial RO membrane and can be well characterized structurally, and thus, these membranes are ideal for characterizing the relationship of the membrane structure with their transport properties. Although mLbL membranes have been studied in previous work, there are still technical challenges in both the synthesis and the analysis of the membrane. The current synthesis work has confirmed the reduction of surface roughness and the ability to target a specific film thickness by mLbL deposition, but the number of layers needed to target a specific thickness is also seen to be very sensitive to a variety of factors. In particular, experimental efforts have demonstrated that thickness, atomic composition, and degree of cross-linking for these mLbL membranes can vary significantly depending on the substrate surface (e.g., silicon wafer plasma treated to generate hydroxyl groups, gold coated silicon wafer, poly(acrylic acid) polymer support, and poly(vinyl alcohol) coated on silicon wafer), the solvents used (toluene, acetone, isopropanol, and methanol), and a number of other experimental variables. These important membrane properties can lead to significant variations in water permeability and salt rejection for RO separation applications. The ultrathin membranes present additional challenges for determining the detailed atomic structure and chemical composition of membranes with a variety of standard analytical techniques because of issues with the signal-to-noise ratio and interfering signals from the support layer. There is a crucial necessity for understanding both structural and chemical properties for ultrathin membranes, including among them variables such as TMC growth rate, MPD/TMC ratio, and number of deposition cycles. The understanding of how these variables can lead to variations in membrane physicochemical properties (such as density, molecular orientation, thickness, the degree of cross-linking, and atomic compositions) will help in designing the next generation of materials. Atomistic molecular simulations can be used to examine local structural and chemical properties at a molecular level that support, complement, and enhance these experimental methods to drive the discovery of new materials. Specifically, computational analysis can provide insights into important structural information such as thickness, molecular orientations, local densities, the degree of cross-linking, and chemical compositions among others. These insights are not only important to understanding mLbL membranes but also to B

DOI: 10.1021/acs.jpcb.6b02801 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B

generalized Amber force field (GAFF),51 and water molecules were modeled using the TIP4P/2005 potential.52 The nonbonded potential uses a shifted LJ 12-6 potential, where LJ interaction energies that have a pair distance larger than 15 Å are set to zero. Furthermore, the Lorentz−Berthelot combining rule was used to determine the LJ parameters for unlike interactions.53 Long-range electrostatics were calculated with the particle−particle particle−mesh solver.54 Partial charge distributions for monomers were calculated using ab initio calculations for a monomer model in the gas phase at the HF/ 6-31G* level of theory followed by a restrained electrostatic potential (RESP) charge fitting procedure.55,56 Partial charges for repeating molecular segments are provided in the Supporting Information (see Table S1). The Gaussian 03 software program was used for all RESP charge derivations,57 and the LAMMPS58 program was used for both energy minimization and molecular dynamics (MD) simulations with a time step of 1 fs using the velocity-Verlet algorithm. Finally, it is important to mention that the all-atom optimized potential for liquid simulation (OPLS-AA)59 and the generalized Amber force field (GAFF)51 were also evaluated for a more limited set of simulations, as discussed in the Results and Discussion section. 2.2. Virtual Synthesis of Interfacial Polymerized Polymeric Membranes. The Polymatic simulated polymerization algorithm and a compression and decompression scheme were employed to generate IP-PA membranes.32,33 It is important to note that, in this work, IP-PA membranes were only investigated in order to validate the molecular models later used to create mLbL membranes. The main focus of this study is the structure generation of these mLbL membranes and their characterization through key structural properties. The IP membrane generation performed in this study consists of four steps: (1) Monomers were packed randomly into a simulation box under periodic boundary conditions at a low density of 0.3−0.4 g/cm3 followed by an energy minimization and a short 10 ps MD simulation at 300 K. (2) Polymerization via Polymatic at low density. (3) Compression and decompression of the polymerized sample, “21-steps”. (4) Hydration of the dry polymer membrane with water. The initial periodic box was randomly packed with 249 monomers of MPD-L, 102 monomers of TMC-L, and 98 monomers of TMC-C. These molecular ratios were chosen such that the ratio of cross-linked to linear segments in the final polymerized membrane was 49:51 as reported experimentally by the work of Kim et al.60 The polymerization steps were performed between “reactive atoms” on neighboring monomers within a cutoff distance of 6.25 Å. This arbitrary cutoff distance was selected, since it corresponds to the first minimum observed after the first maximum in the radial distribution function between reactive atoms, and allowed the formation of long polymeric chains. During the polymerization step, additional charges on reactive atoms (qpolym = ±0.5e) were used, as described in ref 33 to accelerate the polymerization process. The virtual polymerization was carried out until no more bonds could be found within 6.25 Å for a period of at least 500 ps. Next, the polymerized systems were compressed and decompressed using the “21-step” MD protocol with Tmax = 1000 K, Pmax = 5 × 104 atm, Tfinal = 300 K, and Pfinal = 1 atm.34 Similar to previous work, dry polymeric membranes were hydrated by inserting a number of water molecules corresponding to 23 wt % water.21,22,24 This water content was chosen according to reported experimental data for the TMC/MPD polymeric

worth noting that, although there are studies that have investigated polyamide chain/carbon nanotube systems, fluids confined in polyamide-6,6, polyamide-6,6 on graphene sheets and confined between graphene sheets,40−43 to our knowledge, there are no previous simulation work on PA-mLbL membranes (or any other mLbL membranes), and thus, this work represents the first systematic molecular simulation study of PA-mLbL membranes composed of MPD and TMC. First, the membrane is built one molecular layer at a time via substrate deposition, “bonding”, and removal of excess solvated monomers. We utilize Polymatic to form networks by connecting repeat units of TMC and MPD monomers deposited on a silicon substrate. The resulting simulated structures are characterized with respect to (a) TMC monomer growth rate, (b) MPD/TMC ratio, and (c) the number of deposition cycles. The simulation results presented cover key structural and chemical properties, such as film growth rate, density, free volume, molecular orientation, the degree of crosslinking, and atomic compositions. The rest of this Article is arranged as follows: Details of our computational methods are described in section 2. In section 3.1, we present the main results for algorithm and model validation using IP-PA bulk membranes, and in section 3.2, we present the main outcomes for ultrathin layer-by-layer membranes deposited on silicon substrate. Conclusions are given in section 4.

2. SIMULATION METHODOLOGY 2.1. Model. Chemical structures for meta-phenylenediamine (linear and end terminal) as well as for benzene-1,3,5tricarboxylic acid chloride (linear, cross-linked, and end terminal) monomers used in this work are shown in Figure 1. The transferable potential for phase equilibria (TraPPE)44−50 was used for nonbonded Lennard-Jones (LJ) interactions. Parameters for bonded interactions were taken from the

Figure 1. Chemical structures of (a) meta-phenylenediamine (MPD) and benzene-1,3,5-tricarboxylic acid chloride (TMC), (b) MPD repeat segment (MPD-L), TMC linear repeat segment (TMC-L), and TMC cross-linked repeat segment (TMC-C), and (c) polymer end segments (MPD-E and TMC-E). C

DOI: 10.1021/acs.jpcb.6b02801 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B

Figure 2. Visualization of first 1/2 cycle TMC monomers bonded to the silicon substrate (top view) for a TMC monomer growth rate of (a) 0.4 monomers/nm2, (b) 0.8 monomers/nm2, (c) 1.2 monomers/nm2, (d) 1.6 monomers/nm2, and (e) 2.0 monomers/nm2. Silicon, oxygen, hydrogen, and TMC atoms are colored yellow, red, white, and blue, respectively. Visualization of a top view of the first 1/2 cycle for four independent simulations for a TMC monomer growth rate of 0.4 monomers/nm2 is presented in Figure S3.

monomers were “removed from the simulation box” followed by an energy minimization, and a short 20 ps NVT MD simulation at 300 K. (4) Next, the current “silicon substrate simulation box” (that now includes the first layer of TMC bonded monomers) and the MPD-L (box 3) are combined followed by an energy minimization, and a short 5 ps NVT MD simulation. (5) MPD-L monomers are randomly bonded to the TMC-C monomers on the surface targeting a MPD/TMC ratio (R) of 1.3, 1.5, or 1.7 (with only one site of MPD-L monomer bonding to a TMC-C monomer). (6) At last, excess nonbonded MPD-L monomers are “removed from the simulation box” followed by an energy minimization, and a short 20 ps NVT MD simulation. This procedure completes the first cycle. The next cycle then begins with the combination of the “current simulation box” (silicon substrate, the first layer of TMC, and MPD monomers), and the TMC-C monomer simulation box 2. From the second cycle onward, during the first step TMC-C monomers are randomly bonded to the “free MPD-L surface” targeting coverages of 0.4, 0.8, 1.2, 1.6, or 2.0 monomers/nm2 (with only one site of TMC-C monomer bonding to MPD-L sites). As a reminder, this range was selected to study a similar range of molecular surface coverage observed in earlier experimental work on silicon substrates.20,50,63 This procedure is repeated until the desired number of cycles (N) is reached. Once N cycles are completed, virtual cross-linking between MPD-L and TMC-C monomers is carried out until no more bonds could be found within the preset 6.25 Å cutoff for a period of at least 500 ps. Finally, nonbonded MPD-L and TMC-C end polymer segments were converted to MPD-E, and TMC-E, followed by an energy minimization, and a 1 ns NVT MD simulation. All MD simulations were carried in NVT ensemble (T = 300 K, and Lx = 49.84 Å, Ly = 49.84 Å, and Lz = 300.00 Å). Using the virtual ultrathin layer-by-layer membrane generation procedure described above, 56 independent polymeric membranes were built. Each of the membranes consists of 20 cycles. Simulation cycles were conducted varying the number of TMC monomers bonded to the surface (S) and MPD/TMC

membranes.61 Finally, the hydrated membranes were energy minimized, and MD simulations were carried out in the isobaric−isothermal ensemble (T = 300 K and P = 1 atm) for a period of 1 ns to generate a final hydrated structure. In all systems, four independently generated samples were used to calculate the mean and standard deviation of specific properties. 2.3. Virtual Synthesis of Ultrathin Layer-by-Layer Membranes. For the investigation of ultrathin cross-linked PA films based on mLbL deposition on silicon substrates,12,13 the following model system was utilized. Initially, three separate simulation boxes were prepared. The first simulation box contains the silicon substrate, and the second and third boxes contain pure TMC-C and MPD-L (200 molecules each), respectively. The simulation box for the silicon substrate is elongated in the z-direction with Lx = 49.84 Å, Ly = 49.84 Å, and Lz = 300.00 Å. The origin of this box contains a slab of βcristobalite (≈23 Å thick in the z-direction) with its (1 0 0) surface (the x−y plane) exposed.62 Following the work of Zhuravlev et al.,50 the β-cristobalite surface was functionalized with silanol groups and periodic boundary conditions were applied in the x and y directions, mimicking an infinitely large surface. For both TMC-C and MPD-L simulation boxes, energy minimizations were performed followed by short 10 ps NPT MD simulations at 1 atm and 300 K while maintaining Lx = Ly = 49.84 Å (same xy dimensions as the silicon substrate box). To generate mLbL cross-linked PA thin film, first the Polymatic simulated polymerization algorithm was employed for each layer. Virtual synthesis of a single cycle membrane consists of six steps: (1) Initial configurations were prepared by combining the simulation boxes that contain the silicon substrate (box 1) and the TMC-C (box 2), followed by an energy minimization, and 5 ps NVT MD simulation at 300 K. (2) TMC-C monomers were randomly bonded to the surface silanol sites targeting coverages (S) of 0.4, 0.8, 1.2, 1.6, and 2.0 monomers/ nm2 (see Figure 2). The same bonding criterion as in section 2.2 was employed (with only one site of TMC-C monomer bonding to a silanol site). (3) Excess nonbonded TMC-C D

DOI: 10.1021/acs.jpcb.6b02801 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B ⎛ Lwet − Ldry ⎞ Swelling % = ⎜ ⎟ × 100 Lwet ⎝ ⎠

ratio (R). Both TMC monomers per surface area and MPD/ TMC ratios were kept constant between cycles for a given membrane. This strategy was employed specifically to study the effect of monomer ratios and surface coverage on the membrane cross-linking degree, composition and structural properties. Similar to the IP simulations, four independently generated samples were used to calculate the mean and standard deviation of specific properties (see Figure S3) for each system.

(1)

where Lwet and Ldry are the Lz simulation box dimensions of wet and dry polymer samples, respectively. We observed a swelling percentage close to 9% for all three models, as reported in Table 1. This agrees well with experimental data where swelling percentages between 4 and 12% were reported.65,66 The good agreement in both hydrated density and swelling percentages, in conjunction with previous work where dry densities of different polymers have been extensively validated [28, 46, 50− 54], provide a good indication that the dry densities obtained in this work should be on the right track. Next, we compared pore size distributions (PSD) with experimental data. PSD of the hydrated PA samples were calculated using Pore Blazer version 2.0.67 PSD represent the change of void volume while increasing a spherical probe diameter (Figure 3). In Figure 3, r is the probe diameter, V (r)

3. RESULTS AND DISCUSSION 3.1. Model Validation. In this work, we first validate the model, by evaluating three different force fields with the same structure generation methodology. We characterize the virtual samples and compare them against available experimental and simulation data. The rationale for exploring three different force field models (TraPPE, OPLS-AA and GAFF) was to determine the model that could accurately predict the properties of interest for IP membranes. As mentioned above, in this work IP-PA membranes were investigated in order to validate the molecular models used for mLbL membranes. For these IP-PA membranes, we first compared the hydrated density (see Table 1) to available experimental and molecular Table 1. Predicted Densities and Swelling of IP-PA Membranes Obtained in This Study for Several Force Fieldsa

a

force field

dry density (g/cm3)

hydrated density (g/cm3)

swelling %

TraPPE OPLS-AA GAFF

1.3014 1.3352 1.3582

1.2781 1.2981 1.3181

91 91 91

The subscripts indicate the standard deviation in the final digit.

simulation data. We choose this property since in previous simulation work, this has been the only thermodynamic property consistently compared in all molecular simulation studies.21−26,28−31 The dry density was calculated using the total polymer mass and the total volume of the simulation box, whereas the hydrated density was calculated using both polymer and water mass and the total volume of the simulation box. For the three force fields studied (TraPPE, OPLS-AA, and GAFF), the average hydrated densities were found to be 1.2781, 1.2981, and 1.3181 g/cm3 respectively (the subscripts indicate the standard deviation in the final digit), which is in good agreement with experimental hydrated density of 1.31 g/ cm3,24,61,64 and molecular simulation hydrated densities of 1.296−1.500 g/cm3.21−26,28−31 Although there are variations in hydrated densities with respect to the model used in this work, it is important to note that the maximum difference in density between models is less than 0.04 g/cm3, which is still within the range of values obtained in previous experimental studies.24,61,64 For the three force fields studied, the average dry densities were found to be 1.3014, 1.3352, and 1.3582 g/cm3 respectively. These results are consistent with previous simulation work, where dry densities on membranes were reported between 1.17 and 1.31 g/cm3.30,31 The variations observed in the models studied in this work, appears to be mainly due to differences in LJ parameters originated through the parameterization strategy employed during the force field development stage. We then evaluated swelling of the hydrated polymer sample (see Table 1), where swelling percentage is defined as65

Figure 3. Pore size distribution for simulated samples of hydrated IPPA membranes, using TraPPE (black line), GAFF (blue line), and OPLS-AA (red line) force fields.

corresponds to the monotonically decreasing cumulative pore volume function (normalized by total pore volume). In other words, this is the normalized volume of the void space that can be covered by spheres of diameter r or smaller. The pore size distribution was calculated by differentiation of the cumulative pore volume function V (r) as a function of probe diameter (−dV(r)/dr).68 The PSD profiles of the bulk IP-PA samples using the different force fields are shown in Figure 3. All three models result in very similar PSD profiles with a significant amount of the pores with the size of less than 6 Å. This is comparable with the experimental pore sizes between 4.4 and 5.0 Å that used positron annihilation lifetime spectroscopy (PALS) and Synchrotron SAXS.66,69 However, there are other PSD experimental studies that used PALS, where the data were interpreted with a bimodal distribution. The authors obtained a first distribution, between 2 and 5 Å, that was related to “network pores”, and a second distribution, between 7 and 8 Å, corresponding to “aggregate pores”.60,70 A bimodal distribution within these pore sizes was not observed during the current simulation study, nor other experimental or previous simulations.21,66,69 It is important to note that pore size distributions of IP-PA membranes can vary with respect to membrane properties such as the degree of cross-linking, monomer ratios, thickness, the amount of water, membrane E

DOI: 10.1021/acs.jpcb.6b02801 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B synthesis procedure, and post-treatment techniques. It is also important to note that although there are no significant differences between hydrated PSD profiles for the three different models, there are deviations in the PSD profiles for the dry samples obtained using different force fields (see Figure S5). This highlight that PSD for dry samples are more sensitive to the force field selected, as it has been observed previously.35 However, the selection of the force field in this work is based on the prediction of several thermodynamic properties, not only PSD, as well as computational efficiency. Most importantly the general tendencies and findings of this work do not change based on the force field selected. For example, PSD of the hydrated samples (Figure 3) consists of relatively larger pore sizes compared to dry PSDs (see Figure S5) which are expected due to swelling of the polymer framework. The average peak position for the hydrated samples is around 3.5−4.0 Å, with a distribution of pores up to 6.5 Å, and similar for all models, while dry samples have distinct peak positions for each force field, but all less than 2 Å. Overall, the consistency between the simulation results and experimental observations show that the simulation models employed in this study can provide a reasonable IP-PA membrane for further investigations. Going forward, we adopted the same molecular method for the virtual synthesis of PA-mLbL membranes on silicon substrates. However, for the mLbL membranes, only the TraPPE force field was used because on one hand, bulk membrane validation did not point toward a significant advantage of using one model over the other, and on the other hand the TraPPE representation uses a united atom mapping compared to all atoms OPLS-AA, and GAFF models. This difference between TraPPE and the other two models leads to approximately 31% decrease in the number of sites for the TraPPE model, which is less computationally expensive. 3.2. Ultrathin Layer-by-Layer Membranes Deposited on Silicon Substrate. To obtain a molecular level understanding of the relationship between membrane variables and physicochemical properties of the PA membrane, the TMC monomer growth rate and MPD/TMC ratio during mLbL deposition cycle were systematically varied. Growth rates and monomer ratios between cycles were kept constant for a given membrane. This allows us to investigate the relationship between monomer growth rate, monomer ratio, and membrane structural properties. Experimentally it has been suggested that the PA-mLbL film growth rates between deposition cycles is constant, hinting that molecular growth rates and monomer ratios between cycles might be similar for a given PA-mLbL membrane.11,13,20 3.2.1. Membrane Thickness. Results for PA-mLbL film thickness with respect to the number of deposition cycles are shown in Figure 4, and respective values of the slopes are given in Table 2. For film thickness calculations, the membrane/ vacuum interface was arbitrarily taken as the point where the density of the membrane reaches 0.9 g/cm3 (see Figure 5). The thickness of the membrane is calculated as the distance between the silicon substrate surface (≈ 23 Å) and the membrane/ vacuum interface. It can be seen that the PA film growth rate increases as a function of both TMC monomer growth rate and MPD/TMC ratio. The lowest film growth rate (2.148 Å/cycle) was observed for S = 0.4 and R = 1.3, and the highest film growth rate (7.621 Å/cycle) was observed for S = 2.0 and R = 1.7.

Figure 4. Thickness of the mLbL polymeric membrane with respect to the number of deposition cycles for MPD/TMC ratio = 1.7 (top), MPD/TMC ratio = 1.5 (middle), and MPD/TMC ratio = 1.3 (bottom). The polymer membrane/vacuum interface is arbitrarily taken as the point where the density of the membrane reaches 0.9 g/ cm3 in Figure 5. The colors indicate the TMC monomer growth rates per surface area: black = 0.4 monomers/nm2, red = 0.8 monomers/ nm2, blue = 1.2 monomers/nm2, magenta = 1.6 monomers/nm2, and orange = 2.0 monomers/nm2.

Table 2. Film Growth Rates for Varying MPD/TMC Monomer Ratios (R) and TMC Monomer Growth Rates per Surface Area (S = TMC Monomers/nm2)a TMC monomers/nm2 S S S S S

= = = = =

0.4 0.8 1.2 1.6 2.0

R = 1.3 (Å/cycle)

R = 1.5 (Å/cycle)

R = 1.7 (Å/cycle)

2.148 3.432 4.645 5.611 6.442

2.302 3.563 4.786 5.801 6.822

2.394 3.723 5.291 6.449 7.621

a

The subscripts indicate the standard deviation, for four independent samples, in the final digit.

Figure 5. Density profiles of 20 cycle mLbL polymer membranes calculated for MPD/TMC ratio = 1.7 (top), MPD/TMC ratio = 1.5 (middle), and MPD/TMC ratio = 1.3 (bottom). The colors indicate the TMC monomer growth rates per surface area: black = 0.4 monomers/nm2, red = 0.8 monomers/nm2, blue = 1.2 monomers/ nm2, magenta = 1.6 monomers/nm2, and orange = 2.0 monomers/ nm2. At the origin of this box (z-axis = 0) is the slab of β-cristobalite with the surface exposed to the PA membrane close to z-axis ≈ 23 Å. F

DOI: 10.1021/acs.jpcb.6b02801 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B Experimentally, it has been reported that film thickness calculated using X-ray reflectivity to exhibit relatively linear film growth rates with respect to the number of deposition cycles.11,13,15,20 Although average film growth was assumed to be linear in previous experimental work, a closer look reveals that significant variations with respect to the number of cycles can be present. For example, the film thickness for PA-mLbL membranes synthesized on silicon substrate using acetone or methanol as solvent exhibited 25, 100, and 200 Å thickness after 15, 30, and 60 cycles respectively (the authors report an average 3.3 Å/cycle film growth rate) and PA-mLbL membranes synthesized on silicon substrate using isopropanol as solvent exhibited 100, 150, and 330 Å thickness after 15, 30, and 60 cycle’s respectively (the authors report an average 5.5 Å/cycle film growth rate).13 Furthermore, PA-mLbL membranes synthesized on silicon substrate using toluene as solvent exhibited roughly 200 Å thickness after 40 cycle’s (average 5.0 Å/cycle film growth rate)15 whereas PA-mLbL membranes synthesized on a base layer of poly(vinyl alcohol) (substrate) using toluene as solvent exhibited 260 Å thickness after 30 cycle’s (average 8.6 Å/cycle film growth rate).11 This example illustrates the fact that film growth rate varies significantly depending on both the substrate and solvent used for the PAmLbL depositions process. A comparison of these experimental results with the simulation results presented in this work leads to the finding that if the membrane film growth rate is not constant with respect to the number of deposition cycles heterogeneous membranes are being formed. In other words, even small variations in film growth rates indicate significant variations in TMC monomer growth rates and/or MPD/TMC ratios between depositions cycles resulting in relatively heterogeneous membranes. Although it is not computationally feasible in this work to look into varying membrane variable (TMC growth rates and MPD/TMC ratios) between cycles as this would result in significantly larger number of simulation systems, the structural properties (density, free volume, atomic fractions, angle distributions and degree of cross-linking) obtained through molecular simulations for membranes with constant variables between cycles provide insight on how structural properties of experimental membranes vary with varying film growth regions. This is especially advantageous due to the fact that no structural properties other than film thickness and average atomic compositions for the entire membranes (not decomposed with respect to membrane depths) are reported in any of the previous experimental work on PA-mLbL membranes.11,13,15,16 and thus the results of this work provide new insights regarding mLbL membranes. The structural properties obtained from the simulation can thus be used to understand how key structural properties such as density, free volume, packing, or atomic fractions affect film growth rate and thus correlate with both water permeability and salt rejection. 3.2.2. Densities Profiles and Free Volume. Density profiles of the PA-mLbL membrane as a function of the z-coordinate (perpendicular to the vacuum/membrane interface) are shown in Figure 5. The density of PA-mLbL membrane increases as a function of the TMC monomer growth rate, and no significant variations with respect to different MPD/TMC ratios were observed. The lowest density ≈1 g/cm3 was observed for S = 0.4, and the highest density ≈1.5 g/cm3 was observed when S = 2.0. Furthermore, the free volume of the PA-mLbL membrane as a function of the z-coordinate is shown in Figure 6. The free volume was calculated using a bin size of 10 Å (in the z-axis)

Figure 6. Free volume of the mLbL polymeric membrane with respect to the z-axis for MPD/TMC ratio = 1.7 (top), MPD/TMC ratio = 1.5 (middle), and MPD/TMC ratio = 1.3 (bottom). The colors indicate the TMC monomer growth rates per surface area: black = 0.4 monomers/nm2, red = 0.8 monomers/nm2, blue = 1.2 monomers/ nm2, magenta = 1.6 monomers/nm2, and orange = 2.0 monomers/ nm2. The z-axis is the same as that represented in Figure 5.

and the void volume in specific bins were plotted against the zaxis center of the bin coordinate. From this figure, it is evident that the free volume in the PA-mLbL membrane closer to the silicon surface is higher for TMC monomer growth rates of 0.4 and 0.8. In contrast, no significant variations in free volume with respect to the depth of the membranes were observed for higher TMC monomer growth rates of 1.2, 1.6, and 2.0 monomers/nm2. These observations are consistent with density profiles (see Figure 5), where the density of the membranes was relatively low closer to the silicon surface for TMC monomer growth rates of 0.4 and 0.8 monomers/nm2. Moreover, free volume analyses clearly illustrate that the TMC monomer growth rates have a significant influence on the free volume of the mLbL-PA membranes which would in return influence both water permeability and salt rejection. These effects will be systematically studied in future work, where we will conduct MD simulations of hydrated mLbL-PA membranes. Comparing to the bulk PA membrane hydrated density of 1.3 g/cm3, the mLbL membrane density is relatively low for growth rate of 0.4 and 0.8 TMC monomers/nm2, high for growth rates of 1.6 and 2.0 TMC monomers/nm2, and only similar values were observed (between 1.25 and 1.35 g/cm3) for growth rates of 1.2 TMC monomers/nm2. Taking into account the fact that significantly different growth rates were observed in experimental studies, our findings imply that the experimental mLbL membranes with different growth rates exhibit significantly different densities for up to 20 cycles. 3.2.3. Angle Distributions and Hydrogen Bonding. For the mLbL PA membranes, we also monitored the average angle distribution between the aromatic plane of a TMC and a MPD monomer and the xy-plane (see Figure S2) as a function of number of cycles. The angle distribution was evaluated for different TMC growth rates, MPD/TMC ratios, and the number of deposition cycles. In Figure 7, we present the angle distribution of the aromatic rings and the xy-plane with varying TMC growth rates, and MPD/TMC ratios for 20 cycle membranes. Table S6 shows the average angle (θ) between the aromatic rings and the xy-plane with varying TMC growth rates, MPD/TMC ratios, and the number of deposition cycles. G

DOI: 10.1021/acs.jpcb.6b02801 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B

rejection and needs to be investigated in future work. Similar analysis for the angle formed between the CO vector and the vector normal to the surface, and the angle formed between the NH vector and the vector normal to the surface were conducted, but no variations with respect to TMC growth rates, MPD/TMC ratios, or number of deposition cycles were observed (see Figure S2 and Figure S4). Next, the silanol-polymer hydrogen-bonds are analyzed (see Table 3) as this is an important factor determining the Table 3. Average Number of Hydrogen Bonds per Silanol Groups in Systems for Varying MPD/TMC Monomer Ratios (R) and TMC Monomer Growth Rates per Surface Area (S = TMC Monomers/nm2)a Figure 7. Distribution of the angle formed between the vector normal to the aromatic plane and the vector normal to the surface for 20 cycle membranes (see Figure S2). The average angle (θ) between the aromatic rings and the xy-plane is also given inside the respective figures. Top row = MPD/TMC = 1.7, middle row = MPD/TMC = 1.5, bottom row = MPD/TMC = 1.3. The colors indicate the TMC monomer bonding per surface area: black = 0.4 monomers/nm2, red = 0.8 monomers/nm2, blue = 1.2 monomers/nm2, magenta = 1.6 monomers/nm2, and orange = 2.0 monomers/nm2. The subscripts indicate the standard deviation in the final digit.

TMC monomers/nm2 S S S S S

Nθ Bin × NTotal

0.4 0.8 1.2 1.6 2.0

R = 1.5

R = 1.7

0.081 0.203 0.314 0.435 0.453

0.092 0.172 0.294 0.395 0.477

0.081 0.193 0.275 0.385 0.456

a

The subscripts indicate the standard deviation, for four independent samples, in the final digit.

interactions between the polymer and the substrate.40,42 Hydrogen bonds were calculated by taking silanol “OH” group as the donor and the polymer “O” or “N” as the acceptor. Following previous work, the hydrogen bonds are defined according to a geometric criterion in which the distance between the donor group and the acceptor to be less than 0.29 nm with a limiting angle of 130°.41 Our results show that (1) as the TMC growth rate increases, the average number of hydrogen bonds per silanol group increases while (2) no significant variations with respect to different MPD/TMC ratio is observed. The former is expected since the number of “O” increases with TMC growth rate for the first 1/2 cycle. Changes in the MPD/TMC ratio in the following cycle have a lesser influence on the average number of hydrogen bonds per silanol groups, as the distance between the “N” acceptors and the silanol “OH” groups is larger. 3.2.4. Cross-Linking Degree and Atomic Fractions. We also calculated the cross-linking degree (cross-linked = TMC-C) with varying TMC growth rates, MPD/TMC ratios, and the number of deposition cycles (see Table 4) and atomic fractions with varying TMC growth rates and MPD/TMC ratios (see Table S5). The cross-linking degree increases as the MPD/ TMC ratio increases. No significant variation was observed for different TMC growth rates or the number of deposition cycles. It is worth noting that, since the TMC growth rate and MPD/ TMC ratio are kept constant between cycles, we do expect the degree of cross-linking between cycles to be similar. The lowest cross-link degree (61.0%) was observed for MPD/TMC ≈ 1.3, and the highest cross-link degree (93.7%) was observed for MPD/TMC ≈ 1.7. The variations in the degree of cross-linking in this mLbL membrane with respect to the MPD/TMC ratio is an important factor, as membrane swelling, due to hydration, would be influenced by the degree of cross-linking, which in return will impact both water permeability and salt rejections. Experimentally, PA-mLbL membranes are suggested to exhibit a considerably higher degree of cross-linking13 compared with bulk PA membranes,60 implying that the PA-mLbL membrane cross-linking degree is well over 50%. This is consistent with the findings presented in this work, which is the first to study

h(θ) is the probability of finding a molecule at a specific angle of (θ) as calculated by h(θ ) =

= = = = =

R = 1.3

(2)

where Bin is the bin size used to divide the range of angles (5°), Nθ is the number of configurations with an angle of θ and NTotal is the total number of configurations considered for angle distribution calculations. An angle of 0° indicates that the rings of the TMC and MPD molecule remain flat with respect to the surface, and 90° indicates the TMC, and MPD molecules lie perpendicular to the surface. From Figure 7, it can be observed that, on average, none of the species prefer to lie flat with respect to the surface; instead, they tend to orient at a tilted angle. Although it is expected free benzene will orient parallel to the substrate,43 in this study the polyamides are grafted to the substrate, and thus they orient with a tilted angle relative to the surface due to a compromise between steric hindrance and energetic effects. Our results show that, (1) as the TMC growth rate increases, the angular distribution becomes narrower and the average angle values increase, (2) no significant variations with respect to different MPD/TMC ratio is observed, and (3) as the number of deposition cycles increases, the average angle values increase. Considering the fact that an increase in the MPD/TMC ratio results in a higher number of MPD monomers bonding to TMC during each cycle, it might be expected that the average planar angles will vary due to interaction affecting packing, such as steric hindrance, bonding and electrostatics. In this study, however, it is observed that changes in the number of MPD monomers per cycle does not affect angle orientations. This is consistent with the density and free volumes profiles observed in Figures 5 and 6, where significant variations in both density and free volume are observed with varying TMC monomer growth rates but not with varying MPD/TMC ratios. These results show that packing of the monomers is significantly influenced by the TMC monomer growth rate, but not by the MPD/TMC ratio. This might also influence both water permeability, and salt H

DOI: 10.1021/acs.jpcb.6b02801 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B

MPD and TMC. Membrane models synthesized via IP were characterized via hydrated density, swelling percentage, and pore size distribution analyses, and good agreement was found against available experimental and molecular simulation data. Membrane models synthesized via molecular-layer-by-layer deposition were characterized via thickness, film growth rate, density, free volume, molecular orientation, the degree of crosslinking, and atomic composition. It was found that increasing the TMC monomer growth rate or MPD/TMC ratio increased the film growth rate and a constant film growth was observed while keeping the TMC monomer growth rate and MPD/ TMC ratio constant. The density of the mLbL membrane was sensitive to the TMC growth rate and increased with increasing TMC monomer growth rate. No significant variations in density were observed with respect to varying the MPD/TMC ratio. It was also found that, on average, TMC and MPD monomers prefer to orient at a tilted angle with respect to the substrate surface, where the average tilted angle increased with respect to increasing TMC growth rate and the number of deposition cycles with no significant variations with respect to variations of the MPD/TMC ratio. Finally, it was observed that the degree of cross-linking increased (atomic fractions of “C” to increase, “O” to decrease, and “N” to increase) as the MPD/ TMC ratio was increased and no significant variations with varying TMC growth rate or number of deposition cycles were observed. This study clearly illustrates that changes in experimental conditions have a significant impact on physicochemical properties of the mLbL membrane and potentially have an important effect on the membrane performance. In future work, we will conduct a systematic study of hydrated membranes to investigate the relationship between membrane performance and synthesis conditions that can be used in designing and optimizing the next generation mLbL for relevant industrial applications.

Table 4. Cross-Linked TMC Monomer Percentage Calculated from Four Independently Generated mLbL PA Membranes for Different MPD/TMC Monomer Ratios (R), TMC Growth Rates per Surface Area (S = TMC/nm2), and Number of Cycles (C)a

a

R

S

1.7 1.5 1.3 1.7 1.5 1.3 1.7 1.5 1.3 1.7 1.5 1.3 1.7 1.5 1.3

0.4 0.4 0.4 0.8 0.8 0.8 1.2 1.2 1.2 1.6 1.6 1.6 1.7 1.5 1.3

C = 10 cross-link % C = 15 cross-link % C = 20 cross-link % 891 794 612 922 832 652 932 843 662 931 842 661 941 841 661

901 792 622 941 802 622 941 821 641 941 831 652 941 831 661

912 801 643 921 812 641 931 811 652 941 812 631 931 831 641

The subscripts indicate the standard deviation in the final digit.

mLbL membranes by molecular simulations. In previous IP-PA simulation work, cross-linking degrees between 20 and 85% were studied.21−31 However, those results cannot be compared with the ones presented here, since the IP-PA and mLbL membranes are fundamentally different. Similarly to crosslinking, no significant variations in atomic fractions were observed with varying TMC growth rate or the number of deposition cycles. In contrast, the atomic fractions of “C” and “N” increase while “O” decreases with increasing MPD/TMC ratio (see Table S5). We also analyzed the atomic fractions of the mLbL membranes as a function of distance from the silicon surface (see Figure S6). No significant variations in atomic fractions with respect to the depth of the mLbL membranes were observed, suggesting that the PA-mLbL membranes are homogeneous if the TMC growth rates and MPD/TMC ratios are kept constant between deposition cycles. These results indicate that the degree of cross-linking and compositions (atomic fractions of “C” to increase, “O” to decrease, and “N” to increase) are strongly dependent on the MPD/TMC ratio but relatively independent of the TMC growth rate or the number of deposition cycles. In other words, this indicates that the membrane thickness and chemical composition are independent of one another (if the TMC growth rate and MPD/TMC ratio are kept constant between cycles). This is a rather different conclusion than the one obtained for IP-PA membranes where the thickness and local compositions are closely coupled. Furthermore, the PA-mLbL membrane thermo/structural properties (thickness, density, free volume, and orientation) are strongly coupled to the TMC growth rate, while chemical properties (degree of cross-linking and compositions) are strongly coupled to the MPD/TMC ratio. This is an important advantage for membrane optimization processes when comparing to IP-PA membranes.

■

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jpcb.6b02801. Schematic representations, force field parameters, angle distributions, simulation snapshots, pore size distributions, density profiles, membrane thickness, and atom compositions (PDF)

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]fl.edu. Phone: 352-294-3488. Present Address

§ T.P.L., C.M.C.: Department of Chemistry, University of Florida, Gainesville, FL 32611.

Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS We thank Kyle E. Hart, Abhishek Roy, Steve Rosenberg, Ian Tomlinson, Bob Cieslinski (Dow Chemical Company); Michael Fortunato, Michael Hickner, Tawanda Zimudzi (Penn State); Christopher M. Stafford and Edwin P. Chan (NIST) for helpful discussions and anonymous reviewers for their valuable suggestions. In addition, we thank the Dow Chemical Company for funding. High-performance computational resources for this research were provided by the Research

4. CONCLUSIONS In this study, the physicochemical properties of polyamide membranes composed of MPD and TMC were investigated. We present the first molecular simulation study of PA polymer membranes synthesized via mLbL deposition, consisting of I

DOI: 10.1021/acs.jpcb.6b02801 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B

(21) Ding, M.; Ghoufi, A.; Szymczyk, A. Molecular Simulations of Polyamide Reverse Osmosis Membranes. Desalination 2014, 343, 48− 53. (22) Harder, E.; Walters, D. E.; Bodnar, Y. D.; Faibish, R. S.; Roux, B. t. Molecular Dynamics Study of a Polymeric Reverse Osmosis Membrane. J. Phys. Chem. B 2009, 113, 10177−10182. (23) Hughes, Z. E.; Gale, J. D. A Computational Investigation of the Properties of a Reverse Osmosis Membrane. J. Mater. Chem. 2010, 20, 7788−7799. (24) Kotelyanskii, M.; Wagner, N.; Paulaitis, M. Atomistic Simulation of Water and Salt Transport in the Reverse Osmosis Membrane Ft-30. J. Membr. Sci. 1998, 139, 1−16. (25) Luo, Y.; Harder, E.; Faibish, R. S.; Roux, B. Computer Simulations of Water Flux and Salt Permeability of the Reverse Osmosis Ft-30 Aromatic Polyamide Membrane. J. Membr. Sci. 2011, 384, 1−9. (26) Xiang, Y.; Liu, Y.; Mi, B.; Leng, Y. Hydrated Polyamide Membrane and Its Interaction with Alginate: A Molecular Dynamics Study. Langmuir 2013, 29, 11600−11608. (27) Ding, M.; Szymczyk, A.; Ghoufi, A. On the Structure and Rejection of Ions by a Polyamide Membrane in Pressure-Driven Molecular Dynamics Simulations. Desalination 2015, 368, 76−80. (28) Ding, M.; Szymczyk, A.; Goujon, F.; Soldera, A.; Ghoufi, A. Structure and Dynamics of Water Confined in a Polyamide ReverseOsmosis Membrane: A Molecular-Simulation Study. J. Membr. Sci. 2014, 458, 236−244. (29) Gao, W.; She, F.; Zhang, J.; Dumee, L. F.; He, L.; Hodgson, P. D.; Kong, L. Understanding Water and Ion Transport Behaviour and Permeability through Poly(Amide) Thin Film Composite Membrane. J. Membr. Sci. 2015, 487, 32−39. (30) Kolev, V.; Freger, V. Hydration, Porosity and Water Dynamics in the Polyamide Layer of Reverse Osmosis Membranes: A Molecular Dynamics Study. Polymer 2014, 55, 1420−1426. (31) Shen, M.; Keten, S.; Lueptow, R. M. Dynamics of Water and Solute Transport in Polymeric Reverse Osmosis Membranes Via Molecular Dynamics Simulations. J. Membr. Sci. 2016, 506, 95−108. (32) Abbott, L.; Colina, C. Polymatic: A simulated polymerization algorithm, v. 1.0, 2015. Available at https://nanohub.org/resources/ 17278. (33) Abbott, L. J.; Hart, K. E.; Colina, C. M. Polymatic: A Generalized Simulated Polymerization Algorithm for Amorphous Polymers. Theor. Chem. Acc. 2013, 132, 1−19. (34) Larsen, G. S.; Lin, P.; Hart, K. E.; Colina, C. M. Molecular Simulations of Pim-1-Like Polymers of Intrinsic Microporosity. Macromolecules 2011, 44, 6944−6951. (35) Hart, K. E.; Abbott, L. J.; Colina, C. M. Analysis of Force Fields and Bet Theory for Polymers of Intrinsic Microporosity. Mol. Simul. 2013, 39, 397−404. (36) Hart, K. E.; Abbott, L. J.; McKeown, N. B.; Colina, C. M. Toward Effective Co2/Ch4 Separations by Sulfur-Containing Pims Via Predictive Molecular Simulations. Macromolecules 2013, 46, 5371− 5380. (37) Hart, K. E.; Colina, C. M. Estimating Gas Permeability and Permselectivity of Microporous Polymers. J. Membr. Sci. 2014, 468, 259−268. (38) Abbott, L. J.; Hughes, J. E.; Colina, C. M. Virtual Synthesis of Thermally Cross-Linked Copolymers from a Novel Implementation of Polymatic. J. Phys. Chem. B 2014, 118, 1916−1924. (39) Abbott, L. J.; Colina, C. M. Porosity and Ring Formation in Conjugated Microporous Polymers. J. Chem. Eng. Data 2014, 59, 3177−3182. (40) Eslami, H.; Mueller-Plathe, F. Structure and Mobility of Nanoconfined Polyamide-6,6 Oligomers: Application of a Molecular Dynamics Technique with Constant Temperature, Surface Area, and Parallel Pressure. J. Phys. Chem. B 2009, 113, 5568−5581. (41) Eslami, H.; Muller-Plathe, F. Molecular Dynamics Simulation of Water Influence on Local Structure of Nanoconfined Polyamide-6,6. J. Phys. Chem. B 2011, 115, 9720−9731.

Computing and Cyberinfrastructure unit of Information Technology Services at Pennsylvania State University.

■

REFERENCES

(1) Wang, J.; Dlamini, D. S.; Mishra, A. K.; Pendergast, M. T. M.; Wong, M. C. Y.; Mamba, B. B.; Freger, V.; Verliefde, A. R. D.; Hoek, E. M. V. A Critical Review of Transport through Osmotic Membranes. J. Membr. Sci. 2014, 454, 516−537. (2) Lee, K. P.; Arnot, T. C.; Mattia, D. A Review of Reverse Osmosis Membrane Materials for DesalinationDevelopment to Date and Future Potential. J. Membr. Sci. 2011, 370, 1−22. (3) Geise, G. M.; Lee, H.-S.; Miller, D. J.; Freeman, B. D.; McGrath, J. E.; Paul, D. R. Water Purification by Membranes: The Role of Polymer Science. J. Polym. Sci., Part B: Polym. Phys. 2010, 48, 1685− 1718. (4) Petersen, R. J. Composite Reverse Osmosis and Nanofiltration Membranes. J. Membr. Sci. 1993, 83, 81−150. (5) Shannon, M. A.; Bohn, P. W.; Elimelech, M.; Georgiadis, J. G.; Marinas, B. J.; Mayes, A. M. Science and Technology for Water Purification in the Coming Decades. Nature 2008, 452, 301−310. (6) Elimelech, M.; Phillip, W. A. The Future of Seawater Desalination: Energy, Technology, and the Environment. Science 2011, 333, 712−717. (7) Cadotte, J. E.; Petersen, R. J.; Larson, R. E.; Erickson, E. E. A New Thin-Film Composite Seawater Reverse Osmosis Membrane. Desalination 1980, 32, 25−31. (8) Ghosh, A. K.; Jeong, B.-H.; Huang, X.; Hoek, E. M. Impacts of Reaction and Curing Conditions on Polyamide Composite Reverse Osmosis Membrane Properties. J. Membr. Sci. 2008, 311, 34−45. (9) Freger, V. Nanoscale Heterogeneity of Polyamide Membranes Formed by Interfacial Polymerization. Langmuir 2003, 19, 4791−4797. (10) Coronell, O.; Mariñas, B. J.; Cahill, D. G. Depth Heterogeneity of Fully Aromatic Polyamide Active Layers in Reverse Osmosis and Nanofiltration Membranes. Environ. Sci. Technol. 2011, 45, 4513− 4520. (11) Johnson, P. M.; Yoon, J.; Kelly, J. Y.; Howarter, J. A.; Stafford, C. M. Molecular Layer-by-Layer Deposition of Highly Crosslinked Polyamide Films. J. Polym. Sci., Part B: Polym. Phys. 2012, 50, 168− 173. (12) Chan, E. P. Deswelling of Ultrathin Molecular Layer-by-Layer Polyamide Water Desalination Membranes. Soft Matter 2014, 10, 2949−2954. (13) Chan, E. P.; Lee, J.-H.; Chung, J. Y.; Stafford, C. M. An Automated Spin-Assisted Approach for Molecular Layer-by-Layer Assembly of Crosslinked Polymer Thin Films. Rev. Sci. Instrum. 2012, 83, 114102. (14) Choi, W.; Gu, J.-E.; Park, S.-H.; Kim, S.; Bang, J.; Baek, K.-Y.; Park, B.; Lee, J. S.; Chan, E. P.; Lee, J.-H. Tailor-Made Polyamide Membranes for Water Desalination. ACS Nano 2015, 9, 345−355. (15) Chan, E. P.; Young, A. P.; Lee, J. H.; Stafford, C. M. Swelling of Ultrathin Molecular Layer-by-Layer Polyamide Water Desalination Membranes. J. Polym. Sci., Part B: Polym. Phys. 2013, 51, 1647−1655. (16) Gu, J. E.; Lee, S.; Stafford, C. M.; Lee, J. S.; Choi, W.; Kim, B. Y.; Baek, K. Y.; Chan, E. P.; Chung, J. Y.; Bang, J. Molecular Layer-byLayer Assembled Thin-Film Composite Membranes for Water Desalination. Adv. Mater. 2013, 25, 4778−4782. (17) Kohli, P.; Blanchard, G. J. Applying Polymer Chemistry to Interfaces: Layer-by-Layer and Spontaneous Growth of Covalently Bound Multilayers. Langmuir 2000, 16, 4655−4661. (18) Kohli, P.; Blanchard, G. J. Design and Demonstration of Hybrid Multilayer Structures: Layer-by-Layer Mixed Covalent and Ionic Interlayer Linking Chemistry. Langmuir 2000, 16, 8518−8524. (19) Major, J. S.; Blanchard, G. J. Covalently Bound Polymer Multilayers for Efficient Metal Ion Sorption. Langmuir 2001, 17, 1163−1168. (20) Lomadze, N.; Perez, M.; Prucker, O.; Ruhe, J.; Reinecke, H. Step-and-Repeat Assembly of Molecularly Controlled Ultrathin Polyaramide Layers. Macromolecules 2010, 43, 9056−9062. J

DOI: 10.1021/acs.jpcb.6b02801 J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B

W.; Wong, M. W.; Gonzalez, C.; Pople, J. A. Gaussian 03 2004, revision C.02. (58) Plimpton, S. Fast Parallel Algorithms for Short-Range Molecular Dynamics. J. Comput. Phys. 1995, 117, 1−19. (59) Jorgensen, W. L.; Maxwell, D. S.; Tirado-Rives, J. Development and Testing of the Opls All-Atom Force Field on Conformational Energetics and Properties of Organic Liquids. J. Am. Chem. Soc. 1996, 118, 11225−11236. (60) Kim, S. H.; Kwak, S.-Y.; Suzuki, T. Positron Annihilation Spectroscopic Evidence to Demonstrate the Flux-Enhancement Mechanism in Morphology-Controlled Thin-Film-Composite (Tfc) Membrane. Environ. Sci. Technol. 2005, 39, 1764−1770. (61) Mi, B.; Cahill, D. G.; Marinas, B. J. Physico-Chemical Integrity of Nanofiltration/Reverse Osmosis Membranes During Characterization by Rutherford Backscattering Spectrometry. J. Membr. Sci. 2007, 291, 77−85. (62) Wyckoff, R. W. Ix. Die Kristallstruktur Von B-Cristobalit Sio2 (Bei Hohen Temperaturen Stabile Form). Z. Kristallogr. - Cryst. Mater. 1925, 62, 189−200. (63) Peri, J.; Hensley, A., Jr The Surface Structure of Silica Gel. J. Phys. Chem. 1968, 72, 2926−2933. (64) Zhang, X. J.; Cahill, D. G.; Coronell, O.; Marinas, B. J. Absorption of Water in the Active Layer of Reverse Osmosis Membranes. J. Membr. Sci. 2009, 331, 143−151. (65) Freger, V. Swelling and Morphology of the Skin Layer of Polyamide Composite Membranes: An Atomic Force Microscopy Study. Environ. Sci. Technol. 2004, 38, 3168−3175. (66) Lee, J.; Doherty, C. M.; Hill, A. J.; Kentish, S. E. Water Vapor Sorption and Free Volume in the Aromatic Polyamide Layer of Reverse Osmosis Membranes. J. Membr. Sci. 2013, 425−426, 217− 226. (67) Sarkisov, L.; Harrison, A. Computational Structure Characterisation Tools in Application to Ordered and Disordered Porous Materials. Mol. Simul. 2011, 37, 1248−1257. (68) Gelb, L. D.; Gubbins, K. Pore Size Distributions in Porous Glasses: A Computer Simulation Study. Langmuir 1999, 15, 305−308. (69) Singh, P.; Ray, P.; Xie, Z.; Hoang, M. Synchrotron Saxs to Probe Cross-Linked Network of Polyamide ‘Reverse Osmosis’ and ‘Nanofiltration’membranes. J. Membr. Sci. 2012, 421−422, 51−59. (70) Sharma, V.; Singh, P.; Gautam, S.; Maheshwari, P.; Dutta, D.; Mukhopadhyay, R. Dynamics of Water Sorbed in Reverse Osmosis Polyamide Membrane. J. Membr. Sci. 2009, 326, 667−671.

(42) Eslami, H.; Muller-Plathe, F. How Thick Is the Interphase in an Ultrathin Polymer Film? Coarse-Grained Molecular Dynamics Simulations of Polyamide-6,6 on Graphene. J. Phys. Chem. C 2013, 117, 5249−5257. (43) Eslami, H.; Behrouz, M. Molecular Dynamics Simulation of a Polyamide-66/Carbon Nanotube Nanocomposite. J. Phys. Chem. C 2014, 118, 9841−9851. (44) Chen, B.; Potoff, J. J.; Siepmann, J. I. Monte Carlo Calculations for Alcohols and Their Mixtures with Alkanes. Transferable Potentials for Phase Equilibria. 5. United-Atom Description of Primary, Secondary, and Tertiary Alcohols. J. Phys. Chem. B 2001, 105, 3093−3104. (45) 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. (46) Wick, C. D.; Stubbs, J. M.; Rai, N.; Siepmann, J. I. Transferable Potentials for Phase Equilibria. 7. Primary, Secondary, and Tertiary Amines, Nitroalkanes and Nitrobenzene, Nitriles, Amides, Pyridine, and Pyrimidine. J. Phys. Chem. B 2005, 109, 18974−18982. (47) Stubbs, J. M.; Potoff, J. J.; Siepmann, J. I. Transferable Potentials for Phase Equilibria. 6. United-Atom Description for Ethers, Glycols, Ketones, and Aldehydes. J. Phys. Chem. B 2004, 108, 17596−17605. (48) Rafferty, J. L.; Siepmann, J. I.; Schure, M. R. Retention Mechanism for Polycyclic Aromatic Hydrocarbons in Reversed-Phase Liquid Chromatography with Monomeric Stationary Phases. Journal of Chromatography A 2011, 1218, 9183−9193. (49) Bai, P.; Tsapatsis, M.; Siepmann, J. I. Trappe-Zeo: Transferable Potentials for Phase Equilibria Force Field for All-Silica Zeolites. J. Phys. Chem. C 2013, 117, 24375−24387. (50) Zhuravlev, N. D.; Siepmann, J. I.; Schure, M. R. Surface Coverages of Bonded-Phase Ligands on Silica: A Computational Study. Anal. Chem. 2001, 73, 4006−4011. (51) Wang, J.; Wolf, R. M.; Caldwell, J. W.; Kollman, P. A.; Case, D. A. Development and Testing of a General Amber Force Field. J. Comput. Chem. 2004, 25, 1157−1174. (52) Abascal, J. L. F.; Vega, C. A General Purpose Model for the Condensed Phases of Water: Tip4p/2005. J. Chem. Phys. 2005, 123, 234505. (53) Allen, M. P.; Tildesley, D. J. Computer Simulation of Liquids; Oxford University Press: New York, 1989. (54) Hockney, R. W.; Eastwood, J. W. Computer Simulation Using Particles; Taylor & Francis Group, LLC: New York, 1988. (55) Bayly, C. I.; Cieplak, P.; Cornell, W.; Kollman, P. A. A WellBehaved Electrostatic Potential Based Method Using Charge Restraints for Deriving Atomic Charges: The Resp Model. J. Phys. Chem. 1993, 97, 10269−10280. (56) Cornell, W. D.; Cieplak, P.; Bayly, C. I.; Gould, I. R.; Merz, K. M.; Ferguson, D. M.; Spellmeyer, D. C.; Fox, T.; Caldwell, J. W.; Kollman, P. A. A Second Generation Force Field for the Simulation of Proteins, Nucleic Acids, and Organic Molecules. J. Am. Chem. Soc. 1995, 117, 5179−5197. (57) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. A.; Montgomery, J. A; Vreven, T.; Kudin, K. N.; Burant, J. C.; Millam, J. M.; Iyengar, S. S.; Tomasi, J.; Barone, V.; Mennucci, B.; Cossi, M.; Scalmani, G.; Rega, N.; Petersson, G. A.; Nakatsuji, H.; Hada, M.; Ehara, M.; Toyota, K.; Fukuda, R.; Hasegawa, J.; Ishida, M.; Nakajima, T.; Honda, Y.; Kitao, O.; Nakai, H.; Klene, M.; Li, X.; Knox, J. E.; Hratchian, H. P.; Cross, J. B.; Bakken, V.; Adamo, C.; Jaramillo, J.; Gomperts, R.; Stratmann, R. E.; Yazyev, O.; Austin, A. J.; Cammi, R.; Pomelli, C.; Ochterski, J. W.; Ayala, P. Y.; Morokuma, K.; Voth, G. A.; Salvador, P.; Dannenberg, J. J.; Zakrzewski, V. G.; Dapprich, S.; Daniels, A. D.; Strain, M. C.; Farkas, O.; Malick, D. K.; Rabuck, A. D.; Raghavachari, K.; Foresman, J. B.; Ortiz, J. V.; Cui, Q.; Baboul, A. G.; Clifford, S.; Cioslowski, J.; Stefanov, B. B.; Liu, G.; Liashenko, A.; Piskorz, P.; Komaromi, I.; Martin, R. L.; Fox, D. J.; Keith, T.; Al-Laham, M. A.; Peng, C. Y.; Nanayakkara, A.; Challacombe, M.; Gill, P. M. W.; Johnson, B.; Chen, K

DOI: 10.1021/acs.jpcb.6b02801 J. Phys. Chem. B XXXX, XXX, XXX−XXX