An investigation into the crystal structures of Na-J(BW) and Li-A(BW)

Studies in Surface Science and Catalysis,volume 158 J. (~ejka,N. Zilkovfiand P. Nachtigall (Editors) 9 2005 Elsevier B.V. All rights reserved.

105

An investigation into the crystal structures of Na-J(BW) and Li-A(BW) D. Salih a, B. Slater a, D . W . Lewis b, M.A. G r e e n a'b and W. van B e c k c

aDavy Faraday Research Laboratory, The Royal Institution of Great Britain, 21 Albemarle Street, London W 1S 4BS, United Kingdom. bCentre for Theoretical and Computational Chemistry, Department of Chemistry, University College London, 20 Gordon Street, London W 1H 0AJ, United Kingdom. CESRF (European Synchrotron Radiation Facility), 6 Rue Jules Horowitz, BP 220, 38043 Grenoble CEDEX 9, France. A combined computational and experimental study has been undertaken to resolve cationdirecting effects. Li-A(BW) and Na-J(BW) have different topologies but are synthesised from almost identical starting ingredients, the alkali metal being the only variable. This exemplar reaction therefore allows the effect of the cations upon crystallisation and nucleation processes to be directly inferred. We report the minimum energy structures of Li-A(BW) and Na-J(BW) obtained by application of an atomistic approach and the experimental synthesis of Li-A(BW) and Na-J(BW). The synthesis was prompted by inconsistencies between the calculated Na + cation positions of Na-J(BW) and those determined from previous experimental work. With the support of molecular dynamics simulations, we present a redetermination of cation positions as part of an ongoing investigation into the crystal growth mechanism of these two materials. 1. INTRODUCTION Since the first commercialisation of zeolites, much of synthetic chemistry reported in the literature has focused on empirical approaches to tailor-make zeolites, with less emphasis upon why or how these phases are formed [ 1]. An array of factors influence which phase is formed, arising from competing processes during zeolite nucleation and growth [1]. The difficulty in deciphering exactly how microporous materials form is due in part to persistent questions about the identity and role of solution species in directing growth. Several attempts have been made to explore the solution species present during zeolite growth, both experimentally, through spectroscopic methods [1,2], and computationally [3-6], exploring minimum energy structures of possible precursor units. These works support theories of polyand mono-meric species crystallising from solution upon pre-formed solid surfaces in the solution [ 1,7]. The broader aim of this work is to understand the role that the cation plays in influencing the species present in solution and to predict how the cation influences the reaction energetics and mechanisms of attaching species to the crystal surface. The surface structures and crystal morphologies of zeolites can give great insight into how and why the growth of certain

106 structures occur [8]. To determine the crystal growth mechanisms, it is necessary to predict the most stable surface structures, as these define the rate determining steps in the crystal growth process. Here we present the first step in exploring the zeolite growth mechanisms by modelling the minimum energy configurations of the bulk zeolite structures of Li-A(BW) and Na-J(BW). Li-A(BW) and Na-J(BW) are synthesised from gels containing A1203, SiO2, H20 and alkali under identical hydrothermal conditions [9-12]. Media rich in lithium favours the growth of Li-A(BW) [Li4(H20)41[AI4Si4016]whilst sodium rich solution favours the growth of Na-J(BW) ]Na+3(H20)1.sI[A13Si3OI2] and critically, as is evident from the structural formulae, the Si:A1 ratio is 1:1. Consequently, the ion present in solution must have a directing influence on the zeolite growth process and hence the framework topology formed. A study of the two zeolites therefore affords insight not only into the general mechanism of zeolite growth but also the structure directing effects of the extra-framework cations Li + and Na + independent of other structure directing influences. The cations are expected to affect the distribution of oligomers and also the energy barriers of forming particular oligomers in solution. Consequently, we expect to understand how these oligomers are incorporated into the growing crystal surfaces, independent of the other synthesis parameters. Although synthesised by Barrer and White in the early 1950s [13,14], the structure of LiA(BW) and Na-J(BW) was not reported until over 30 years later. Both Li-A(BW) and NaJ(BW) were reported to comprise of 4, 6, and 8-membered rings, connected differently and forming two distinct topologies, ABW and JBW, respectively. Hydrated cations lie within the 8-ring channels of both Li-A(BW) and Na-J(BW) and anhydrous cations are present only in the 6-ring channels of Na-J(BW). The structural details of Li-A(BW) are well documented, supporting the initial model proposed by Norby et al. [ 15-17]. The structure of J(BW) zeolites containing germanium framework species and extra-framework potassium as well as sodium interstitial cations have been reported [18,19], but the structure of pure Na-J(BW), containing sodium cations within the channels has been reported with less confidence. The structure proposed by Hansen et al. [20] identified two partially occupied hydrated sodium cation sites, labelled as Na(3) and Na(31) respectively. The Na(31) sites were proposed to replace 20% of the equivalent Na(3) sites. The structure of Na-J(BW) was also characterised by Ragimov et al. [21], although the zeolite was identified simply as a sodium aluminosilicate and the distribution of the two hydrated sodium cation sites within the 8-ring channels was noted to be "distributed statistically". In order to predict surface structure, an accurate representation of the zeolites bulk structures is a pre-requisite. Here we present preliminary results of computational atomistic modelling of the minimum energy configurations of the zeolite bulk structures. Inconsistencies with published experimental data prompted the synthesis of Na-J(BW) and modelling of the thermal behaviour of Na + cations from which a re-determination of the cation positions was made. 2. STATIC LATTICE MODELLING The structural and thermodynamic properties of numerous zeolites have been extensively and very successfully studied using static lattice methods [22,23]. In this instance, the minimum energy configuration of the bulk zeolite structures of Li-A(BW) and Na-J(BW) were obtained by application of the static energy minimisation program GULP [24]. Species within the 3D simulation cells were described using the Born model of ionic solids, where each atom is represented by a point charge and non-Coulombic interactions are described by van der Waals attractive and repulsive forces under periodic boundary conditions. Zeolite materials possess

107 the highly polarisable oxygen ion within their framework and may contain water molecules within their pores and channels. Thus, the polarisability of the oxygen ions can be approximated by the shell model of Dick and Overhauser [25]. The intermolecular potentials used to model the interactions between the component species of Li-A(BW) and Na-J(BW) were those previously fitted to dense aluminosilicate zeolites and applied to describe a number of dense zeolite frameworks by Lewis et al. [26-28] and are therefore expected to be readily transferable to these systems. Water is described by potentials due to de Leeuw et al. [29]. For each material, the proposed cell was relaxed at constant pressure such that the all species and cell parameters are free to vary. Note that each structure is relaxed to mechanical equilibrium at an effective temperature of OK.

2.1. Results: Li-A(BW) The minimum energy configuration of Li-A(BW) is in good agreement with the experimental model proposed by Norby et al. [15]. The unit cell parameters are predicted within 3% and the framework bond lengths within 1% of the experimental model (Fig. 1). The principal structural discrepancy is in the orientation of the water molecules, where the lone pairs are observed to coordinate more strongly to the Li + cations in the minimum energy configuration (Fig. 1), thus underestimating the Li-water separation. The lithium ions within the experimental model were described as lying in a nearly perfect tetrahedron of oxygen atoms, three framework O (2), (3), (4), and one water Ow oxygen atom [15,17]. The simulated Li + coordination sphere agrees to within 6% of the experimental model but displays a distorted tetrahedral geometry with highly anisotropic Li-O distances in

(b) Fig. 1. (a) Li-A(BW) experimental and (b) optimised model. Lithium ions are depicted as black spheres, the framework as grey sticks and water molecules are displayed as cylinders (view along [ool]) Table 1 Comparison of experimental [15, 17] and calculated Li-A(BW) Li + coordination sphere: Li-O distances (A), and percentage difference between the calculated and experimental model. O(1-4) are framework oxygen atoms whereas Ow denotes a water oxygen atom. Li-O distances Experimental Model Optimised Model Difference (%) Li-Ow

1.97(1)

1.85

-5.75

Li-O2

2.00(2)

2.03

1.70

Li-O3

1.91(1)

1.97

3.08

Li-O4

1.98(2)

2.03

2.37

Li-O1

3.24(1)

3.37

3.82

108 comparison with the experimental model (Table 1). The distortion of the LiO4 tetrahedron indicates possible failures of the fitted Buckingham potential which describes the Li-O interaction. The pronounced decrease in the Li-Ow distance suggests that the potential describing the Li-Ow interaction is too attractive. However, it is possible that this overestimation is simply a thermal effect, a postulate that we will explore using molecular dynamics. Nevertheless, the existing model is in an acceptable agreement with published data and can therefore be used for (interatomic potential) surface structure predictions.

2.2. Results; Na-J(BW) The minimum energy configuration of the l x l x l unit cell of Na-J(BW) is in good agreement with that proposed by Hansen and F~ilth [20] as their 'general' proposed structure, i.e. full occupation of the Na(3) site and no occupation of the Na(31) sites. The unit cell parameters agree to within 2%, and the framework bond length and angles agree to within 3% of the experimental model. In order to correctly model the experimentally determined unit cell composition of Na-J(BW), the possible permutations of the 20% occupancy of Na(31) cation sites were considered by increasing the size of the periodic unit cell to form a supercell of lxlx5 Na-J(BW) unit cells (a=16.426, b=15.014, c=26.120A). The permutations considered were: (a) Na(3) sites with occupancy of 1, (b) 1 in 5 Na(3) sites substituted for an Na(31) site located as far apart as possible (Fig. 2.), (c) 1 in 5 Na(3) sites substituted for an Na(31) site located as close together as possible, and (d) Na(31) sites with occupancy of 1 using the experimental model as the starting configuration. Conflicting with both experimental models, the calculated minimum energy structure was that in which all Na(3) sites, and hence no Na(31) sites, were occupied by the hydrated Na + cations. Therefore, in an attempt to provide new experimental data, upon which the credibility of the findings from the modelling of Na-J(BW) can be measured, a synthesis of Na-J(BW) was undertaken.

(b)

t

,! Fig. 2. (a) Na-J(BW) experimental model with Na(3) sites fully occupied and (b) 1 in 5 Na(3) sites substituted for an Na(31) site located as far apart as possible. Na+ cations in Na(3) sites are modelled as black spheres and those to that in a Na(31) site as a grey sphere, with the framework and water modelled as grey sticks.

109 3. E X P E R I M E N T A L P R O C E D U R E

A prerequisite for understanding cation directing effects is that both Li-A(BW) and Na-(JBW) are synthesised under identical conditions, differing only in the alkali present in solution. The approaches used to synthesise Na-J(BW) were therefore utilised to synthesise Li-A(BW) for consistency. A recently proposed synthesis method for both Na-J(BW) and Li-A(BW) [12] using metakaolin, resulted in the successful synthesis of both zeolites. Here uncalcined kaolin (-~46% SiO2, -~39% A1203) rather than the meta-kaolin precursor was used, preparing gels of composition A1203"2SIO2"1.45Na20"31H20 and A1203"2SIO2-1.45Li20-31H20 treated under hydrothermal conditions at 473K for 96h. X-ray diffraction data obtained (at 298K) on the BM1B station (X = 0.7997 A) at the European Synchrotron Radiation Facility, Grenoble, confirmed the framework type of our LiA(BW) and Na-J(BW) samples. Refinement of the structure of Na-J(BW) by application of Rietveld refinement and Le Bail extraction for two impurity phases (which were assigned to sodalite and vishnevite (Na6.sKl.zCa0.12(Si6A16024)(SO4)0.96(H20)2, isostructural to hydroxycancrinite (Nas(Si6A16Oz4)(OH)l.4(CO3)'3(H20)6.35)) gave preliminary results which indicate approximately 40% occupation of Na(31) sites, about double that reported by Hansen et al. [20] (Table 2). These initial results give an indication of the positions and occupations of the two hydrated Na + sites, as the presence of two other phases induces greater errors and difficulty in refinement of a single phase within a sample. Table 2 H~,drated Atom Na(3) Na(31)

sodium atomic positions and populations x y ,, z 0.210 0.510 0.325 0.6 + 0.024 0.212 -0.008 0.213 0.4 + 0.024

25000 -

20000 15000

E 10000 O

o

5ooo

0

IIII IIIIIIIIIIIIIIHIIIIllllllIIIHIliiNlililllliniI

I

I I I I I I I II il

I

'

1'o

I IIII

'

Ill Ill Ill Ill II Illllllllllllllllllllllllllllllllllllllllllllll Ill IIIIIIIILIHIIliL IlllHIIlil Nlli IIIIIIIIIIIIIlillllllllllllllll

io

'

3'o

'

4'o

IIIIlilllllllllllii

'

5'o

2 Theta, deg

Fig. 3. Refined XRD pattern of the prepared Na-J(BW) sample (wRp = 11.5%). Observed and difference pattern below the base line. Vertical tickmarks below the XRD pattern identify the peak positions of the phases, 1 (Na-J(BW)), 2 (sodalite) and 3 (vishnevite) respectively.

110 Since the experimental diffraction data was obtained at room temperature and that OK atomistic energy minimisation ignores thermal factors present at ~273K, modelling the dynamics of the system at higher temperatures was performed to understand thermal sensitivity of cation site occupations. 4. M O L E C U L A R DYNAMICS SIMULATIONS The dynamics of the framework and interstitial species were modelled by application of molecular dynamics. Evaluation of Newton's laws of motion gives the position and velocities of each atom at each time-step, allowing the evolution of the system over time to be observed. Molecular Dynamics (MD) simulations were implemented using the DL_POLY code [30]. In this study the isobaric-isothermal NPT ensemble was applied. A time-step of 0.5fs (0.0005ps) was used for a period of up to 112.5ps. Simulations were performed within periodic boundary conditions and the minimum image convention at temperatures of 100, 200, 300 and 400K. The system was allowed to equilibrate at each temperature over a timescale of up to 37ps. Once equilibrated to the target temperatures, the final high temperature configurations were optimised by quenching to OK, to assess the stability of the cation site occupations. The atomic trajectories sampled during the simulation were analysed to identify whether migration of Na(3) to Na(31) sites occurred. Above 300K, cation motion was activated and 0.504.0/"'i (b) 2< 0.45.}i. (a) ?< 3.50.40r / ,' 3.0 71/ ,, 0.35-1!!'I .-/"- . . . . . . . . . . . . . . . ::>.....

lJ \\'i

ii~

,";.... "....

lh

........................... ..

0

"t,.,,\,

2.5

t~

\\

i5 2.0

9 0.20 ~,'i,",~:..:2;2:~:-~.:~~-~~~ "'."~t " ;~. , . i t '-~-- ' ,::+-~.~%-~-2:.:Ls _ :=~-:~

\

1.5

\-.. ""-.. .....

co 1.0 v,~ ,A r, r..... _ . t-

o. o o os

0.5

.....

0 . 0 0 ~

,

0

(c)

I

~

8_i'

0.0 ~ " 0

,

Time / ps

12 t

....................................................

87.5

, Time / ps

35]

;-

---~

, 82.5

(d) F"

7< 30

/ ' .... .,'--

"~ 9 254~/

///"

"~ .......

~ - ~

~

~ : ~ ~

.......

.....................

0 13-

._~ 1:3

6

~-

6

u- 10 m ~'~,;....

4

A

t\ '\

4 0" 09

\

r-

t

56.5

112.5

2

,

0

T i m e / ps

i

112.5

0

'

0

I

T i m e / ps

112.5

Fig. 4. Mean square deviations (A2) (y axis) of hydrated sodium cations within the Na-J(BW) channels over a time period (x axis) of approximately lOOps at (a) 100K, (b) 200K, (c) 300K and (d) 400K with the MSD between 4-8A.2 and time 60-112.5ps highlighted inset. The x-axis in each depicts time in time-steps of the evolution of the simulation.

lll migration of hydrated sodium cations were observed along the a c plane, populating Na(31) sites. To quantify which cations move from which sites, the sites were labelled and the mean square displacement (MSD) of all 96 Na(3) sites within the simulation cell plotted as a function of time (Fig 4.). Note that the average distance between the two hydrated Na(3) and Na(31) is 2.89.A, and hence a total cumulative MSD of approximately 8.36~ indicates a characteristic migration ofNa + from the Na(3) to the Na(3 l) site. Although we do not observe an MSD of-8/~ until 300K, at 200K it can be seen that the sodium ions are activated and vibrate around their mean Na(3) positions. At 300K clear migration can be seen, approximately 5:91 (-1:18) Na + cations migrate to Na(31) sites. The occupation of Na(31) sites are found to increase with temperature; at 400K the ratio of Na(31)/Na(3) is observed to become 9:87(-1:9), approximately double that at 300K. In each case, when quenched to 0K the Na(31) sites remain occupied with no migration back to the original Na(3) site. Analysis of the trajectories at 400K shows a linear translation within a channel (parallel to c), in which a series of Na(3) cations migrate along the z axis, through the channel. The MSDs were deconvoluted into cartesian components which indicated that the MSD in the z axis was approximately 20 times larger than that of x and y. Over the duration of the simulation, we only observed this migration once and hence longer timescales and a larger simulation cell is required to accurately resolve how frequently this event occurs. In Fig. 4(d), one cation is found to displace rapidly into an as yet uncharacterised location -~5]~ from its original site. Almost concurrently, 4 cations become displaced into Na(31) locations, which over the course of the simulation, percolate to neighbouring Na(3) sites approximately 5.57A away from their original position, effecting a translational shift or current. This observation suggests a potential co-operative diffusion mechanism that has been observed in far more porous zeolites by, for example, Auerbach et al. [31 ]. 5. DISCUSSION Modelling of the Li-A(BW) phase gave structural geometries in agreement with previous experimental studies. However, a modelling study of Na-J(BW) revealed discrepancies between the predicted occupation at OK and reported experimental data. Structural refinement of a 3 phase sample of Na-J(BW) yielded Na(3) and Na(31) positions and relative occupations of 0.6 and 0.4 respectively, considerably higher than that previously reported. However, the molecular dynamics simulations showed a pronounced increase in the occupation of Na(31) sites with temperature. Moreover, once migration from Na(3) to Na(31) has occurred, the reverse migration does not occur on a fast timescale and is not promoted by reducing temperature. During synthesis therefore, activated Na(31) sites would be expected to remain occupied. It is conceivable that fluctuations in temperature during synthesis due to heating apparatus and local hot-spots in the synthesis mixture could lead to high occupations of Na(31). Clearly further simulation and experimental work is needed to understand the thermal dependence on occupation of the Na(31) site. However, these initial results indicate that migration is activated between 200 and 300K giving a lower bound for the activation barrier. Simulations carried out at a range of temperatures would allow us to extract the activation barrier and hence deduce what temperature would give rise to the populations extracted from the refinement, which will provide an additional test of the integrity of both the simulation and the refinement. In future work, predictions of the Li-A(BW) and Na-J(BW) surface structures will be given. The templating effects of Na + and Li+ ions in aluminosilicate rich solution and their

112

effects upon the size and distribution of oligomers will be discussed, making use of the methodology recently applied to study siliceous solution fragments [32]. Using constrained optimisation procedures to estimate barrier heights of precursor condensation reactions, we will present preliminary predictions of the solution species and how these units crystallise from solution upon the crystal surface. The conclusions drawn from these computational experiments will be used to provide preliminary insights into the growth processes and the role of cation direction within solution and at the crystal surface during zeolite crystallisation. REFERENCES

[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14]

[15] [16] [17]

[~8] [191 [20] [21] [22] [23] [24]

[25] [26] [27] [28] [29]

[30] [3~] [321

R. Szostak, Molecular Sieves, Blackie Academic and Professional, 1998. C.E.A. Kirshhock, R. Ravishankar, F. Verspeurt, P.J. Grobet, P.A. Jacobs, J.A. Martens, J. Phys. Chem. B., 103 (1999). C.R.A. Catlow, D.S. Coombes, D.W. Lewis, J.C.G. Pereira, Chem. Mater., 10 (1998) 3249. J.C. Pereira, C.R.A. Catlow, G.D. Price, Chem. Commun., (1998) 1387. B.P. Feuston, S.H. Garofalini, J. Phys. Chem., 94 (1990) 5351. N.Z. Rao, L.D. Gelb, J. Phys. Chem. B., 108 (2004) 12418. S. Bosnar, T. Antonic, J. Bronic, B. Subotic, Microporous Mesoporous Mater., 76 (2004) 157. M. Chiu, B. Slater, J.D. Gale, Angew. Chem., Int. Ed. 44 (2005) 1213. R.M. Barrer, E.A.D. White, J. Chem. Soc., (1951) 1267. R.M. Barrer, E.A.D. White, J. Chem. Soc., (1952) 1561. P. Norby, H. Fjellv~g, Zeolites., 12 (1992) 898. D-C. Ling, X-W. Xu, F. Zuo, Y.-C. Long, Microporous Mesoporous Mater., 70 (2004) 63. R.M. Barrer, Pure Appl. Chem., 51 (1979) 1091. R.M. Barrer, Hydrothermal Chemistry of Zeolites, Academic Press, London, 1982. P. Norby, A. Norlund Christensen, I.G.K. Andersen, Acta Chem. Scan., A40 (1986) 500. P. Norby, Zeolites, 10 (1990) 193. E. Krough Andersen, G. Ploug-Sorensen, Z. Kristallog., 176 (1986) 67. A.M. Healey, G.M. Johnson, M.T. Weller, Microporous Mesoporous Mater., 37 (2000) 153. A.M. Healey, P.F. Henry, G.M. Johnson, M.T. Weller, M. Webster, A.J. Genge, Microporous Mesoporous Mater., 37 (2000) 165. S. Hansen, L. Fglth, Zeolites, 2 (1982) 162. K.G. Ragimov, M.I. Chiragov, N.M. Mustafaev, K.S. Mamedov, Sov. Phys. Dokl., 23 (1978) 697. C.R.A. Catlow, Modelling of Structure and Reactivity in Zeolites, Elsevier Academic Press Limited, London, 1992. C.R.A. Catlow, R.A. van Santen, B. Smit, Computer Modelling of Microporous Materials, Elsevier Academic Press Limited, London, 2004. J.D. Gale, J. Chem. Soc. Faraday Trans., 93 (1997) 629. B.G. Dick, A.W. Overhauser, Phys. Rev., 112 (1958) 90. D.W. Lewis, C.R.A. Catlow, Faraday Discuss., 106 (1997) 451. N. Almora-Barrios, A. Gomez, A.R. Ruiz-Salvador, M. Mistry, D.W. Lewis, Chem. Commun., (2001) 531. D.W. Lewis, A.R. Ruiz-Salvador, N. Almora-Barrios, A. Gomez, M. Mistry, Mol. Simul., 28 (2002) 649. N.H. de Leeuw, S.C. Parker, Phys. Rev. B., 58 (1998). T.R. Forester, W. Smith, J. Mol. Graphics, 14 (1996) 136. S.M. Auerbach, Inter. Rev. Phys. Chem., 19 (2000) 155. M.J. Mora-Fonz, C.R.A. Catlow, D.W. Lewis, Angew. Chem., Int. Ed. (submitted, 2004).