Introduction
Palygorskite (Plg) is a fibrous clay mineral (Galán, Reference Galán1996). Due to its significant porosity and specific surface area, Plg has been used widely in industry and engineering and, therefore, is a very important non-metallic mineral (Singer & Galán, Reference Singer and Galán2000). In practical applications, Plg is used as an adsorbent to treat toxic metal ions, organic contaminants, etc., and also as a molecular sieve to separate oil and water (Shariatmadari et al., Reference Shariatmadari, Mermut and Benke1999; Frost et al., Reference Frost, Xi and He2010; Chen et al., Reference Chen, Liu, Li, Chen, Chang, Kong and Frost2011; Sheikhhosseini et al., Reference Sheikhhosseini, Shirvani, Shariatmadari, Zvomuya and Najafic2014; Rusmin et al., Reference Rusmin, Sarkar, Biswas, Churchman, Liu and Naidu2016; Liu et al., Reference Liu, Wei, Liang, Chen, He, Chen, Xi, Chen, Han and Zhu2018; Zhu et al., Reference Zhu, Zhang, Wang, Wen, Su, Zhu, He and Xi2018). These applications are all based on the unique nano-sized pore structures and interfacial properties of Plg. The physical and chemical properties of the pores are of the utmost importance, therefore.
Palygorskite is an aluminosilicate with Fe as the most common isomorphic substitution (Singer et al., Reference Singer, Huertos and Galán2011). The Plg reserve in the Jiangsu-Anhui region is the largest in China (Gao et al., Reference Gao, Chen, Wu, Huang and Wang2009) and several studies have reported that Fe is a common feature of the Plg from this area (Yi et al., Reference Yi, Li, Tian and Zheng1995; Long et al., Reference Long, McDonald, Facheng, Houjei, Zili and Xu1997; Yang & Chen, Reference Yang and Chen2004). Fe substitution usually happens in the octahedral sheets according to Liu et al. (Reference Liu, Chen, Chang, Chen, Qing, Xie and Frost2013). The influence of Fe substitution on the properties of Plg pores has not been investigated previously, however.
The pores of Plg are typically nanosized. Current experimental techniques to quantify directly the pore properties are still difficult and very few studies have been reported. Extended X-ray absorption fine structure (EXAFS) has been applied to study the location of zeolitic water in the crystal tunnels and the dehydration of Plg (Post & Heaney, Reference Post and Heaney2008). To the current authors’ knowledge, systematic research on the microstructure and mobility of cations and water molecules in the tunnels of Plg is lacking.
With the development of computational simulations, attempts have been made to develop a reliable forcefield for various clay minerals (Teppen et al., Reference Teppen, Rasmussen, Bertsch, Miller and Schäfer1997; Titiloye & Skipper, Reference Titiloye and Skipper2001; Cygan et al., Reference Cygan, Liang and Kalinichev2004). ClayFF (Cygan et al., Reference Cygan, Liang and Kalinichev2004) is the most widely used clay force field. The molecular dynamics (MD) method has been applied widely in the fields of mineral science and geochemistry for decades (Mulla et al., Reference Mulla, Cushman and Low1984; Skipper et al., Reference Skipper, Refson and McConnell1991; Chang et al., Reference Chang, Skipper and Sposito1995; Liu & Lu, Reference Liu and Lu2006; Wang et al., Reference Wang, Kalinichev and Kirkpatrick2006; Zhang & Choi, Reference Zhang and Choi2006; Kumar et al., Reference Kumar, Kalinichev and Kirkpatrick2007; Anderson et al., Reference Anderson, Ratcliffe, Greenwell, Williams, Cliffe and Coveney2010; Greathouse et al., Reference Greathouse, Cygan, Fredrich and Jerauld2016; Giri et al., Reference Giri, Teixeira and Cordeiro2018). The interface properties of smectites have been investigated extensively by using MD simulations (Ferrage et al., Reference Ferrage, Lanson, Sakharov, Geoffroy, Jacquot and Drits2007, Reference Ferrage, Lanson, Michot and Robert2010; Liu et al., Reference Liu, Lu, Wang and Zhou2008; Greathouse et al., Reference Greathouse, Hart, Bowers, Kirkpatrick and Cygan2015). In contrast, only a few MD studies were concerned with fibrous clay minerals. Fois et al. (Reference Fois, Gamba and Tilocca2003) tried to introduce molecular dynamics simulations to explain the stability of a palygorskite-indigo complex (Maya blue paint). The structure of water molecules in the pores of sepiolite and palygorskite was studied by Ockwig et al. (Reference Ockwig, Greathouse, Durkin, Cygan, Daemen and Nenoff2009) but those authors did not study the spatial distribution and mobility of water molecules. By using MD simulation, Zhou et al. (Reference Zhou, Lu and Boek2016) investigated the distributions and dynamics of zeolitic water in sepiolite pores and found that their mobility is very low due to the nano-confinement effects.
In order to reveal the effect of Fe substitution on the distribution and mobility of water and cations, which is difficult to do experimentally, systematic MD simulations were performed to investigate the microscopic properties of the pores of Fe-containing and Fe-free Plg. Through detailed comparisons between Fe-containing and Fe-free Plg, the influence of Fe substitution could be determined.
Methods
Models
The initial structures of Fe-containing and Fe-free palygorskites (denoted as Fe-Plg and Al-Plg, respectively) are shown in Fig. 1. Their chemical formulae are Na0.5(Si7.5Al0.5)(Mg2Al2)O20(OH)2(OH2)4·n(H2O) and Na0.5(Si7.5Al0.5)(Mg2FeAl)O20(OH)2(OH2)4·n(H2O). Here OH meant hydroxyl, OH2 was the coordinated water of Mg, H2O stood for zeolitic water, and n was taken as 3.5 in this study. The lattice parameters were: a = 13.286 Å, b = 17.848 Å, c = 5.242 Å, α = 90°, β = 107.56°, γ = 90°. These parameters were taken from the results for Plg samples from Anhui, China (YF Cai). The atomic coordinates of the Plg unit cell were modified from Post and Heaney (Reference Post and Heaney2008). The simulation supercells consisted of 8 unit cells (2a×1b×4c) with 3D periodic boundary conditions.
Computation Details
The force field used for the MD simulation was ClayFF(Cygan et al., Reference Cygan, Liang and Kalinichev2004) (details in Table 1). This force field has been applied successfully in simulating clay minerals including smectites, kaolinite, sepiolite, and palygorskite (Cygan et al., Reference Cygan, Greathouse, Heinz and Kalinichev2009; Ockwig et al., Reference Ockwig, Greathouse, Durkin, Cygan, Daemen and Nenoff2009; Anderson et al., Reference Anderson, Ratcliffe, Greenwell, Williams, Cliffe and Coveney2010; Underwood et al., Reference Underwood, Erastova and Greenwell2016).
ClayFF is a non-bonded forcefield for clay framework. The bond stretch term is required only by the hydroxyl groups.
Non-bonded van der Waals interactions are calculated through the Lennard-Jones 12-6 potential term:
where ε and σ represent the van der Waals depth and length, respectively. and σ ij = 0.5(σ i + σ j ). The electrostatic interactions are calculated through coulombic potential terms and treated using the Ewald summation.
where q i and q j are the partial charge of atom i and j. e is the charge of the electron,
NVT MD simulations were carried out using the LAMMPS package (Plimpton, Reference Plimpton1995). Each simulation was run for 5 ns with a time step of 1 fs, and the last 3 ns was taken as the production run.
Results and Discussions
Distributions of Water in Pores
The two profiles of the density distributions of water O in Plg pores along the y direction (i.e. parallel to the tunnels) are very similar (Fig. 2). The snapshot for Al-Plg (Fig. 3) was used as an example because the structures in Fe-Plg and Al-Plg were almost the same. By integrating the density distributions and the snapshots it was clear that the sharp peaks on the density distributions denoted the coordinated water of Mg and the shoulders in the middle were zeolitic water.
In the simulation periods of the two Plg models, all coordinated water molecules of Mg atoms vibrated only near the equilibrium positions but could not escape. The radial distribution function (RDF) (Fig. 4) for water molecules around Mg presented a very sharp peak at ~2.18 Å, and its corresponding coordination number (CN) was 2. This peak denoted the coordinated water.
Taking Al-Plg as an example, Fig. 5 showed a snapshot of Mg coordinated water in the pores. Two coordinated water molecules for each Mg and the 4 O atoms in the crystal structure together formed the six-fold coordination of Mg. The agreement between the simulation results of Fe-Plg and Al-Plg indicated that the presence of Fe in the crystal lattice did not affect the local coordination structure of the edge Mg atoms.
On the density distributions (Fig. 2), zeolitic water molecules did not show obvious peaks, indicating that they did not have specific distributions, but could diffuse freely. H-bonds (Fig. 3) between zeolitic water and six-membered ring oxygen and between the upper zeolitic water and lower zeolitic water could be found clearly. From the comparison of two density distribution curves of Fe-Plg and Al-Plg (Fig. 2), Fe in the structure did not influence the spatial distribution of zeolitic water. The parameters feo and mgo (Table 1) were similar and the interactions between zeolitic water and octahedral Fe/Al were also similar, therefore. On the other hand, the distributions of zeolitic water were affected mainly by interaction with closer Na+.
Distribution of Na+
The distribution of Na+ ions derived from Al-Plg channels is shown as an example (Fig. 6), and the situation in Fe-Plg was very similar. Na+ was located above the six-memberedSi–O ring on the surface and close to the substituting ion. Na+ ions, therefore, interacted directly with surface O atoms and the water molecule O (Fig. 7). According to the RDF-CN curve (Fig. 8), each Na+ ion was coordinated with three zeolitic water molecules, on average, and with two surface O atoms, but did not interact directly with the coordinated water of Mg (Fig. 9). The distance between Na and bound water is >3 Å. The coordinated water of Mg, therefore, vibrated around Mg only. The two RDF-CN curves almost overlapped. The comparisons revealed that the influence of Fe on the distribution and structure of Na+ was negligible.
Mobility of Pore Species
The MSD (mean squared displacement) curves of the species in Plg tunnels are noted to be very similar (Fig. 10). The curves of Mg-coordinated water were close to 0 Å2, corresponding to the rather rigid coordination as discussed above. The curves of Na+ ions were almost horizontal, indicating very limited mobility. Using the two models, the MSDs of zeolitic water were notably greater and showed obvious diffusion within the tunnels. The self-diffusion coefficients (D) of zeolitic water was calculated from the MSD curves (Eq. 4). The sampling frequency of MSD was every 100 fs and the sample time was 2 ns. These comparisons showed that the self-diffusion coefficient of zeolitic water in Fe-Plg was a little greater than that in Al-Plg. The magnitudes of self-diffusion coefficients were both ~10–11 m2/s, however. The Fe-substitution in the structure, therefore, did not affect the mobility of the species significantly.
The MSD pattern of Na+ showed a plateau which indicated that the movement of Na+ was very restricted by the Plg tunnels. Previous studies showed that in the interlayer space of smectites, Na+ ions favor being hydrated and, thus, have notably greater mobility. Their self-diffusion coefficients obtained by fitting MSD curves were of the order of ~10–10 m2/s. New Na+ forcefield parameters were used by Ho et al. (Reference Ho, Criscenti and Greathouse2019) in ClayFF and those authors found that the mobility of Na+ increased, but the Na+ MSDs still showed a plateau with new Na+ parameters in Plg. On the contrary, K+ ions had relatively low mobility because they preferred staying above the surface Si–Osix-membered rings and bonded simultaneously with surface oxygen atoms and interlayer water molecules. As shown above, Na+ ions were always fixed above the six-membered rings and, therefore, scarcely showed obvious diffusive behavior. This feature was very similar to that of K+ ions in smectites. The special distribution of Na+ was caused by the unique nano-sized tunnels of Plg.
The self-diffusion coefficients of zeolitic water calculated for the two Plg systems were very close, i.e. ~10–11 m2/s. This value was much smaller than that of bulk water (~10–9 m2/s) and also than that of the interlayer water of montmorillonites (~10–10–10–9 m2/s) (Holmboe & Bourg, Reference Holmboe and Bourg2013; Zhang et al., Reference Zhang, Lu, Liu, Zhou and Zhou2014). These values were close to the diffusion coefficient of zeolitic water in sepiolite pores calculated by Zhou et al. (Reference Zhou, Lu and Boek2016). The poor mobility of zeolitic water molecules indicated that the Plg pores had a very strong spatial confining effect on them, which was similar to that of sepiolite.
The mobility results of Na+ ions and water molecules showed that the one-dimensional nano-sized tunnels of Plg imposed very strong spatial limits on the species. These observations infer that other molecules such as pollutants could also be fixed firmly in the pores.
Conclusions
In the present study, molecular dynamics simulations were carried out to simulate Fe-containing and Fe-free Plg systems, with the focus on the influence of Fe substitution on the physical and chemical properties of the Plg pores. Through detailed analyses of the MD trajectories, the distribution and mobility of water and cations in the pores were derived. The following conclusions were drawn: (1) Mg-coordinated water vibrated near the equilibrium position and could not diffuse; (2) zeolitic water had much poorer mobility than the bulk and interlayer water in montmorillonites due to the very strong nano-confining effect of the pores; and (3) Na+ ions were distributed above the Si–Osix-membered rings, and they showed almost no diffusivity. Systematic comparison indicated that the difference between Fe and non-Fe systems was negligible, inferring that the environmentally important properties of Fe-containing Plg, such as adsorption and fixation of toxic substances, were similar to those of ordinary Plg,
Acknowledgments
The authors acknowledge the National Natural Science Foundation of China (Nos. 42002036 and 41572027) and are grateful to the High-Performance Computing Center (HPCC) of Nanjing University for the numerical calculations (on its blade cluster system) used here.
Funding
Funding sources are as stated in the Acknowledgments.
Declarations
Conflict of Interest
The authors declare that they have no conflict of interest.