Scholarly article on topic 'PIM-1/graphene composite: A combined experimental and molecular simulation study'

PIM-1/graphene composite: A combined experimental and molecular simulation study Academic research paper on "Chemical engineering"

Share paper
Academic journal
Microporous and Mesoporous Materials
OECD Field of science
{"Polymers of intrinsic microporosity" / Graphene / Composite / "CO2 adsorption" / "Membrane separation"}

Abstract of research paper on Chemical engineering, author of scientific article — Aleksandra Gonciaruk, Khalid Althumayri, Wayne J. Harrison, Peter M. Budd, Flor R. Siperstein

Abstract This work presents a combined molecular simulation and experimental study to understand the effect of graphene on the packing and gas adsorption performance of a new class of polymers, known as polymers of intrinsic microporosity (PIMs). PIMs can be processed to membranes or other useful forms and their chemistry can be tailored for specific applications. Their rigid and contorted macromolecular structures give rise to a large amount of microvoids attractive for small molecule adsorption. We show that the presence of graphene in the composite affects the structure of the membrane as evidenced by the change in colour and SEM micrographs, but it does not reduce significantly the adsorption capacity of the material.

Academic research paper on topic "PIM-1/graphene composite: A combined experimental and molecular simulation study"

Microporous and Mesoporous Materials xxx (2014) xxx-xxx

Contents lists available at ScienceDirect

Microporous and Mesoporous Materials

journal homepage:

PIM-1/graphene composite: A combined experimental and molecular simulation study

Aleksandra Gonciaruka, Khalid Althumayrib, Wayne J. Harrison b, Peter M. Buddb, Flor R. Sipersteina'*

a The School of Chemical Engineering and Analytical Science, The University of Manchester, M13 9PL, United Kingdom b The School of Chemistry, The University of Manchester, M13 9PL, United Kingdom


This work presents a combined molecular simulation and experimental study to understand the effect of graphene on the packing and gas adsorption performance of a new class of polymers, known as polymers of intrinsic microporosity (PIMs). PIMs can be processed to membranes or other useful forms and their chemistry can be tailored for specific applications. Their rigid and contorted macromolecular structures give rise to a large amount of microvoids attractive for small molecule adsorption. We show that the presence of graphene in the composite affects the structure of the membrane as evidenced by the change in colour and SEM micrographs, but it does not reduce significantly the adsorption capacity of the material. © 2014 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license (http://

Article history: Received 13 May 2014 Accepted 3 July 2014 Available online xxxx


Polymers of intrinsic microporosity



CO2 adsorption

Membrane separation

1. Introduction

At the core of most gas separation processes we can find membranes, adsorbents or absorbents. Membrane separations are one of the most energy-efficient tools that have minimum environmental impact compared to conventional technologies such as cryogenic distillation, or even gas absorption. However, membrane technology is not yet employed at its full capacity and new materials are needed to exploit its potential. For example, CO2 removal from natural gas streams by membrane technology represents less than 5% of the market [1]. Currently, absorption with amine-based liquids is the most widely used and well-established CO2 separation technology. However, the method demands high energy consumption due to the low solubility of CO2 which then requires heating and cooling large volumes of liquid, high rate of solvent consumption, solvent recovery, corrosion problems and complex control of the process. Membrane-based technologies are a competitive alternative since membranes do not require regeneration, no phase change occurs, the process is single stage and no moving parts are involved. Nevertheless currently used membranes suffer from low permeability, so that large areas are required to permeate the gas, or low selectivity requiring multistage processes to reach

* Corresponding author. E-mail addresses: (A. Gonciaruk), (K. Althumayri), waye. (W.J. Harrison), (P.M. Budd), (F.R. Siperstein).

the desired purity. By designing robust and efficient materials that could treat high gas volumes, smaller membrane areas could be used reducing the total cost of equipment, or smaller pressure gradients reducing the operating costs.

For membrane gas separations, polymers offer a bouquet of advantages; their chemistry can be tailored to achieve great selectivity, membranes can be flexible, light and very thin due to ease of processability. However the most selective polymers such as polyi-mides, polysulfones and polycarbonates do not have sufficient free volume and so permeability is too low. Over the past few decades high-free volume polymers, such as PTMSP [2], PMP [3] and PIMs [4-7], were developed based on chain rigidity and interchain separation. However such polymers exhibit a decrease in permeability over time due to structure collapse. Cross-linking the chains and addition of nanoparticles into the polymer framework were shown to be useful for preventing physical ageing [8,9]. Nevertheless the trade-off between permeability and selectivity, as well physical ageing of high free volume polymers, drives a search for new building blocks and synthesis routes that could improve and control polymer architecture.

The new class of materials that are currently of great interest, polymers of intrinsic microporosity (PIMs), could also be promising sorbents for efficient gas separation. PIMs belong to a family of amorphous glassy polymers that form microporous solids simply because they possess highly rigid and contorted structures which prevent them from filling the space efficiently. There are a variety of molecules [10] that could be used to synthesize the PIMs due to their contorted structure imposed by either a spiro-centre or by 1387-1811/© 2014 The Authors. Published by Elsevier Inc.

This is an open access article under the CC BY license (

A. Gonciaruk et al./Microporous and Mesoporous Materials xxx (2014) xxx-xxx

steric crowding around the covalent bond. Thus necessary properties of PIMs can be potentially tailored by introducing a suitable co-monomer and/or functional group into the polymer chain, enhancing performance of the sorbent such as selectivity, capacity, solubility and stability.

The combination of microporosity and ability to generate solu-tion-processable film-forming material offers unique benefits in membrane technology. PIM membranes were investigated in per-vaporative separation of phenols from aqueous solutions, showing selectivity and total fluxes comparable to those obtained with conventionally used rubbery polymers [11]. Later applications for gas separation membranes were successful as well, showing that PIM-

1 possesses high permeability exceeded only by highly permeable polymers such as PTMSP and Teflon AF2400, whilst its selectivity was significantly higher [7]. PIMs crossed Robeson's 1991 upper bound [12] for several important gas-pairs such as O2/N2 and CO2/CH4, which contributed to its revision in 2008 [13]. The bound shows the trend of selectivity against permeability towards common gas pairs for many membrane materials. Polymeric materials as gas separation membranes suffer from a well-known trade-off between permeability and selectivity; those with high selectivity possess poor permeability properties and vice versa.

Incorporation of an additional component to already existing polymer offers a fast and cost-effective alternative for production of new high-performance materials. Mixed matrix membranes (MMM) are currently recognised as a competitive approach [14]. It was shown that mixing polymer with inorganic adsorbents such as fumed silica, carbon black or carbon nanotubes can increase permeability of pristine PIM-1, whilst the selectivity can be affected positively or negatively depending on the filler, other membrane properties and gas pairs [14-17].

The aim of this work is to investigate PIM packing behaviour in the presence of graphene slabs, and understand how it affects PIM structural properties and subsequently adsorption of carbon dioxide (CO2). Although it is expected that the permeability of the composite material will be different to that for PIM-1, it is beyond the scope of this work to assess the permeability of the composite materials. We are mainly interested in understanding the effect on selectivity, because based on the information available in the literature, variations in selectivity can be hard to predict by intuition. It is expected that graphene incorporated in PIM-1 can potentially disrupt packing of polymer chains resulting in increased free volume or create additional voids at the interface between itself and polymer. For this purpose, we employed a combination of virtual molecular models of the composite followed by gas adsorption simulations, as well as experimental characterisation of the composites. The model of the composite system provides an atomistic level insight into molecular structure, whilst the experimental study on adsorption will complement the model and serve as a validation basis.

2. Experimental method

Adsorption of CO2 in the sample was studied by the static gravimetric technique using Hiden Isochema's (UK) intelligent gravimetric analyser (IGA-001). The sample was pre-treated in order to release any contaminants. Pretreatment involved thermal depressurising to a vacuum using a turbomolecular pump. The sample was maintained at 373 K for 3 h and 8 h at 473 K using an electric furnace. The temperature was increased at a rate of

2 K/min. Helium adsorption was measured prior to collecting the CO2 isotherm in order to account for the buoyancy effect and to calculate sample density. CO2 isotherms were measured gradually increasing the pressure from 0 to 20 bar at constant temperature, either 293 K or 333 K. The system was allowed to equilibrate at each pressure point for a maximum of 3 h.

2.1. Dual-mode model

Collected adsorption data was fitted to the dual-mode model to allow for data processing and comparison between different materials. To represent adsorption of gases in glassy polymers the dualmode (DM) isotherm model is conventionally used. The model has a fundamental basis, which postulates that one population of gas is dissolved in a fraction of the solid according to the Henry's law just like a sorption in rubbery polymers and another population behaves as in the Langmuir model, i.e., a number of independent and constant energy sites are available where molecules can be adsorbed; at low pressures the adsorption is proportional to the gas pressure whereas at high pressures a saturation capacity is reached. The classical expression takes the following form:

n = ni + n2 = kiP +

mk2 P TTkP

where n1 and n2 is the amount adsorbed based on Henry's law and Langmuir model, respectively, k1 is Henry's law dissolution constant, P is pressure, m is the Langmuir saturation capacity constant and k2 is the Langmuir hole affinity constant. k2 increases with increase in gas-solid interaction energy and decrease with temperature. Fig. 1 shows a graphic representation of the DM model including the contribution of Langmuir and Henry's law models.

The data obtained are valuable in understanding how graphene affects the adsorption performance of P1M-1. The low pressure region will provide insight into the composite's affinity for CO2, whilst the high pressure region will indicate how graphene affects free volume and mobility of P1M-1 chains, i.e., swelling.

2.2. Enthalpies of adsorption

Combining isotherms at different temperatures it is possible to calculate enthalpy of adsorption, AH, and, subsequently, predict loadings at other temperatures of interest. The equation is as follows:


d ln P

where P is the pressure and T is the temperature of a system with n moles adsorbed. Pressure values at a given CO2 loading were calculated once the experimental isotherms were fitted to the DM equation.

2.3. Samples

2.3.1. Gas

Carbon dioxide (CO2) and Helium (He) of purity 99,995% (4.5 grade) and 99.999% (5.0 grade), respectively, were used as received from the manufacturer, BOC Gases. Helium was used for buoyancy correction as discussed.

2.3.2. Composite

The material tested contained a 1 wt% loading of graphene. PIM-1-graphene composite forms a green membrane (Fig. 2) which is

evidently different from the bright yellow PIM-1 membranes. Gravimetrically measured density for the composite was 0.999 ± 0.021 g/cm3.

Fig. 3 shows SEM images of the cross section of a pure PIM-1 membrane suggesting it is a fairly uniform material. Fig. 4 shows SEM images of the cross section of a PIM-1/graphene membrane which is clearly different from the pure PIM-1 membrane. The membrane cross section appears to split into two sections, with the lower section showing a more raised and jagged structure. The membrane structure could have been compromised during the preparation, although different preparation methods (including

A. Gonciaruk et al./Microporous and Mesoporous Materials xxx (2014) xxx-xxx

cutting with scissors and snapping the membrane) showed similar features. 1t is possible that snapping the membrane does not produce as clean a break as would be desired but that does not prevent us from concluding that there are structural differences between the two materials. 1t is immediately obvious, in comparison to pure P1M-1, that the composite membrane has a much rougher texture. There is also a much larger number of visible macropores in the internal structure. These features, suggest that the graphene may have influenced the packing of the P1M-1 macromolecules. The larger pore range indicates that the composite material would exhibit higher levels of permeability and faster adsorption but could also be less selective with certain gas separations.

3. Computational method

Models were constructed and computational results were obtained using software programs from Materials Studio® V5.5.3, Accelrys Inc. (San Diego, CA). Interactions between atoms were described using the Dreiding force field [18]. The Lennard-Jones 12-6 potential was used to model van der Waals interactions, whilst electrostatic interactions were calculated using Coulomb's law. Approximate partial atomic charges were specified by the charge equilibration QEq method [19] as implemented in the software. A three-site model was used for the CO2 molecule where two oxygen and carbon atoms are explicitly modelled.

3.1. Structure generation

The polymer builder module implemented in the software is not capable of constructing ladder polymers such as P1M-1. Therefore, based on Heuchel et al.'s procedure [20] the problem was tackled by breaking one of two 5-membered rings in order to create a single-bonded polymer backbone (Fig. 5). Hydrogen atoms numbered 1 and 2 were defined as head and tail atoms, respectively. Two monomers were selected to represent possible configurations of the polymer chain due to different orientations of methyl functional groups (Fig. 5), bonded to 5-membered ring. The group can be oriented either in the same or opposite (referred

to as cis and trans, respectively) direction taking the fused ring system as a reference. Repeat units were selected randomly and connected through carbons 3 and 4, removing head and tail hydrogens during the construction process. Ten random 11-15 monomer long chains were created. Examples of polymer chains are provided in Fig. 6.

Graphene sheets were created in planar form by connecting six-membered carbon rings (Fig. 7). The edge carbon atoms of the graphene were saturated by adding hydrogens.

The structures were randomly packed in an amorphous three-dimensional periodic box at low density using the Amorphous cell module in Monte Carlo fashion. The initial system contained 10 P1M-1 chains of various lengths and one large graphene sheet (Fig. 7) yielding a total of 7752 atoms in the model system. Such composition corresponds to a weight ratio 1:10 (graphene:P1M-1). Three independent models were created to obtain average properties.

Selection of low initial density is dictated by the stiffness of the polymers. Unlike flexible polymers, P1M-1 has a limited number of conformations and therefore cannot be packed directly to final density. 1nstead molecules are placed in a low density box and final configuration is reached during a series of molecular dynamic simulations. Larsen et al. [21] developed a new generic scheme based on that of Karayiannis et al. [22], where the P1M model is consistently compressed and relaxed to experimental density. The slow decompression scheme is stated to be reliable in terms of obtaining realistic density without the need to compare with experimental data. This approach has also been successfully used to model other materials in which their contorted structure prevents them from packing efficiently [23,24]. Simulation conditions are outlined in supporting information.

The initial large graphene slab was split into two and four parts to determine the graphene sheet size effect on the composite properties (Fig. 7). The two additional systems were created containing the same polymer/graphene overall composition, which resulted in either two intermediate or four small sheets of graphene and the same number of P1M-1 chains as in the system with one large graphene. The diameter of the smallest graphene sheet (Fig. 7) is similar to the size of P1M-1 monomer. 1t is expected that smaller sheets will occupy pores more efficiently and, consequently, reduce adsorption capacity and gas diffusivity. Small sheets may also travel in the polymer framework more easily and stack to each other, affecting mechanical properties. Large graphene sheets may serve as a barrier blocking pathways to pores during the adsorption process, especially in membrane separations. This will worsen molecule diffusivity and transport through the membrane as well as reduce accessible free volume. Large sheets also possess higher surface area which may significantly worsen mechanical contact between graphene and polymer. Phase separation may occur due to agglomeration of large sheets, which in turn may make the composite more brittle than pristine polymer. Therefore it is expected to find an intermediate graphene sheet that ensures good mechanical properties and is sufficiently large not to block the material's pores. The resulting composites are labelled Composite L, Composite M and Composite S; the last letter in the name defines the size of the sheets used, i.e., L - large, M - medium and S - small.

All atom electrostatic charges and Lenard-Jones parameters for P1M-1, graphene and CO2 molecules are available in the Supplementary information.

3.2. Structural characterisation

The models were characterised and compared in terms of density, accessible nitrogen surface area, and pore size distribution (PSD). For pore volume, helium (He) atom with kinetic diameter of 2.6 A is used. Poreblazer software [25] was used to generate

Fig. 5. P1M-1 repeat units. Colour code: black - carbon, grey - hydrogen, red -oxygen, and blue - nitrogen. Numeration: 1 and 2 - head and tail atoms, 3,4 and 5 -carbon atoms to be connected during chain construction. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

PSD. Geometric surface area is defined by a line that the centre of the probe draws whilst rolling along the van der Waals surface of the adsorbent. The accessible surface area of an adsorbent is defined in the same way; however regions that cannot be accessed externally are excluded. For calculations of accessible surface area, a nitrogen molecule (kinetic diameter 3.68 A) is chosen because it is the usual probe used in a BET experiment. Of course, if one wants to be consistent, surface area should be calculated also from simulated nitrogen isotherms at 77 K as it was shown that BET and geometric surface areas are not always comparable [26,27]. However such simulations are computationally demanding and disagreement between surface area obtained from different methods is often insignificant [26,28,29].

Poreblazer employs a Monte Carlo procedure to generate PSD. The tested pore is divided into bins. A point is placed in the space of a bin randomly and is tested for overlaps. If no overlaps occur, the largest possible sphere particle is generated that contains the point and does not overlap with the adsorbent. The bin is then incremented by one and the procedure is repeated. Cumulative pore volume function V(d) is generated representing the volume that can be occupied by a probe of diameter d or smaller. PSD function dV(d)/dd can be obtained by differentiating V(d).

Density of the model is simply calculated by dividing model mass by the total volume of the simulation box. However it is thought that skeletal volume should be used instead for more

Table 1

Structural properties of PIM-1/graphene composite and PIM-1.

Density (g/cm3) Skeletal density (g/cm3) Accessible nitrogen surface area (m2/g)

Composite L 0.863 ± 0.026 1.007 ±0.008 853 ±136

Composite M 0.828 ± 0.037 0.998 ± 0.007 983 ±151

Composite S 0.872 ±0.017 1.006 ±0.002 791±108

Composite experimental - 0.999 ± 0.021 -

P1M-1 simulated this work 0.833 ± 0.046 0.977 ±0.013 901 ± 207

P1M-1 experimental this work - 0.948 ± 0.008 -

P1M-1 simulated [reference] - 0.94-1.40 [5,20,21,31,32,37] 435 [20] 448 [21], 830 [27], 940 [31]

P1M-1 experimental [reference] - 0.94-1.4 [31,38,39] 760-875 [6,11,38]

Fig. 8. Representative arrangement of PIM-1 chain fragments on graphene sheet.

Pore width, Â

Fig. 9. Pore size distribution comparison between composite model with varying graphene sheet size and pure PIM-1. Simulated PSD of pure PIM-1 are reprinted from ref [21] and ref 2 [31].

A. Gonciaruk et al. / Microporous and Mesoporous Materials xxx (2014) xxx-xxx

correct comparison with experimental data [30]. Skeletal densities are calculated from the following equation:

upore g

where pexp and psim are skeletal and simulated densities, respectively, Dpore is the volume accessible for helium atom (1.3 A) and g is the mass of the system.

3.3. Simulation details of CO2 adsorption

The sorption module, adsorption isotherm task in Material Studio employs Grand Canonical Monte Carlo simulations (GC MC). Adsorption of molecules was allowed only in the accessible volume defined with the CO2 molecule (kinetic diameter 3.3 A) which is the same as in the experiment. A combination of translation, rotation, insertion, and deletion steps were performed for a total of 5.5 x 106 equilibration and production steps. Interactions between atoms were again described using the Dreiding force field; however Lenard-Jones parameters were adjusted to match experimental and simulated isotherms at low pressures. Gas adsorption was simulated at 293 K over a pressure range from 0 to 20 bar.

4. Results and discussion

4.1. Structural properties

Simulated and experimental structural properties of P1M-1/ graphene composite and pure P1M-1 are provided in Table 1. Properties of P1M-1 simulated in this work compare well with experimental and simulated data available in the literature. This indicates that the generic Dreiding force field is capable to predict correct structures of P1M-1. Some deviations between simulated values and those measured experimentally are observed. They may be due to the presence of defects, residual solvent, impurities and/or kinetically inaccessible regions in experimental samples that are not captured in the ideal simulated sample, or to uncertainties in the force field parameters used, which were not derived for this specific case.

The predicted structures of composites loaded with large and small platelets were slightly denser than simulated P1M-1 and had smaller accessible surface area. However the system with intermediate graphene sheets had higher surface area compared to other composite systems and pure P1M-1. This suggests that it is indeed possible to change the structural properties of the polymer matrix by carefully controlling graphene size.

Visualisation of simulated systems revealed that P1M-1 fragments arranged themselves parallel to the graphene sheet (Fig. 8). The separation between P1M-1 fragments and graphene sheet was about 3.5-4 A, similar to the distances between stacked layers in graphite. Some of the P1M-1 fragments stacked onto the graphene sheet as in hexagonal phase graphite or by repeating a hexagonal ring pattern. However most of P1M-1 chain fragments were constrained by an arrangement of the whole chain and therefore tended to align rather randomly on the graphene sheet. No additional voids were created in the interface between polymer matrix and graphene sheets. Nevertheless such stacking indicates good interface adhesion between the graphene sheet and polymer, which probably facilitates graphene dispersion within the polymer matrix and strengthens the resulting composite material. The mobility of P1M-1 may also be affected due to the contact between some chain fragments and the graphene sheet, which potentially can help controlling the polymer ageing.

Average pore size distributions of the composite and P1M-1 simulated in this work are provided in Fig. 9. PSD calculated for

simulated P1M-1 by G.S. Larsen et al. from two references [21,31], are also included for comparison. Although both of these PSD are generated for P1M-1 using the same method, the two PSDs are slightly different due to one of them being shifted towards higher pore width. This indicates that many possible arrangements occur whilst packing quirky structures such as P1M-1 into amorphous systems. Therefore the differences between PSD of such structures should not be over interpreted.

There might be seen some insignificant differences between PSD of different structures. Most pore sizes are scattered around an average value of 3 A, although PSDs generated in this work are slightly shifted to narrower pore widths compared to the reference PSDs of P1M-1. Nevertheless all PSDs tended to follow a similar trend: peaks arose at similar pore width values and all PSDs produced shoulders towards higher pore sizes.

4.2. CO2 adsorption

The effect of graphene presence in P1M-1 was also tested for carbon dioxide (CO2) adsorption. Adsorption of CO2 was simulated

Fig. 10. Experimental and simulated CO2 isotherms at 293 K. Simulated isotherm calculated using (a) original L-J parameters and (b and c) reduced L-J interaction strength, e, by 24%. Data in (b) and (c) is the same, with (c) showing the detail of the low pressure region. Lines represent the fitting obtained from the DM model. Experimental isotherm of pure PIM-1 is taken from [36].

Table 2

Dual-mode constants at 293 K.

Composite PIM-1

ki (mmol g 1 bar 0.15 0.13

k2 (bar-1) 0.86 0.67

m (mmol g-1) 3.12 3.77

AH (kJ mol-1) 26 24

only in the system with the large graphene sheet (Composite L) and in pure P1M-1. The simulated isotherms compared with those obtained in experimental samples are provided in Fig. 10. As expected, the simulated isotherm follows Langmuirian behaviour, i.e., CO2 adsorption increases significantly at low pressures, whereas at higher pressures a saturation capacity is reached. Experimental isotherms on the other hand tended to have a linear relationship at high pressures between adsorbed CO2 loading and pressure and did not level off even at 20 bar pressure. This indicates possible polymer swelling and increased dominance of gasgas interactions. The difference between the shape of simulated and experimental isotherms is also expected as the framework of the system is considered to be rigid, i.e., adsorption takes place only in fixed free volume regions. 1n real conditions, on the other hand, polymer chains can re-orientate, adjusting to pressure and loading changes. This may affect free volume change leading to greater loadings, especially at higher pressure. This behaviour is observed in both composite and pristine P1M-1. This effect has

already been captured by Holck et al. in their work [32] where they showed the same discrepancies between experimental and simulated isotherms for pure P1M-1 and other swelling glassy polymers. 1n order to match isotherm points at higher pressures it was required to create an additional "swollen" model of the material by introducing gas molecules into the system and allowing material to rearrange its configuration.

From the experimental isotherms it is obvious that addition of graphene has not affected adsorption performance of P1M-1, as the isotherm shape and adsorption capacity is almost the same in both materials. This is also reflected in the DM constants (Table 2) where m, k1 and k2 values, denoting adsorption capacity, swelling and affinity between CO2 and adsorbent, respectively, are almost the same between the composite and P1M-1. Moreover, calculated heats of adsorption are also similar for both samples.

The simulations show a slightly different CO2 adsorption in composite and P1M-1 at pressures higher than 1 bar, which is attributed to different surface areas of the two systems. 1n simulations, both pure P1M-1 and the composite reach the same CO2 loadings up to approximately 1 bar. 1n this pressure region, gas-adsorbent interactions are the most important and no swelling is expected to occur. Comparing adsorption at low pressures will indicate which of the systems possess higher affinity towards adsorbed gas. The fact that both materials adsorb the same amount of CO2 at low pressure indicates that P1M-1 chains possibly stack closely to graphene, lim-

Fig. 11. Representative snapshots of final configuration boxes of Composite M (a) and Composite S (b). Atoms of graphene sheets are coloured in black, all other carbon atoms are dark grey, hydrogens are light grey, nitrogens are blue and oxygen atoms are red. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

A. Gonciaruk et al./Microporous and Mesoporous Materials xxx (2014) xxx-xxx

iting access towards its surface for CO2 molecules, or attraction between CO2 and graphene is simply very similar to the attraction between CO2 and polymer. This encouraging finding suggests that selectivity towards CO2 should not be worsened in such composite membranes. There have been several attempts reported in the literature to enhance P1M-1 adsorption performance by incorporating nanoparticles such as functionalised carbon nanotubes [16], fumed silica [33] and zeolitic imidazolate framework Z1F-8 [34]. 1n all of the cases permeabilities were increased whereas selectivities for CO2/CH4 and CO2/N2 gas pairs were worsened or increased insignificantly.

The adsorption isotherms calculated using default L-J parameters stored within the Dreiding force field overestimated CO2 adsorption (Fig. 10a). By optimising the Dreiding force field we were able to match experimental and simulated isotherms. 1t was achieved by scaling down the strength of the interaction (e) which describes van der Waals interactions between non-bonded atoms. The scaling factor was determined by empirically fitting the calculated adsorption isotherm of CO2 to experimental data measured at 293 K. The main focus was to reproduce the adsorption isotherm at the low pressure region where van der Waals interactions are dominant and it is assumed that no swelling occurs. The obtained scaling factor is 0.76, which reduces the strength of attraction insignificantly considering that only one parameter and only for CO2 molecule atoms is scaled down. This scaling is similar to that required to model accurately CO2 adsorption in other microporous materials [35].

4.3. Graphene size effect

The size of graphene sheets may affect adsorption performance of an adsorbent. We observed that in all boxes of Composite S phase separation occurred, as stacking of two or three graphene sheets was observed (Fig. 11b). The agglomeration occurred at early stages before compression when system density was low (0.169 g cm-3). 1n a real P1M-1/graphene composite, the agglomeration may occur during the preparation procedure, before a dense polymer membrane is formed, which would be a similar case to the simulations.

Graphene size effect is reflected in the structural properties provided in Table 1. Density of the system with the medium graphene sheets is slightly lower than the density of the other composite systems, whilst surface area is larger and exceeds the surface area of pure P1M-1. No agglomeration of graphene molecules was observed in all three boxes of this system (Fig. 11a). This suggests that the hypothesis mentioned earlier that there is an optimum size of graphene sheets that can enhance the material's properties is confirmed: the medium graphene sheet size increases accessible surface area rather than blocking or occupying the pores of the material. However the small difference between properties of the systems studied, including their PSDs, suggests that all systems are quite similar in their structure configurations. Therefore it seems that graphene sheet size, over the size range investigated, does not have a major effect on the bulk structural properties of P1M-1. Further research is recommended to identify the size range of graphene sheets that would affect the structure of the polymer and the adsorption properties, which should probably be larger than the ones presented in this work.

5. Conclusions

The objective of this study was to create a model of P1M-1 loaded with graphene, to understand how graphene affects polymer packing and how it is reflected in structural properties and

subsequently adsorption of carbon dioxide compared to pristine P1M-1 adsorbent. It is evident that graphene has some effect on the bulk structure of pure P1M-1, i.e., the membrane changes its colour and gains a ruptured surface compared to a smoother P1M-1 surface. The rougher structure of the composite's surface suggests a possible increase in permeability and faster adsorption. However the molecular model of the composite demonstrates that the polymer's structural properties such as density, surface area and PSD are preserved in the composite. Moreover, similar adsorption isotherms of pristine P1M-1 and P1M-1/graphene composite suggest that there is no significant effect on affinity towards CO2 caused by the presence of graphene. This leads to expect that graphene can increase P1M-1 membrane permeability without affecting its selectivity.


This work was supported by EPSRC grant EP/K016946/1. AG is grateful for the postdoctoral scholarship from the School of Chemical Engineering and Analytical Science at the University of Manchester.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, in the online version, at 07.007.


[1] R.W. Baker, K. Lokhandwala, 1nd. Eng. Chem. Res. 47 (7) (2008) 2109-2121.

[2] T. Masuda et al., J. Am. Chem. Soc. 105 (25) (1983) 7473-7474.

[3] A. Morisato, 1. Pinnau, J. Membr. Sci. 121 (2) (1996) 243-250.

[4] N.B. McKeown et al., Macromolecules 43 (2010) 5287-5294.

[5] W. Fang, L. Zhang, J. Jiang, Mol. Simul. 36 (12) (2010) 992-1003.

[6] P.M. Budd et al., Chem. Commun. (2004) 230-231.

[7] P.M. Budd et al., J. Membr. Sci. 251 (1-2) (2005) 263-269.

[8] S.D. Kelman et al., J. Membr. Sci. 320 (1-2) (2008) 123-134.

[9] P. Bernardo, E. Drioli, G. Golemme, 1nd. Eng. Chem. Res. 48 (10) (2009) 46384663.

[10] N.B. McKeown, P.M. Budd, Macromolecules 43 (12) (2010) 5163-5176.

[11] P.M. Budd et al., Adv. Mater. 16 (5) (2004) 456-459.

[12] L.M. Robeson, J. Membr. Sci. 62 (2) (1991) 165-185.

[13] L.M. Robeson, J. Membr. Sci. 320 (1-2) (2008) 390-400.

[14] T.-S. Chunga et al., Prog. Polym. Sci. 32 (2007) 483-507.

[15] S. Matteucci et al., J. Membr. Sci. 307 (2008) 196.

[16] M.M. Khan et al., Nanoscale Res. Lett. 7 (2012) 504.

[17] S. Kulprathipanja, R.W. Neuzil, N.N. Li, Separation of Fluids by Means of Mixed Matrix Membranes, Allied-Signal 1nc., US, 1988.

[18] S.L. Mayo, B.D. Olafson, W.A. Goddard, J. Phys. Chem. 94 (1990) 8897-8909.

[19] A.K. Rappe, W.A. Goddard, J. Phys. Chem. 95 (1991) 3358-3363.

[20] M. Heuchela et al., J. Membr. Sci. 318 (1-2) (2008) 84-99.

[21] G.S. Larsen et al., Macromolecules 44 (17) (2011) 6944-6951.

[22] N.C. Karayiannis, V.G. Mavrantzas, D.N. Theodorou, Macromolecules 37 (8) (2004) 2978-2995.

[23] A. Del Regno et al., 1nd. Eng. Chem. Res. 52 (47) (2013) 16939-16950.

[24] A. Del Regno, F.R. Siperstein, Microporous Mesoporous Mater. 176 (2013) 5563.

[25] L. Sarkisov, A. Harrison, Mol. Simul. 37 (15) (2011) 1248-1257.

[26] L.D. Gelb, K.E. Gubbins, Langmuir 14 (8) (1998) 2097-2111.

[27] K.E. Hart, L.J. Abbott, C.M. Colina, Mol. Simul. 39 (5) (2013) 397-404.

[28] K.S. Walton, R.Q. Snurr, J. Am. Chem. Soc. 129 (27) (2007) 8552-8556.

[29] Y.-S. Bae, A.Ö. Yazaydm, R.Q. Snurr, Langmuir 26 (8) (2010) 5475-5483.

[30] L.J. Abbott, C.M. Colina, Macromolecules 44 (11) (2011) 4511-4519.

[31] G.S. Larsen et al., Adsorption 17 (2011) 21-26.

[32] O. Hölck et al., J. Membr. Sci. 428 (2013) 523-532.

[33] J. Ahn et al., J. Membr. Sci. 346 (2) (2010) 280-287.

[34] A.F. Bushell et al., J. Membr. Sci. 427 (2013) 48-62.

[35] J. Pérez-Pellitero et al., Eur. J. 16 (5) (2010) 1560-1571.

[36] G.S. Larsen, K.E. Hart, C.M. Colina, Mol. Simul. 40 (7-9) (2014) 599-609.

[37] G.S. Larsen, Simulations and Experiments on Gas Adsorption in Novel Microporous Polymers, The Pennsylvania State University, 2011.

[38] N.B. McKeown et al., Angew. Chem. 1nt. Ed. 45 (11) (2006) 1804-1807.

[39] P.M. Budd, N.B. McKeown, D. Fritsch, Macromol. Symp. 245-246 (1) (2006) 403-405.