Scholarly article on topic 'Organic molecules of intrinsic microporosity: Characterization of novel microporous materials'

Organic molecules of intrinsic microporosity: Characterization of novel microporous materials Academic research paper on "Chemical engineering"

CC BY
0
0
Share paper
Academic journal
Microporous and Mesoporous Materials
OECD Field of science
Keywords
{Adsorption / "Microporous organic materials" / "Intrinsic microporosity" / "Molecular simulation" / Macromolecules}

Abstract of research paper on Chemical engineering, author of scientific article — Annalaura Del Regno, Flor R. Siperstein

Abstract Molecular simulations were used in this work to characterise a new class of microporous material: organic molecules of intrinsic microporosity (OMIMs). Molecular dynamics simulations were used to generate the material’s samples, and grand canonical Monte Carlo simulations of argon adsorption were used to ascertain the relationship between the different structures and the observed properties. Packing behavior, porosity and adsorption capacity have been determined for each system. The final density of the material, as well as the surface area and pore volume depend on the ending group’s bulkiness. Bulkier molecules lead to materials with lower densities, but it was found that the adsorption behavior is not just related to the material’s density, but also to the pore size and shape, which are determined by the way the molecules pack. The relationship between adsorption capacity and physical properties were analyzed and the role of surface area, free volume and enthalpic interaction were used to identify different adsorption regimes. It was found that the uptake of argon at low pressure is proportional to the strength of the adsorbent-adsorbate interaction while at moderate pressure it is dependent on the free volume and surface area.

Academic research paper on topic "Organic molecules of intrinsic microporosity: Characterization of novel microporous materials"

.i.irnm '-■■tf/mimmar-

ELSEVIER

Contents lists available at SciVerse ScienceDirect

Microporous and Mesoporous Materials

journal homepage: www.elsevier.com/locate/micromeso

Organic molecules of intrinsic microporosity: Characterization of novel microporous materials

Annalaura Del Regno, Flor R. Siperstein *

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

ARTICLE INFO ABSTRACT

Molecular simulations were used in this work to characterise a new class of microporous material: organic molecules of intrinsic microporosity (OMIMs). Molecular dynamics simulations were used to generate the material's samples, and grand canonical Monte Carlo simulations of argon adsorption were used to ascertain the relationship between the different structures and the observed properties. Packing behavior, porosity and adsorption capacity have been determined for each system. The final density of the material, as well as the surface area and pore volume depend on the ending group's bulkiness. Bulkier molecules lead to materials with lower densities, but it was found that the adsorption behavior is not just related to the material's density, but also to the pore size and shape, which are determined by the way the molecules pack. The relationship between adsorption capacity and physical properties were analyzed and the role of surface area, free volume and enthalpic interaction were used to identify different adsorption regimes. It was found that the uptake of argon at low pressure is proportional to the strength of the adsorbent-adsorbate interaction while at moderate pressure it is dependent on the free volume and surface area.

© 2013 Elsevier Inc. All rights reserved.

Article history:

Received 15 February 2013

Received in revised form 27 March 2013

Accepted 28 March 2013

Available online 8 April 2013

Keywords: Adsorption

Microporous organic materials Intrinsic microporosity Molecular simulation Macromolecules

1. Introduction

Microporous materials are solids that contain interconnected pores of the order of molecular dimensions (less than 2 nm), property that makes them interesting for a series of applications such as heterogeneous catalysis, adsorption, separation and gas storage. During the past decade many materials have been explored, from traditional microporous materials like zeolites, silica and activated carbons to a wide new range of organic based networks like metal-organic frameworks (MOFs) [1-3], covalent organic frameworks (COFs) [4-6] and hyper-crossed-linked polymers (HCP) [7,8].

Crystalline materials have the advantage of a well controlled pore size and shape, nevertheless, many adsorbents used in industry are amorphous, as they can have other desirable properties. New microporous polymers, such as polymers of intrinsic microporosity (PIMs) [9-13], behave in part as typical adsorbents due to their rigid structure, but also as polymers with a higher flexibility than most carbons and silicas. Careful selection of the monomers has led to a variety of microporous polymers: spirobifluorene based polyamydes [14], conjugated microporous polymers [15] and element organic frameworks [16,17].

Crystalline and amorphous microporous materials can also be obtained with molecules. For instance, ordered cage-like materials

* Corresponding author. E-mail address: flor.siperstein@manchester.ac.uk (F.R. Siperstein).

1387-1811/$ - see front matter © 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.micromeso.2013.03.041

can be synthesized through dynamic covalent interactions [18], where a group of atoms (such as a coordination compound or simply an imine) create a temporary bond with the organic cage during its formation, but they are not part of the final compound. The cages then assemble to form crystalline porous materials. Other ordered materials can exploit hydrogen bonding between small molecules leading to microporous molecular crystals, as shown by Mostarlerz and Oppel [19]. Intrinsic microporosity is also observed in triptycene-based materials [20-22] due to their high internal molecular free volume which pack inefficiently. Microporosity therefore, can be generated by chemical-physical interactions and also by a clever manipulation of the structure geometry, both for relatively small molecules and polymers.

Organic molecules of intrinsic microporosity (OMIMs) [13,23] are a new class of molecules formed with similar building block than those used to prepare PIMs, but with a significantly lower molecular weight. They have similar chemistry to PIMs and pack inefficiently due to their highly concave, ''awkward'' shape.

Mathematical studies of two dimensional disks and three dimensional superballs [24,25] proved that a system's packing density can be modulated by changing the grade of its members concavity; the lowest system's density corresponds to the highest number of concave faces of its units. The building blocks of OMIMs were therefore designed to imitate this concave geometry [13,23]; they consist of a core which determines the geometry, and termini that provide the general environment for these materials (Fig. 1).

Fig. 1. Schematic representation of a generic OMIM structure; the core is represented in black, the termini in gray.

A series of biphenyl core OMIMs have been recently reported by Abbott et al. [23]; in this work a series of packing protocols were compared to determine a realistic and consistent model in order to represent accurately these new materials. Three different type of OMIMs were investigated, exploring different termini conformations but preserving the core structure and the aromaticity of each system; trigonal (2,3-dihydroxytriptycene (OMIM-1)), tetrahedral (2,3-dihydroxy-9,9'-spirobisfluorene (OMIM-2)) and octahedral configurations (2,3-dihydroxyhexabenzopropellane (OMIM-3)) were analyzed and the properties of simulated samples compared with experimental wide-angle X-ray scattering (WAXS). This work showed that the relationship between the OMIM structure and the materials properties is complex. Therefore, a better understanding of the packing behavior is essential in order to efficiently screen series of possible termini that will lead to the desired material's performances.

In this work the influence of the termini's chemical nature on gas adsorption was studied. The cross-shaped core was preserved (Fig. 2 - Core) and the environment of the termini was systematically changed. Starting from the simplest OMIMs type, the benzene and the naphthalene based OMIMs (Fig. 2 - termini A and E), different substituents such as tert-butilate, adamantan and triptycene

were added to the basic aromatic linear terminus (Fig. 2 - termini B, C, D for the benzene family, F and G for the naphthalene one).

We aim to understand the packing behavior of these two families, determine the relationship between the terminus chemistry and/or bulkiness with the packing ability and the adsorption properties in order to anticipate the properties of the final material knowing the nature of cores and termini. Different representative samples of each material were generated using molecular dynamics simulation. The structural properties (density, surface area, pore size distribution and porosity) were characterized. We used grand canonical Monte Carlo (GCMC) simulations of argon to investigate the role of the surface area and pore size distribution as well as adsorbent/guest energetics in determining adsorption of small molecules in these novel materials. We found many similarities in the adsorption behavior compared to other porous materials, such as MOFs, where different adsorption regimes can be identified: those proportional to the surface area or pore volume, which suggest that the knowledge developed for other porous materials is transferable to OMIMs despite the disordered nature of the structures they form.

2. Methodology

2.1. OMIMs structure

The two families of OMIMs, the benzene and the naphthalene based OMIMs (Fig. 2 - pink and blue blocks) were represented using fully atomistic models. The interactions between atoms were described using the OPLS force field [26] and the charges were calculated through the charge equilibration algorithm (QEq) [27].

The Qeq method, introduced by Rappe et al. in 1996, has been widely used as it provides a rapid method for estimating point charges located on each atom. The point charges are calculated based on the molecular geometry and experimental atomic properties: atomic electron affinity and ionization potential. In this method, the potential energy of an atom is described as a Taylor expansion of its partial charge. The first and second derivatives

Fig. 2. Chemical structures of termini for benzene family (pink block; A - benzene-1,2-diol, B - 4-tert-butylbenzene-1,2- diol, C - 3,5-di-tert-butylbenzene-1,2-diol, D - 2,3-dihydroxy-triptycene), naphthalene family (blue block, E - naphthalene-2,3-diol, F - 6-tert-butylnaphthalene-2,3-diol, G - 6-adamantannaphathalene-2,3-diol) and core (green block; 4,4'-dicyano-biphenyl) precursors and the resulting cross-shaped OMIMs with related acronyms.

Table 1

21-Step NVT/NPT molecular dynamics compression/decompression protocol [32].

Step Ensemble Conditions3 Length (ps)

1, 2 NVT 600 K, 300 K 50, 50

3 NPT 300 K, 0.02Pmax 50

4, 5 NVT 600 K, 300 K 50, 100

6 NPT 300 K, 0.6Pmax 50

7, 8 NVT 600 K, 300 K 50, 100

9 NPT 300 K, Pmax 50

10, 11 NVT 600 K, 300 K 50, 100

12 NPT 300 K, 0.5Pmax 5

13, 14 NVT 600 K, 300 K 5, 10

15 NPT 300 K, 0.1Pmax 5

16, 17 NVT 600 K, 300 K 5, 10

18 NPT 300 K, 0.01Pmax 5

19, 20 NVT 600 K, 300 K 5, 10

21 NPT 300 K, Pfinal 800

a Pmax = 50,000 bar, Pfinal = 1 bar.

of this potential energy are related to the electron affinity and ionization potential. The method considers that the chemical potential of all atoms are equal at equilibrium. Defining the chemical potential as the derivative of the total electrostatic potential with respect to the atom charge, one obtains a systems of equations that allow calculating the optimum charge distribution [27]. Extensions of this method have been recently proposed for rapid screening of porous materials [28,29]. This method has shown to provide good

Probe (Ar)

Physically accessible pores Kinetically accessible pores

Fig. 3. Schematic representation: (top) surface area and accessible surface area; (bottom) physically accessible pores and kinetically accessible pores.

estimations of the partial charges for polymeric systems, including polyethylene, poly (vinylidene difluoride, poly (tetra-fluoroethyl-ene, poly (oxymethylene), nylon 66 and poly (ether-ether-ketone) [27] polymethyl metacrylate and polystyrene blends [30] as well as nafion and Teflon membranes [31].

Three boxes starting from independent configurations were generated for each OMIM. Molecules were packed randomly into a low density box of approximately 0.1 g/cm3, avoiding overlaps of the van der Waals radii. A 21-step NVT/NPT molecular dynamics (MD) compression protocol was performed.

Details are shown in Table 1.

The method, initially proposed by Larsen et al. [32] in simulations of PIMs, as a modification of the method proposed by Karay-iannis et al. [33] and consists of a compression (steps 1-9, Table 1) and the decompression protocol (steps 10 to 21, Table 1). The final configuration is achieved at the end of the last NPT MD step (step 21, Table 1) consisted of an 800 ps simulation at 1 bar and room temperature. The resulting method which does not required prior knowledge of the material's density has also been used in simulations of hyper-cross-linked polymers [34] and on a series of cross-shaped OMIM systems [23]. The approach has been validated by comparing structural properties of the virtual material to experimental WAXS scattering data, and excellent agreement in both the peak positions and intensities has been found in both cases. All simulations were performed in Gromacs 4 [35].

2.2. Gran canonical Monte Carlo simulations

GCMC simulations of argon adsorption were performed in the Sorption module using Material Studio 5.5 [36]. Argon was described as a single Lennard-Jones sphere and intermolecular interaction between argon and the adsorbent framework were described using the Universal force field [37] and calculated using the Lorentz Berthelot mixing rules. All simulations were carried at the normal boiling temperature for argon (87.15 K), from the pressure of 10-11-10-4 bar (Henry's law region) in order to emphasize a pressure region normally difficult to investigate experimentally. At each pressure, a combination of translation, rotation, insertion and deletion steps were performed for a total of (3 - 10) x 106 MC to obtain equilibrium. The Lennard-Jones potential was used to estimate the van der Waals interactions; no Coulombic interactions were considered as the framework was assumed to be rigid during the GCMC simulations. Canonical Monte Carlo simulations were then performed over one sample box for each system using the locate module in Materials Studio 5.5 [36] to find the preferential, i.e, lowest energy sites for each systems at infinite dilution. Simulations were conducted with a fixed loading of one molecule of argon over one box and consisted of five annealing cycles during which the system was cooled from 10,000 to 50 K. The results were used for a qualitative understanding of differences in heat of adsorption between the OMIM's systems studied.

2.3. Force field

In this work, two different force fields have been used: OPLS [26] and Universal Force field [37]. The first one was used to describe the materials during the MD packing procedure while the second one was used to describe the interactions with argon during the GCMC simulations. The OPLS force field was parametrised directly to reproduce experimental thermodynamic and structural data in fluids, therefore is computationally efficient and highly specific; UFF, on the other hand, was developed based only on the element type, hybridization and connectivity, resulting into a really generic Force field, useful to describe every atom in the periodic table. Few works have been presented where the OPLS-AA force field was used to describe adsorbent frameworks; adsorption of

Table 2

Properties of each independent simulated byphenyl-core OMIMs.

Terminus Box molecules Density (g/cm3) Surface Area (m2/g) Free Volume (cm3/g)

Benzene family

1 1.065(1) 47(3) 0.012(1)

BNZ 2 207 1.073(1) 29(6) 0.007(2)

3 1.057(1) 66(4) 0.016(1)

1 0.864(2) 224(11) 0.059(4)

BNT 2 153 0.876(2) 173(6) 0.045(2)

3 0.882(3) 123(5) 0.030(1)

1 0.798(1) 236(8) 0.060(2)

BDT 2 121 0.790(2) 293(13) 0.078(4)

3 0.796(2) 271(4) 0.072(1)

1 0.860(2) 475(11) 0.125(3)

TRI 2 98 0.847(3) 520(18) 0.142(6)

3 0.840(3) 581(19) 0.158(6)

Naphthalene family

1 1.014(2) 166(2) 0.043(1)

NFT 2 157 1.018(2) 169(8) 0.045(2)

3 1.027(3) 154(3) 0.040(1)

1 0.874(2) 176(11) 0.044(3)

NTB 2 123 0.872(2) 182(8) 0.045(2)

3 0.858(1) 266(5) 0.071(2)

1 0.838(2) 395(9) 0.109(3)

ADA 2 95 0.842(1) 349(12) 0.093(4)

3 0.860(1) 312(10) 0.080(2)

Average and error evaluated over a total of fifteen boxes. Standard deviations are given in parenthesis.

hydrogen, carbon dioxide and methane was modeled by Yang et al. [38-40] in a series of metal organic frameworks while Zhou et al. [41] studied the adsorption of hydrogen in ZIF-8. Even though OPLS-AA successfully described the interaction within the metal organic frameworks, interaction with the adsorbent were either refitted [38-40] or combined with other force fields [41] to obtain a better agreement with the experimental results. In 2003 Vish-nyakov et al. [42] compared argon adsorption isotherms in MOF-5 using different force-fields: UFF, OPLS-AA, and OPLS-UA. The results showed that generic force fields overestimated the Henry's constant, but the UFF resulted to be the most accurate between the generic force fields analyzed, predicting well the filling pressure of the main channels. Given the absence of experimental data for argon adsorption in OMIMs, and based on the previous works, we decided to describe the molecules during the packing procedure using a more specific force field, e.g. OPLS-AA, which was successfully used to describe the organic ligands in metal-organic frameworks, and Universal force field to describe interactions with argon.

2.4. Characterization

Molecular models were characterized in this work by densities, surface areas and pore volumes. All results were averaged over five boxes generated from the final relaxation steps for each independent simulation for a total of fifteen boxes for each system. The

Table 3

Properties of benzene based OMIMs.

systems' snapshots were collected during the last 100 ps of the final relaxation steps with 20 ps gaps between each box. The porosity in the simulations was determined from the geometric surface areas and pore volumes. In this approach the surface area is defined by rolling a probe molecule along the surface of atoms. A probe size equal to the diameter of argon molecule used in the GCMC simulations (dAr = 3.868 A) was used [37]. In our calculation pore volume and surface area take into account all the physical accessible pores; no distinction is made between physically and kinetically accessible pores as the entire porosity is considered in the GCMC simulations of argon adsorption (Fig. 3). The fractional free volume is calculated to compare the porosity of each system independently from their density as

Fractional free volume(FFV) = ^^ vojume

Bulk volume

while the pore size distribution (cm3/(g nm)) is calculated as the differential of the fractional free volume normalized by the density (q: cm3/g) versus the pore diameter (U: nm) for a future comparison with experimental data

Pore size distribution(PSD) =

d(FFV/q) d(U)

All the adsorption isotherms were collected using the final configuration for each independent MD simulation, for a total of three

Terminus Density3 (g/cm3) Surface Areaa (m2/g) Free Volumea (cm3/g) -AHb (kJ/mol) uptakeb@10-" bar (mmol/g) uptakeb@10~4 bar (mmol/g)

Benzene family BNZ 1.065(7)

BNT 0.874(8)

BDT 0.795(4)

TRI 0.849(9)

65(20) 173(43) 267(26) 526(47)

0.012(4) 0.045(12) 0.070(8) 0.141(14)

19.15(78) 18.08(18) 19.28(36) 20.45(50)

3.06x10~5(1.55) 1.48x10~5(0.12) 4.63x10~5(0.77) 20.03x10~5(3.72)

2.164(122) 2.721(146) 3.265(53) 5.185(378)

■ /*\ " BNZ-PSD

2.0 ; ♦ BNT-PSD

? c ■ i m A BDT-PSD

I'-W/ • TRI-PSD

Ol 1.5 »»a A AV d BNZ-FFV

m \\\\ o BNT-FFV

E o q" 1.0 ■ VW A ~A X«, \iVv\ ° BDT-FFV TRI-FFV

0.8 0.7 0.6

0.4 §

0.3 Îl LL

0.2 0.1

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Pore width, nm

o.o .0

10° 10 10 Pressure, bar

Fig. 4. Benzene family: (a) Fractional Free Volume (open symbols) and Pore Size Distribution (full symbols). Results represent the average value over a total of fifteen boxes. Error bars are not included for clarity. Relative errors are smaller than 2% for probe diameters larger then 0.4 A. (b) Argon adsorption isotherms for benzene family OMIMs. Results shown are average values over three boxes obtained from independent initial configurations. Error bars are smaller than the symbol size.

samples analyzed for each OMIM; loadings and heats of adsorption were therefore averaged over three boxes for each system.

3. Results

3.1. Validation of the simulated systems

Molecular models of OMIMs were generated using the 21-steps NVT/NPT MD compression protocol [32,34,23]. This technique is a parameter free method; the final density therefore was not fixed, allowing every species to reach the most favorable packing density. The average calculated properties consisting of five samples for

each independent starting configuration are reported in Table 2. In examining the properties for each data set it is clear that differences in densities between samples generated for the same NPT trajectory are less than 1%. Even the smallest change in density influences strongly the structural properties of the materials, for instance, the benzene OMIM (BNZ) shows a sensible deviation between structural properties and densities displacement. The variation of the systems's densities are lower than 0.1% for every boxes. It is fundamental to underline that the only difference between independent boxes generation was the initial random orientation of molecules in each box; number of molecules, starting volume and initial density were kept constant for each OMIM. Each independent simulations leads to samples that represent a different arrangement of the packed molecules. BNZ that is the densest OMIM, can pack in different ways leading to samples that have a variations in surface area of more than 100%. Similar variations are observed for TRI and BDT, the bulkiest OMIMs studied here, which can clearly orientate their termini creating different pore sizes that may be not accessible to argon probe used for our characterization, despite a similar density. The naphthalene family members show smaller variations between densities, surface areas and pore volumes compared to the benzene family members. Despite having a longer termini, OMIMs in the naphalene have a smaller variation in their packing behavior presenting liquid-crystal type structures. Such behavior is observed in a variety of liquid crystals included naphthalene based liquid crystals, well known in literature [43-45].

3.2. Benzene family

The average calculated properties for the benzene family OMIMs are shown in Table 3. Surface area and pore volume increases with the terminus's bulkiness going from a minimum value for BNZ (SA = 65 m2/g; FV = 0.012 cm3/g) to a maximum observed for TRI (SA = 526 m2/g; FV = 0.141 cm3/g). The SA is inversely pro-protional to the system's densities except BDT; despite having the lowest density, this OMIM shows lower SA and FV than TRI. The presence of two tert-butyl groups, leads to the formation of large number of pores smaller than the argon diameter. This is illustrated with the PSD in Fig. 4(a).

The pore size distribution curves for BNZ, BNT and BDT species has a similar shape with a maximum in the distribution observed for pores of 0.125 nm diameter, suggesting that those materials have pores of comparable shape, with the only difference being the total free volume, directly correlated with the density.

The shape of the PSD curve for TRI suggests a higher amount of larger pores with a maximum at a probe diameter of 0.175 nm. TRI shows the highest concentration of pores with a diameter within the range of 0.175-1 nm than any other member of the family while it has the lowest distribution of pores with a diameter smaller than 0.175 nm compared to the other family members.

GCMC simulations of argon adsorption were collected for the final configuration of the three independent simulation boxes. The calculated adsorption isotherms for the benzene based family are

Table 4

Properties of naphthalene based OMIMs.

Terminus Density3 (g/cm3) Surface Areaa (m2/g) Free Volumea (cm3/g) -AHb (kJ/mol) uptakeb@10~" bar (mmol/g) uptakeb@10~4 bar (mmol/g)

Naphthalene family

NFT 1.019(6) 163(8) 0.043(3) 19.35(54) 6.53x10~5(1.97) 3.010(62)

NTB 0.868(7) 208(43) 0.053(13) 18.89(33) 3.40x10~5(0.74) 3.529(14)

ADA 0.847(10) 352(37) 0.094(12) 19.22(90) 6.47x10~5(4.64) 4.118(348)

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.1 Pore width, nm

10"° 10"' Pressure, bar

Fig. 5. Naphthalene family: (a) Fractional Free Volume (open symbols) and Pore Size Distribution (full symbols). Results represent the average value over a total of fifteen boxes. Error bars are not included for clarity. Relative errors are smaller than 2% for probe diameters larger then 0.4 A. (b) Argon adsorption isotherms for benzene family OMIMs. Results shown are average values over three boxes obtained from independent initial configurations. Error bars are smaller than the symbol size.

shown in Fig. 4(b) while the average calculated properties are shown in Table 3. At low pressure (10-11-10-6 bar) the TRI species presents the highest uptake followed by BDT, BNZ and BNT respectively; at moderate pressure (10-6-10-4bar) an inversion is observed between the BNZ and the BNT species leading to an argon uptake in the order TRI > BDT > BNT > BNZ.

3.3. Naphthalene family

The average calculated properties for the naphthalene based OMIMs are shown in Table 4.

The same trend observed for the benzene family is recognized in the naphthalene one: surface area and pore volume increases with the terminus's bulkiness. The smallest SA and FV are observed for the simplest OMIM of the family, the NFT (SA=163m2/g; FV = 0.043 cm3/g) while the highest value coincided with the ADA (SA = 352 m2/g; FV = 0.094 cm3/g). Densities decrease with increasing the bulkiness of the substituent, as observed previously. Small differences in densities between NTB and ADA lead to a sensible difference in SA and FV. The presence of the tert-butyl groups, as seen before in the BDT case, leads to the formation of small pores, as confirmed by the pore size distribution of the naphthalene family (Fig. 5(a)). NFT and NTB PSDs show a similar trend to the benzene family, with a maximum in the distribution observed

Amount adsorbed®» 10" bar, mmol/g

(b) 22.0

21.5 -

21.0 -

20.5 - • -

1 20.0 - -

19.5 - □

5 19.0 - ■ / yS O

18.5 - ^ -

18.0 - ♦ _

17.5 -

Amount adsorbed@ 10 bar, mmol/g

Fig. 6. Correlations between amount adsorbed and physical properties of the adsorbents. Naphthalene family's members represented by open symbols: □ NFT, 0 NTB, o ADA; benzene family's members represented by full symbols: ■ BNZ, ♦ BNT, N BDT, • TRI. (a) Heat of adsorption vs the amount adsorbed at 10-11 bar. (b) Heat of adsorption vs amount adsorbed at 10-4 bar.

for pores of diameter equal to 0.125 nm. The shape of the PSD curve for ADA, suggests the presence of larger pores with a maximum distribution of pores with a diameter equal to 0.175 nm. Similarly with what observed for TRI, ADA shows an higher concentration of pores with a diameter within the range of 0.175-1 nm and a lower distribution of pores with a diameter smaller than 0.175 nm compared to the other family members. GCMC simulations or argon adsorption were collected for the final configuration of each independent simulation box; the GCMC results were averaged over three boxes for each OMIM. The calculated adsorption isotherms for the naphthalene based family are shown in Fig. 5(b) while the average calculated properties are shown in Table 4. At low pressure (10-11-10-6 bar) ADA and NFT present the maximum uptake followed by NTB; at moderate pressure (10-6-10-4 bar) an inversion is observed between the three species leading to an argon uptake in the order ADA > NTB > NFT.

4. Discussion

The dependency of density on the bulkiness of the terminus has been recognized for benzene and naphthalene families and thoroughly discussed in Sections 3.2 and 3.3.

Looking into the details of the adsorption isotherms of each OMIMs, it is possible to recognize a relationship between the uptake at low pressure and the enthalpic interaction with sorbed molecules; in the benzene family (Table 3) AH at low coverage varies between -20.45(50) kJ/mol in the TRI and -18.08(18) kJ/mol in

(a) 0.20

Amount adsorbed@ 10" bar, mmol/g

(b) 0.20—.—I—.—I—.—I—.—I—.—I—.—

Amount adsorbed@ 10" bar, mmol/g

Fig. 7. Correlations between amount adsorbed and physical properties of the adsorbents. Naphthalene family's members represented by open symbols: □ NFT, 0 NTB, o ADA; benzene family's members represented by full symbols: ■ BNZ, ♦ BNT, N BDT, • TRI. (a) In black free volume vs the amount adsorbed at 10-11 bar; in red surface area vs amount adsorbed at 10-11 bar. (b) In black free volume vs amount adsorbed at 10-4 bar; in red surface area vs amount adsorbed at 10-4 bar.

the BNT samples; in the naphthalene family (Table 4) ADA and NFT show the highest AH (-19.22(90) and -19.35(54) kJ/mol, respectively) while NTB shows the lowest value of AH (-18.85(46) kJ/ mol).

The uptake of argon at the pressure of 10-11 bar (e.g. low coverage) follows an exponential relationship with the heat of adsorption, as shown in Fig. 6(a). No relationship is recognized between heat of adsorption and amount adsorbed at moderate pressure (Fig. 6(b)). At low pressure, no relationship is recognized between the argon uptake and the structural properties such as surface area and free volume (Fig. 7(a)). The relationship between Henry's constant and the enthalpy of adsorption is given by the second virial coefficient [46,47], where the amount of adsorbed gas (nad) is given by

nad = B— (3)

where P is the pressure, R is the gas constant, T is the temperature and B is the second gas-solid virial coefficient. The latter is proportional to the exponential of the gas-solid interaction potential [46,47] and it can be defined as [47]

Fig. 8. Minimum energy pores for benzene family OMIMs; (a) BNZ, (b) BNT, (c) BDT, (d) TRI species: at the top, 3D images of the minimum energy pores, at the bottom, 2D sketch of the pores (gray surface) structures.

B = Vp exp — exp

where Vp is the adsorbent pore volume, AS0 and Ah are respectively the standard entropy and enthalpy of adsorption. Therefore, it is expected that the enthalpy of adsorption at low coverages will

Fig. 9. Minimum energy pores for naphthalene family OMIMs; (a) NFT, (b) NTB, (c) ADA species: at the top, 3D images of the minimum energy pores, at the bottom, 2D sketch of the pores (gray surface) structures.

be proportional to the logarithm of the amount adsorbed, as shown in Fig. 6(a).

Considering that the chemical nature of the OMIMs studied in this work is similar, the explanation for the differences observed in the heat of adsorption has to be searched into the pores structures that each OMIM displays (Fig. 8 and Fig. 9); BNZ and BNT (Fig. 8(a) and (b)) have rectangular cages where different OMIMs orient their termini parallel to each other, leading to a narrow elongated pores; BNT, because of the presence of the tert-butyl group has a larger rectangular pore, compared to BNZ. BDT on the other hand, tends to have square cages (Fig. 8(c)); the presence of two tert-butyl groups forces each molecule to display its own harms perpendicular; the final pore is formed by two or three molecules facing their centres, resulting in more compact pores than BNZ and BNT. TRI shows spherical shaped pores (Fig. 8(d)) where three or four OMIMs dispose their triptycene groups opposite to each other, forming a pore that is larger (in terms of diameter) than the other family's members, but with a larger curve surface, where the sorbed molecule is completely surrounded by aromatic rings.

A similar arrangements of OMIMs is observed in the naphthalene family; NFT and the NTB form rectangular cages where

different OMIMs orient their termini one parallel to each other creating narrow elongated pores (Fig. 9(a) and (b)). Despite the similarity between BNT (Fig. 8(b)) and NTB (Fig. 9(b)), the pores in the NTB case tend to be narrower (behavior that is also confirmed by the slightly higher AH of the latter); the effect of the bulky tert-bu-tyl group at the end of the terminus is not as strong as in the BNT case, being the arms in NTB longer. ADA also create rectangular cages; in this case, however, the OMIMs dispose the terminal adamantan groups at the end of the cage, leading to more compact pores (Fig. 9(c)). At moderate pressure (10~4 bar), argon uptake is proportional to the surface area and the free volume (Fig. 7(b)). At this stage, all the narrowest and the lowest energy pores are completely full, and an inversion is observed; around the pressure of 10~6 bar, both BNT for the benzene family and NTB for the naphthalene one exceed the argon uptake of BNZ and NFT. These results are consistent with the observations of Frost et al. [48] for hydrogen adsorption in metal organic frameworks; in that work a variety of MOFs with same chemistry but different pore volume and surface area were compared; as it has been observed here, the uptake at low pressure was directly related to heat of adsorption while at higher pressure surface area and pore volumes had the major influence on the uptake of small molecules.

5. Conclusions

Molecular simulations of novel organic molecules of intrinsic microporosity were performed. The development of virtual models of amorphous materials is challenging and even when a robust method has been used to generate model structures, certainty of the obtained structures accuracy is difficult to claim without explicit validation against experimental data. Nevertheless, we consider that simulations can provide an advantage in prescreening materials that have not been synthesized. Therefore, transferability of the force-fields used is essential, and future efforts should be devoted to address this issue.

Simulated structures were characterized and compared. The termini bulkiness is inversely proportional to the density. Within a family surface area and free volume are directly related with the bulkiness of the terminus and can be modulated choosing the right substituents. In order to obtain a material with high surface area and free volume rather than low density, the addition of cluttered substituent has to be preferred over the number of substituents added; the presence of two tert-butyl groups in the case of BDT, lead to a low density material, with a large number of small pores, compared to other member of the family. GCMC simulation of argon adsorption isotherms confirmed the results obtained for adsorption of hydrogen in MOFs, where three distinctive regimes were recognized [48]. Adsorption at low pressure is related to the heat of adsorption through an exponential relationship while at higher pressure the uptake of argon is directly related with surface area and porosity of the material. Differences in heat of adsorption between different OMIMs can be explained with the analysis of the pores, where for simple OMIMs like BNF, BNT, NFT and NTB rectangular like pores were observed while for more complex OMIMs pores resulted to be more compact moving from a rectangular closed shape in the ADA case, to a square cage for BDT and a sphere like pore for TRI. TRI terminus leads to the material with best performance at low and high pressures because it creates cavities with strong interactions and curve surface, and packs inefficiently leading to a high pore volume.

Acknowledgments

The authors thank the Engineering and Physical Sciences Research Council for funding (Grant EP/G065144/1). We are

grateful to Neil B. McKneown and Rupert G. D. Taylor from Cardiff University for useful discussions; we thank Carlos Avendaño Jimenez and Richard Gowers from the University of Manchester for helping with the preparation of the manuscript.

References

[1] M. Eddaoudi, D.B. Moler, H. Li, B. Chen, T.M. Reineke, M. O'Keeffe, O.M. Yaghi, Acc. Chem. Res. 34 (2001) 319-330.

[2] M. Eddaoudi, J. Kim, N. Rosi, D. Vodak, J. Wachter, M. O'Keeffe, O.M. Yaghi, Science 295 (2002) 469-472.

[3] N.L. Rosi, J. Eckert, M. Eddaoudi, D.T. Vodak, J. Kim, M. O'Keeffe, O.M. Yaghi, Science 300 (2003) 1127-1129.

[4] A.P. Côté, A.I. Benin, N.W. Ockwig, M. O'Keeffe, A.J. Matzger, O.M. Yaghi, Science 310 (2005) 1166-1170.

[5] H.M. El-Kaderi, J.R. Hunt, J.L. Mendoza-Cortés, A.P. Côté, R.E. Taylor, M. O'Keeffe, O.M. Yaghi, Science 316 (2007) 268-272.

[6] A.P. Côté, H.M. El-Kaderi, H. Furukawa, J.R. Hunt, O.M. Yaghi, J. Am. Chem. Soc. 129 (2007) 12914-12915.

[7] M. Tsyurupa, V. Davankov, React. Funct. Polym. 53 (2002) 193-203.

[8] M. Tsyurupa, V. Davankov, React. Funct. Polym. 66 (2006) 768-779.

[9] P.M. Budd, B.S. Ghanem, S. Makhseed, N.B. McKeown, K.J. Msayib, C.E. Tattershall, Chem. Commun. (2004) 230-231.

[10] B.S. Ghanem, K.J. Msayib, N.B. McKeown, K.D.M. Harris, Z. Pan, P.M. Budd, A. Butler, J. Selbie, D. Book, A. Walton, Chem. Commun. (2007) 67-69.

[11] B.S. Ghanem, N.B. McKeown, P.M. Budd, D. Fritsch, Macromolecules 41 (2008) 1640-1646.

[12] B.S. Ghanem, N.B. McKeown, P.M. Budd, N.M. Al-Harbi, D. Fritsch, K. Heinrich, L. Starannikova, A. Tokarev, Y. Yampolskii, Macromolecules 42 (2009) 78817888.

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

[14] J. Weber, Q. Su, M. Antonietti, A. Thomas, Macromol. Rapid Comm. 28 (2007) 1871-1876.

[15] J. Schmidt, M. Werner, A. Thomas, Macromolecules 42 (2009) 4426-4429.

[16] M. Rose, W. Bohlmann, M. Sabo, S. Kaskel, Chem. Commun. (2008) 2462-2464.

[17] E. Stockel, X. Wu, A. Trewin, C.D. Wood, R. Clowes, N.L. Campbell, J.T. Jones, Y.Z. Khimyak, D.J. Adams, A.I. Cooper, Chem. Commun. (2009) 212-214.

[18] M. Mastalerz, Angew. Chem. Int. Ed. 49 (2010) 5042-5053.

[19] M. Mastalerz, I.M. Oppel, Angew. Chem. Int. Ed. 51 (2012) 5252-5255.

[20] J.H. Chong, M.J. MacLachlan, J. Org. Chem. 72 (2007) 8683-8690.

[21] J.H. Chong, M.J. MacLachlan, Chem. Soc. Rev. 38 (2009) 3301-3315.

[22] J. Chong, S.J. Ardakani, K. Smith, M. MacLachlan, Chem. Eur. J. 15 (2009) 11824-11828.

[23] L.J. Abbott, A.G. McDermott, A. Del Regno, R.G.D. Taylor, C.G. Bezzu, K.J. Msayib, N.B. McKeown, F.R. Siperstein, J. Runt, C.M. Colina, J. Phys. Chem. B 117 (2013) 355-364.

[24] Y. Jiao, F.H. Stillinger, S. Torquato, Phys. Rev. Lett. 100 (2008) 245504.

[25] Y. Jiao, F.H. Stillinger, S. Torquato, Phys. Rev. E 79 (2009) 041309.

[26] W.L. Jorgensen, D.S. Maxwell, J. Tirado-Rives, J. Am. Chem. Soc. 118 (1996) 11225-11236.

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

[28] C.E. Wilmer, K.C. Kim, R.Q. Snurr, J. Phys. Chem. Lett. 3 (2012) 2506-2511.

[29] E. Haldoupis, S. Nair, D.S. Sholl, J. Am. Chem. Soc. 134 (2012) 4313-4323.

[30] M. Drache, J. Reichel, Macromol. Theor. Simul. 11 (2002) 878-883.

[31] E.J. Lamas, P.B. Balbuena, Electrochim. Acta 51 (2006) 5904-5911.

[32] G.S. Larsen, P. Lin, K.E. Hart, C.M. Colina, Macromolecules 44 (2011) 69446951.

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

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

[35] B. Hess, C. Kutzner, D. van der Spoel, E. Lindahl, J. Chem. Theory Comput. 4 (2008) 435-447.

[36] Accelrys Software Inc., Materials Studio, Release 5, Accelrys Software Inc., San Diego, 2007.

[37] A.K. Rappe, C.J. Casewit, K.S. Colwell, W.A. Goddard, W.M. Skiff, J. Am. Chem. Soc. 114 (1992) 10024-10035.

[38] Q. Yang, C. Zhong, J. Phys. Chem. B 109 (2005) 11862-11864.

[39] Q. Yang, C. Zhong, J. Phys. Chem. B 110 (2006) 655-658.

[40] Q. Yang, C. Zhong, J. Phys. Chem. B 110 (2006) 17776-17783.

[41] M. Zhou, Q. Wang, L Zhang, Y.-C. Liu, Y. Kang, J. Phys. Chem. B 113 (2009) 11049-11053.

[42] A. Vishnyakov, P.I. Ravikovitch, A.V. Neimark, M. Bülow, Q.M. Wang, Nano Lett. 3 (2003) 713-718.

[43] J. Svoboda, V. Novotna, V. Kozmik, M. Glogarova, W. Weissflog, S. Diele, G. Pelzl, J. Mater. Chem. 13 (2003) 2104-2110.

[44] T. Mori, M. Kijima, Chem. Lett. 36 (2007) 710-711.

[45] T. Mori, M. Kijima, J. Polym, J. Polym. Sci. A Polym. Chem. 46 (2008) 42584263.

[46] O. Meeks, T. Rybolt, J. Colloid Interf. Sci. 196 (1997) 103-109.

[47] A.L. Myers, Colloid. Surface. A 241 (2004) 9-14.

[48] H. Frost, T. Düren, R.Q. Snurr, J. Phys. Chem. B 110 (2006) 9565-9570.