Predicting the Templating Ability of Organic Additives for the Synthesis

J. Phys. Chem. 1995,99, 11194- 11202

11194

Predicting the Templating Ability of Organic Additives for the Synthesis of Microporous Materials D. W. Lewis2 C. M. Freeman,' and C. R. A. Catlow*,' Davy Faraday Research Laboratory, The Royal Institution of Great Britain, 21 Albemarle St., London WlX 4BS, UK,and Biosym Technologies Inc., 9685 Scranton Road, San Diego, Califomia 92121 Received: March 9, 1995; In Final Form: April 28, 1995@

A combination of several computer modeling techniques has been applied to investigate the ability of organic molecules to template microporous materials. We show that the efficacy of a template can be rationalized in terms of the energetics of the host-template interactions. The calculated geometries of the templatebmework combinations are in excellent agreement with the experimental structural data. The procedures used can successfully identify optimum templates for a given-host and have applications in the synthesis of new microporous materials.

1. Introduction Microporous solids are industrially important materials used for catalysis, ion exchange, and gas separation. Synthesis of such materials usually involves crystallization from a gel medium under hydrothermal conditions. These gels are multicomponent systems containing typically aluminosilicate species. The use of organic bases during the synthesis of clathrasils, zeolites, and other molecular sieves is widespread, a particular motivation being their apparent templating effect.' However, the complex and varied conditions in such syntheses have given rise to several theories of the role of such organic molecules in the crystallization process.2 The early work of B m e r and Denny3 demonstrated that the addition of an organic hydroxide (in this case tetramethylammonium hydroxide) as the base component of the synthesis gel produced new materials. The use of such additives also promotes the rate of crystallization during synthesis;' moreover, the rate can be optimized by modification of the t e m ~ l a t e . ~ Microporous materials are metastable, and as a consequence the relative importance of the thermodynamics and kinetics of formation is crucial. Ostwald's law of successive transformation suggests that (thermodynamicallymore stable) dense phases will result with increasing crystallization time. Although this appears straightforward for reaction mixtures involving only silica, alumina, and inorganic cations the additional stabilization of structures by template molecules can have a significant effect.'%5 For example an aluminosilicate solution shows the following transformation sequence: amorphous =;$ faujasite

ZSM-4'

while the addition of tetramethylammonium to the solution results in amorphous

faujasite

-

gismondine

-

Na-P5

In many systems the absence of a template in the crystallizing gel will lead to the formation of amorphous or dense materials under the same conditions,' highlighting the importance of the template.

1- The Royal Institution of Great Britain. Biosym Technologies Inc. @

Abstract published in Advance ACS Abstracts, June 15, 1995.

0022-3654/95/2099- 11194$09.00/0

In this paper we will be concemed with the structure-directing and templating action of these additives. We also investigate the extent to which we can quantify this effect and thus predict the effectiveness of an additive for the formation of a particular framework. We shall refer to these additives as templates for the remainder of this paper. The degree to which templates are crucial in the synthesis of a particular framework varies.6 Certain templates (such as methanol and ethylenediamine) can be considered as void-filling species and do not significantly contribute to the structure formation.2 We may also find that a particular framework is formed by several different template molecules, a process which might be more appropriately described as structure directing rather than true templating. However, in general, all the templates suitable for a particular framework will possess similar properties-size, shape, basicity, etc.-which direct the gel chemistry toward the formation of particular structural motifs.2 This has been demonstrated by the work of Gies and Marler' for both clathrasils and zeolites. In certain (albeit rare) cases a particular framework can be synthesised only with a limited number of templates or, indeed, a single template, and thus we can consider that in these cases the organic base is behaving as a true template. A prime example is the zeolite ZSM-188which until the recent work of Schmitt and Kennedy9 had been synthesised only in the presence of the triquaternary amine C I ~ H ~ & ~These + . authors used computer graphics and modeling to select tris(2-trimethylammonioethy1)amine as a suitable template for ZSM- 18 and have successfully demonstrated that this molecule is effective in the synthesis of ZSM-1K9 This work has demonstrated how modeling can be useful in synthesis design. However, it must also be noted that gel chemistry is also crucial, and that the templating action will not be effective unless the correct gel environment is achieved.2 The effect of hydration of both the framework fragments'O and of the templates is of particular importance.' Structural details of the location of templates within microporous materials is limited. Such details are available from the single crystal X-ray studies of tetrapropylammoniumin MFI,"-I3 of templates in LEV-type structure^'^ and also of crown ethers in EMT.I5 There are, however, several experimental studies of organic molecules within these materials, primarily of course those of catalytic interest. Powder neutron diffraction,I8 and NMR'9-2' have all been applied in locating absorbed molecules in microporous materials. 0 1995 American Chemical Society

Synthesis of Microporous Materials Although in its infancy, the synthesis of new framework structures by rational design and selection of template molecules is becoming increasingly possible. For example, both SSZ2422and DAF-lZ3were synthesized by careful preselection of a template which would result in channels of the desired size. Furthermore, as already noted, Schmitt and Kennedy used molecular modeling techniques in the design and selection of a new template for synthesising ZSM-18.9 However, we are still some way from understanding the mechanism of zeolite crystallization and in particular framework formation under the influence of structure-directing templates. This paper will highlight the role which computer modeling techniques can play in advancing our understanding in this field and in guiding the synthesis of novel microporous materials. Computer modeling has been shown to be a valuable tool for investigating the structure and reactivity of zeolite struct u r e ~ .As ~ ~with experimental studies, work has concentrated on the and d i f f ~ s i o n of ~ ~molecules -~~ within the pores. These studies demonstrate the accuracy that simulations can achieve in reproducing experimental data. Recent work has highlighted the role of “inverse shape selectivity”, Le., the attractive interaction between sorbate and sorbant, in hydrocarbon sorption and prod~ction.~’The crystallization rate in the presence of different templates has been correlated with calculated interaction energies of the template with the framework.32 The geometry and interactions of template molecules within the Z S M J framework have also been i n ~ e s t i g a t e d , ~ ~ demonstrating that differences in crystal growth are a consequence of differences in the interaction between different geometries of the templates within the framework and that modeling techniques can correctly calculate template geometries. In this paper we shall utilize and expand on the techniques used previously to determine the location of the template and the interactions of the template molecules and the host framework. In addition, we shall determine the mode of template packing. We present results of a study of the interaction between templates and framework structures. We will demonstrate that the nonbonding interactions of the template and the framework are crucial for successful templating. Furthermore, we will show that it is possible to select templates for a given framework on the basis of these calculations. 2. Methodology

Our approach is based on a combined molecular dynamics, Monte Carlo, and energy minimization technique25 to probe the nonbonded interactions between guest molecules and zeolite frameworks. A molecular dynamics (MD) trajectory for the template molecule in the gas phase is used to generate a library of sorbate conformations, each of which is then inserted randomly into the framework using a Monte Carlo procedure. Only lowenergy configurations are retained for subsequent analysis. The MD ensures adequate sampling of the conformational space of the guest molecule. This sampling is most important when studying the adsorption of small flexible molecules within a host. However, it is apparent from noting the dimensions of templates and of the pores which they form, that the templates are not significantly perturbed from their low energy conformers; that is, we do not need to consider, for example, the bent (higher energy) conformers of linear molecules since they will not fit within the dimensions of the frameworks considered. Energy minimization is then applied to these crudely docked structures to yield representative low-energy binding sites for the molecule within the host structure. In many cases, particularly where there is a true templating effect, there is a very close match

J. Phys. Chem., Vol. 99,No. 28, 1995 11195 between the framework and the template and the probability of docking the template using a Monte Carlo sampling method is small. Thus, we must either allow an extended sampling of the framework space, which can be computationally expensive, or simply manually dock the template molecule. Thus for the , case of ZSM-18 and the triquatemary amine C I ~ H & ~ + we have crudely docked the molecule in the framework manually prior to energy minimization. It is possible when templates are docked into a pore structure with only restricted voids that we locate only a local minimum. Simulated annealing34was therefore used with certain structures, such as triquatESM-18 and the bulkier cyclic templates, as an additional step before energy minimization. The annealing procedure allows the system to search dynamically the energy surface to find the global minimum. The procedure involves first heating the system followed by a slow cooling stage before using conventional energy minimization to find the final optimised structure. Two important constraints are placed upon these calculations. First the calculation is carried out on a finite, fixed geometry portion of framework (generally 60 A x 60 8, x 60 A). Therefore, the docking procedure is confined to the central portions of the framework cluster which would be expected to be sufficiently bulklike. After synthesis, the template molecules are removed by calcination. Since this results in only a minimal change in the crystal structure of the framework we do not consider that fixing the framework geometry will have a significant effect on our calculations. Furthermore, the templating action will be more effective if the inclusion of the organic molecules does not perturb the framework structure to any significant degree. The second limitation is the omission of electrostatic interactions in the present calculations, which is clearly a simplifying assumption. However, we have restricted our work to siliceous end members of the zeolite structures, and therefore the electrostatic field that the template molecule experiences is largely uniform. Thus, the omission of Coulombic interactions can be expected to have a comparatively small effect on the calculated structures and relative binding energies of different organic species in a framework structure. Indeed, we shall demonstrate that although there are significant differences in the numerical value of the binding energies on the inclusion of electrostatics, the trends are unchanged as are the geometries of the templates. Furthermore, recent work has demonstrated that there is little electrostatic similarity between different templates which form the same framework and that size and shape is the dominant effect.35The increasing availability of accurate interatomic potentials will make more detailed simulations possible in the future. However, such simulations would demand substantial computational effort at present as inclusion of charges would necessarily require not only a complete treatment of the long-range Coulombic interactions but also (in the case of the cationic templates) a large sampling of counterions and defect models about each potential binding site. This is particularly problematic if the template charge is compensated by the framework, since we have to determine any preferred substitution sites (if any) in the framework. The above procedures allow us to determine low-energy binding sites for a particular framework-template combination. The energies obtained are a measure of the “match” between the sorbate geometry and the topology of the host framework. We can therefore calculate an interaction energy @inter) which includes any effect of conformational change when the template molecule is placed within the framework:

Lewis et al.

11196 J. Phys. Chem., Vol. 99, No. 28, 1995

TABLE 1: Nonbonding Parameters from the cft9Lczeo Force Field“ where Enonbondingis the nonbonded interactions between template and host and Efreeis the minimum energy of the template in the gas phase. The methodology is implemented in the Catalysis36 module of the Insight11 software available from Biosym Technologies Inc. The Discover3’ program was used for the energy minimizations of the trial structures, employing the cff9 1-czeo force field,38339throughout, the nonbonding parameters being listed in Table l. We note that other workers27 have found that the nonbonding terms of this force field, derived from ab initio calculations rather than being fitted to experimental data, underestimate the binding energy of organic molecules within zeolites. However, we shall consider only trends and relative energies within this study. Furthermore, we have carried out limited calculations with the cvff force noted by other workers to give results in good agreement with experimental heats of absorptionz7 These results indicate that the minimized hosvguest structure is almost identical and that the difference in energy between the two force fields is nearly constant (Figure 1). In this context we also note that it is difficult to relate the absolute values to experiment as only a limited amount of thermodynamic data on the template assisted synthesis is available!’ However, we note that with careful parameterization it is possible to obtain accurate heats of sorption in agreement with experiment!z

type sp3 c sp2aromatic C sp3N in amines sp2N in 5- or 6-membered ring hydrogenated sp2N sp2nitrogen in aromatic amines sp3nitrogen in protonated amines sp3 oxygen aromatic sp20 in 5 membered ring sp3 s H bonded to C, H H bonded to N, 0 Si in zeolite 0 in zeolite

r

e

4.0100 4.0100 4.0700 3.5700 4.0700 4.0700 3.2620 3.5350 3.5350 4.0270 2.9950 1.0980

0.0540 0.0640 0.0650

o.Ooo1

3.4506

0.0410

0.1340 0.0650 0.0650 0.2400 0.1090 0.0710 0.0200

0.0130 0.0000 0.1622

‘Bonding parameters are given in ref 38. Nonbonding energy between atoms i and j separated by distance rlJ is given by E = ~,,[2(r*,~/r~~)~ - 3(r*,J/r,J)6], where r*zJ= [(r: + r,9)/2]1/6and = 2.SE,E,(r,3(r,3)/(n6r,9 NmHIrCqlrnLla 4

5

6

7

6

to -- -2

3. Results and Discussion 3.1. Energetics of Host-Template Interactions. 3.1.i. Nonbonding Interactions between Templates and Host. Can we rationalize templating ability in terms of energetics? In attempting to answer this question, we examine first the nonbonded interactions between the template molecule and the zeolite host. We have used the docking technique to obtain minimum energy sites for a large number of experimentally noted template/zeolite combinations listed in Table 2. We find that there is a reasonably good correlation between the nonbonding energy and the number of non-hydrogen atoms in the template (Figure 2) which implies that non-hydrogen atom environments are at least approximately equivalent for known template-framework pairs. However, we should note that there is better correlation if we consider trends amongst the different template types. We can see from closer examination of Figure 2 that the correlation among the diamines and tetraalkylammonium ion homologous series is excellent. This also applies to the bis-quaternary amines. Thus, we observe that the maximization of favorable hosvguest interactions is a characteristic of successful templates in considering chemically and structurally similar system. As a consequence of the definition of the molecular mechanics force field, we should note that it is not strictly correct to compare the energies of molecules comprising different atoms and atom types. However, the comparison of intermolecular energies provides a convenient and reasonable means of quantifying host-guest interactions. We see that for a homologous series the comparison of the energy of different members is valid, since we are only increasing-and not changing-the atom types. However, as we have noted above, the trend in energies when we consider all the template types, in which there are different atom types, is not as good. We also note that, not surprisingly, it is not acceptable to use the number of nonhydrogen atoms as a measure of template “size”. For example cyclopentane is significantly different in size and shape to n-pentane. We have therefore analyzed the energetics with respect to molecular volume (Figure 3). The volume is defined

1-10

Figure 1. Comparison of force fields. Interaction energies are given for the series of n-diamines (diaminoethane to diaminhexane) in silicalite. The differences in energy between the two force fields are given on the right-hand scale. TABLE 2: Framework Types and Templates Investigated (IZA Structure Codes Given for Framework Types in Parentheses) framework types template^^^^^^^^^ primary diamines (H2N(CH2),NH2, x = 2-6) ZSM-5 (MFI) bis-quaternary amines ((CH3)3N+(CHz),+NZSM- 11 (MEL) (CH3)3, x = 3-10) ZSM-12 (MTW) tetraalkylammonium ions ([CH3(CH2),]4Nf, x = 0-3) ZSM-18 (MEI) trisquatemary amine ClsH3&3+ cyclic amines and other heterocycles (5-, 6-, ZSM-23 (MTT) and 7-membered rings) zeolite$ (BEA) dialkylamines (e.g.,CH~CH~CH~NHCHZCHzCH3) faujasite (FAU) trialkylamines (e.g.,tripropylamine) 1.3-diaminopropane mordenite (MOR) Dodecasil 1H (DOH) 1,4, 8, 11-tetraazaundecane Dodecasil3C (MTN) 1, 5 , 8, 12-tetraazadodecane ZSM-22 (TON) 1, 5, 9, 11-tetraazatridecane 1-adamanylamine EU-1 (EUO) chabacite (CHA) piperizines zeolite sigma (DDR) NU-3 (LEV) as that of the overlapping van der Waals radii of the at0ms.4~ The same trends identified above are again evident. However, we now note some improved correlation for certain templates, particularly for the cyclic molecules. Therefore, we conclude that molecular volume provides a good measure of template “size” with respect to the efficacy of the template to form a framework. It is of course simply a quantitative expression of

Synthesis of Microporous Materials

J. Phys. Chem., Vol. 99, No. 28, 1995 11197 TABLE 3: Templates in Dodecasil 3C

0

V

tetraalkylamnium cations bisquatemary amines u,odiamines saturatedcyclic unsaturatedcyclic

-0A

4

v x

3s?

cyclopentane (C5H 10.) tetrahydrofuran (C4H8O) pyrrolidine (C4HgNH) tetrahydrothiophene C (&&$) cyclopentyl amine (CsHs-NHz) furan thiophene quinucilidine ( C ~ H I ~ N ) tetramethylammoniumu triprop ylammonia" ethyl, tripropylammonium"

-80

0

\

6

0

wz -120

5 5 5 5 6 5 5 10 5 4 6

73.0 65.1 82.8 78.0 82.5 54.1 66.0 103.5 79.8 62.7 -95.7

-105.3 -94.4 -94.0 -95.6 -57.9 -60.2 -57.9 -84.3 -75.8 -53.9 -49.2

Possible templates during synthesis using trimethonium.

-160

l

~

4

l

6

8

~

10

l

12

'

l

14

~

16

l

18

~

l

'

l

~

l

'

~

u)

Non-Ha t o m

Figure 2. Interaction energy of experimental frameworkhemplate combinations as a function of non-hydrogen atoms. The straight line is the best fit through all data points. The results for the tetraalkylammonium salts are also highlighted with a line joining the data points

A B Figure 4. Comparison of calculated (dark) and experimental" (light)

O l A

bis-quatemaryamines

-40

1

50

.

1

'

100

lh

200

250

Molecular volume / A3

Figure 3. Interaction energy of experimental frameworkhemplate combinations as a function of molecular volume. The straight line is the best fit through all data points. The results for the tetraalkylammonium salts are also highlighted with a line joining the data points. the closeness of the fit between the host and the template which is evident from experimental data. Gies and Marler concluded in their work that the chemical character of templates with similar shapes had no influence on templating ability.' Our results for monocycles as templates for Dodecasil 3C support this conclusion (Table 3). The interaction energies of tetrahydrofuran, tetrahydrothiophene, cyclopentane, and pyrrolidine with the host are very similar while those of furan and thiopene and that of cyclopentyl amine (all templates for Dodecasil 3C) are significantly different-a consequence of the similar nonbonding characteristics of the atoms types (as defined by the force field) in the molecules. However, it may be that the methods used to derive the nonbonding parameters are not sensitive enough to distinguish

positions of TPA in ZSM-5. (A) Isolated TPA molecule. (B) Position when two TPA molecules are used in the calculation.

between similar atom species. Nonetheless, it is an important confirmation of this general conclusion and one which can be utilized in future template selection and design. The effect may be particularly important when considering the solubility of templates, as these similarly shaped molecules can have very different physical properties. The hydration of framework building units and the interaction of these hydrated species with hydrated template species is another question which should be considered, especially as this difficult field of study is becoming accessible to modelling techniques. lo 3.I . ii. Calculated Structural Properties of the TemplateHost Interactions. To determine if our methodology is correctly locating template positions, we have compared the calculated minimum-energy template location with the limited experiment structural data available. On comparing our calculations for TPA in silicalite with the single crystal X-ray structure' (Figure 4a), we note that the calculated position for the N atom is offset from the experimental position by approximately 0.4 A. We attribute this small displacement mainly to the omission of other template molecules, and we shall demonstrate later how the inclusion of the nearest-neighbor template molecules greatly improves the agreement with the experimentally determined position. We also find that our calculated geometries are in excellent agreement with the experimental geometry of 1-aminoadamantane and N-methylquinuclidinium in NU-3 l4 and our minimized structure is close to that proposed8 for C I * H ~ ~ N ~ ~ + in ZSM-18 (Figure 5). 3.I . iii. Eflect of Neglecting Coulombic Interactions. A major approximation made in this work has been the neglect of Coulombic interactions. We shall consider how significant are the contribution of Coulombic interactions in determining templating action and the extent to which this simplification can be justified. The calculations on the tetraalkylammonium cations in the ZSM-5 framework were repeated with the

Lewis et al.

11198 J. Phys. Chem., Vol. 99, No. 28, I995 Relut ive Energy (Charges) (&\/mol) -1200

-1400

-1000

-800

-600

-400

-200

0

Figure 5. Calculated position of the triquatemary amine template in ZSM-18. Two views are given. TABLE 4: Interaction Energies of Tetraalkylammonium Cations in Siliceous ZSM-5" EintJkJmol-' re1 EintJkJ mol-' charges

X

Figure 7. Interaction energies of tetraalkylammonium cations in

template

charges

no charges

no charges

TMA TEA TPA TBA

-8004.0 -8733.4 -9007.7 -9378.3

-5 1.7

0

0.0

-92. 1 -133.9 -165.5

-689 -921 - 1260

-40.3 -82.1 - 113.7

siliceous ZSM-5. Plot of the relative interaction energy for both charged and uncharged systems plotted against each other. Energy is defined as in Table 4.

a The differences are given relative to the TMA interaction energy, Le., the relative interaction energy is defined as E(TxA)-E(TMA) where TxA is TMA, TEA, TPA, and TBA.

-7m

'

"

-80 -100

. -120

.95w 1

1-180

Figure 6. Interaction energies of tetraalkylammonium cations in siliceous ZSM-5. The interaction energy is plotted for both charged (left axis) and uncharged systems.

inclusion of Coulombic interactions during the final minimization stage, although such interactions were still ignored in the Monte Carlo docking stage of the procedure. While using a finite framework cluster, we must ensure that the cluster size and more importantly the location of the template within that cluster do not affect the results. It is clear that a template which is not centrally located within a finite cluster will not experience the same Madelung field as in the infinite crystal. Inclusion of charges therefore adds considerably to the amount of work required in constraining the docking procedure to the central portion of the framework clusters which now have to be large enough to ensure satisfactory summation of the long-range Coulombic interactions. In this context we note again that one of the aims of this work was to allow the rapid evaluation of the energetics and structures of a range of templates and this procedure requires not only considerably longer calculations but also does not lend itself to an automated procedure. For the calcylations including electrostatics it was necessary to employ 80 A cubic clusters, terminated with hydroxyl groups with the docking of the template restricted to the central unit cell. The interaction energies obtained with and without charges are given in Table 4 and illustrated in Figures 6 and 7. From Figure 6 it is evident that the inclusion of electrostatics does not alter the overall trend. Furthermore, the final geometries

Figure 8. Comparison of the most stable configuration of TPA in siliceous ZSM-5 with (darker lines) and without (lighter lines) charge interactions.

obtained are very similar regardless of the neglect or inclusion of charges as illustrated in Figure 8. Thus we conclude that the neglect of Coulombic interactions affects only the total energy and not the nonbonding component which appears, for this particular set of templates in ZSM-5, to be confirmed as the dominant factor in determining templating ability. However, although this result vindicates the approximation used in this work, we must take care when considering more polar species, especially those which have a dipole moment; here we have considered molecular ions which are symmetrical in their charge distribution. Furthermore, there will be an obvious need to consider electrostatic terms if frameworks with an overall charge are considered. The inclusion of such terms will require much more careful selection of the framework

Synthesis of Microporous Materials cluster in terms both of size and the space into which the templates can be docked. Many of these problems would be removed if the framework were considered as a periodic system, but this would require the framework structure to provide full charge compensation for the organic ions. Furthermore, a scheme for determining any preferential siting or ordering of aluminium or other heteroatom would have to be determined. Indeed, given that this information is not available from experimenal data, much theoretical work has been carried out to determine the siting of Al and other metals in a variety of zeolites which have generally concluded that several sites will be o c ~ u p i e d . ~ - ~We ’ have therefore restricted the present study to siliceous frameworks, and it is clear that calculations involving aluminosilicate frameworks would require extensive further work. We further note that templating effects appear to be more important in the synthesis of siliceous s t r u ~ t u r e s . ~ * ~ ~ ~ We must also consider the effect of template-template charge interactions, particularly if the templates are charged. We note that in such species such as quaternary amines that the charge is distributed primarily onto the hydrogens of the alkyl chains, which will lead to repulsion between neighboring templates. When charges are included on the bis-quaternary amines in our calculations in unidimensional systems (where we expect this effect to be most pronounced), we note that the templatetemplate replusions affect the packing arrangment of the templates (discussed below), reducing the effective template concentration per unit cell. However, these interactions do not affect the relative stability of the templates in the zeolite. But, careful consideration must be made of such interactions in onedimensional channels, and this effect will be considered in future work along with template/framework charge interactions. We are confident therefore that the neglect of electrostatic terms will not affect our overall conclusions, enabling us to use a method which is considerably quicker and simpler to implement than if such terms had to be considered. We again emphasise the advantage of the current methodology in rapidly scanning a large number of template/framework combinations. In addition, these results support the view which has been proposed by many experimental studies’ and reviews2.6 that templating effects are primarily governed by steric and van der Waals interactions. Nevertheless, further studies are required to characterize fully the effect of aluminium (and other heteroatoms) on the templating action and framework stability. 3.2. Predicting Templating Ability. For the technique to be able to be used in a predictive way, to assist with synthesis design, it must be able to select correctly a suitable template for a given framework. We have tested this by studying in more detail two series of templates. The tetraalkylammoniumcations, from tetramethylammonium (TMA) to tetrabutylammonium (TBA), have been used extensively in synthesis and are of particular interest since they can form intersecting threedimensional channel systems. The bis-quaternary amines (general formula (CH3)3N+-(CH2),-+N(CH3)3, x = 3-8 and referred to as tri-, tetra-, penta-, hexa-, hepta-, and octamethonium ions), on the other hand tend to form two-dimensional channels. These compounds are widely used and have been the subject of an extensive synthetic We have tested each of these two series by locating minimum-energy sites for each molecule within all the frameworks synthesized by that series. We shall consider the two series separately. 3.2.i. Tetraalkylammonium Cations. We find that experimental template/framework combinations (Table 5) exhibit the most favorable nonbonding interactions. Thus templates which are too small for a given framework are only weakly bound,

J. Phys. Chem., Vol. 99, No. 28, I995 11199 TABLE 5: Nonbonded Interactions Energy of Tetraakylammonium Cations in Various Siliceous Frameworks” ZSM-5

Emw (kJ mol-’)

(kT mol-’) TMA

TEA TPAb TBA (I

-51.7 -92.1 -133.9 -165.5

P

ZSM- 1 I

E,“,C,

-38.7 -73.0 -119.9 -159.5

TMA TEA

TPA TBAb

Einter

(kJ mol-I)

TMA TEA‘ TPA TBA

-43.1 -104.7 -83.4 -56.7

Template predicted as effective template are emboldened. Ex-

perimental template.

TABLE 6: Nonbonded Interactions Energy of Bis-quaternary Amines and the Siliceous Frameworks of EU-1 and ZSM-23” ~~

EU- 1

ZSM-23 -60.7

pentamethonium

- 18.6 - 125.8 - 147.4‘

hexamethonium heptamethonium octamethonium

-94.9 -126.1

- 100.7’

trimethonium tetramethonium

- 128.5’

- 10.9 -13.8 -24.8

-32.8’

Templates predicted as effective template are emboldened. Experimental template.50 while repulsive forces act upon those which are too large. The method correctly identifies TEA as the best template for zeolite /? and TBA for ZSM-11 as found experimentally. However, we find an anomaly for Z S M J with TBA being calculated as more stable than the experimental template, P A . This is, however, to be expected, as in the calculation (which we carried out on a single template molecule) the longer alkyl chains are able to fit the pores of ZSM-5 without hindrance, increasing the nonbonded interactions with the framework and thus increasing the binding energy. We shall demonstrate later how inclusion of packing effects removes this anomaly. 3.2.ii. Bis-Quatemaiy Amines. From Table 6 we see that again for this series, the calculations on single templates are reasonably successful at predicting the correct template for a given host. The pore structure of EU-1 consists of a straight channel which has side pockets on alternate sides of the channel. We find that the minimum-energy structures of all the molecules tested involve the molecule fitting into these side pockets. Thus as we increase the size of the template, they first fill the side pocket, then the channel and-for the largest molecules-bend into the channel (Figure 9). The experimental templates are those which fill the pocket and channel with the least amount of distortion of the molecule. We will see later that after including packing effects, we find that the two experimental templates have the maximum binding energy. ZSM-23 consists of a straight one-dimensional channel system. We therefore find that larger (linear) templates are generally favored over the smaller ones, as a consequence of the increase in the number of nonbonding interactions with the increase in the number of atoms in the template. However, there are exceptions. Most significantly we find that the trimethonium ion has the larger interaction energy since the main contributions to the nonbonding energy is from the amine end groups which in trimethonium are in better contact with the framework; with the larger molecules, the end groups lie parallel to the channel. It is clearly more difficult to predict correctly the templating ability for systems with one-dimensional channels from singlemolecule calculations.

11200 J. Phys. Chem., Vol. 99, No. 28, 1995

Figure 9. Calculated positions of octamethonium (top left), tetramethonium (top right), pentamethonium (bottom left), and hexamethonium (bottom right) in EU-1.

TABLE 7: Stabilization Energy of Tetraalkylammonium Template-Framework Combination When Two Adjacent Template Molecules Are Included template/framework

AEpack(kJ mol-')

TPNZSMJ TBAESM-5 TBNZSM-11 TPNZSM- 1 1

+14.9

-29.7 -18.3 -8.5

Although trimethonium is found experimentally to template the formation of Dodecasil 3C,50we have been unable to dock successfully this molecule within the framework. Although energy minima can be found, the geometry of the organic molecule is unfeasibly strained. We therefore conclude that any organic species occluded within Dodecasil 3C in this synthesis is not trimethonium but some smaller product of the decomposition of trimethonium. We note that the synthesis was carried out at the relatively high temperature of 180 "C and that trimethonium may decompose at this temperature. Possible decomposition products such as tetramethylammonium and trimethylamine will fit within the cages of Dodecasil3C (Table 3). 3.3. Packing Considerations. Until now, we have considered only isolated template molecules within the framework. However, templating and structure directing require the template molecules to self-organize and pack effectively within the framework. Therefore, we have expanded the calculations to include two (and in certain cases more) template molecules: one positioned at the lowest energy site for an isolated molecule and another at the next translationally equivalent site. We can define the packing energy as

where E1 and E, are the nonbonding energy of one and n template molecules in the host lattice, respectively. In the ZSM-5 framework we find that the butyl chains of TBA are too long to allow neighboring channel intersections to be occupied, which results in an unfavorable packing energy (Table 7). Conversely, two TPA molecules pack at adjacent sites satisfactorily, and indeed the system is stabilized by favorable nonbonding interactions between the two molecules. We find a similar effect for TBA (the experimental template) in ZSM- 11. TPA in ZSM- 11, although more stable than a single molecule, is not as favored as TBA in the same framework. Thus

Lewis et al. on including the interaction energies of the template molecules we can predict successfully the correct tetraalkylammonium template for each framework. The addition of further template molecules has also improved the geometry of the TPA with respect to the experimental structure (Figure 4b) to give an almost perfect match. The minimum energy sites in EU-1 suggest that the templates are located in the side pockets of the channels. It is reasonable therefore to assume that the templates will occupy each pocket, thus forming a close-packed array of templates which mirrors the geometry of the framework. We have therefore calculated packing energies for the most likely templates (from the isolated molecule calculations). Tetra-, penta-, and hexamethonium ions all fit within the length of the pocket and the channel (Figure 9). However, tetramethonium is too short to effectively to fill this void and the resulting packing energy (as is the single molecule energy) is not as favorable as those for penta- and hexamethonium. Octamethonium on the other hand is too long to lie perpendicular to the channel and has to bend into the channel. Thus although the packing energy appears to be favorable, we note that it is only possible to fit one template molecule per two side pockets. Thus it is unlikely that this arrangement would result in the formation of the EU-1 framework. Furthermore, it seems unlikely that the required conformation of the template could be achieved for an extended array of molecules. We therefore conclude that only pentamethonium and hexamethonium possess the ability to form EU1, in agreement with experiment. The question of template packing is thus straightforward for both the case of the tetraalkylammonium ions, where the channels involved are three-dimensional and the template molecules are sited at the intersections, and for EU-1 where the nature of the channel shape has an important effect. However, when we consider the packing of molecules within one-dimensional channels systems or systems where the template shape does not imply siting at intersections or in pockets and cavities, we must take care when considering packing effects. In these cases, the framework structure does not introduce any constraints on the siting of the template molecules. In particular we should consider how the template molecules may align with the repeat unit of the framework and how this affects the distance between templates and also the role of the solvent. Consider the templating of ZSM-23. Here we find that a single trimethonium ion has the most favorable interaction energy (Table 6). However, it is the slightly less favored heptaand octamethonium ions which are found to form the framework. Attempts at packing trimethonium ions result in unfavorable packing energies in which the amine end groups point in the same direction (Figure 10). Thus, if the molecules are to close pack such that the end groups "interlock" alternate molecules must be rotated along the axis of the channel, which reduces the nonbonding interaction of the molecules and results in an unfavorable packing energy (Table 8). Heptamethonium ions can pack with interlocking end groups as a result of the odd number of carbon atoms between the end groups and the geometry of the minimum energy binding conformation. Thus it is possible to close pack the molecules in a favorable manner (Figure 10). Although the end groups of octamethonium do not allow close packing in the lowest energy conformer (as is the case for heptamethonium), the rotation of every other molecule does not significantly reduce the interactions with the framework. Furthermore, the calculated packing arrangement results in the template molecules being commensurate with the repeat length of the framework (also found for heptamethonium).

J, Phys. Chem., Vol. 99, No. 28, 1995 11201

Synthesis of Microporous Materials

TABLE 8: Stabilization Energy of Bis-Quaternary Ammonium Templates-Framework Combination When Multiple Template Molecules Are Included framework

EU-1

template

no. of molecules

tetramethonium

2

pentamethonium

3

2

-5.4

3 octamethonium

2

-6.0 -1.6 -2.9 -9.7

trimethonium

2

+5.4

3 2 3 2

+7.0 -2.9 -2.2 -0.3

hexamethonium

2 3

ZSM-23

AEpack

(kT mol-’) -2.0 -2.6

heptamethonium octamethonium

-0.4 3 frameworks would enable us to clarify the packing arrangements in these systems. We note that it is possible to infer the number of templates per unit cell from elemental analysis data, and such results will therefore provide a good indication of the packing of the template within the channels of the zeolite. We further note that zeolites containing one-dimensional channels are often disordered in the direction of the channel, for example, ZSM-

i‘_

34.40

Figure 10. (a) Calculated packing geometries of tri-, hepta-, and octamethonium in ZSM-23. (b). The configuration of heptamethonium in ZSM-23 (hydrogen atoms are omitted for clarity). The inset shows in further detail the interlocking of the methyl groups by the rotation about the C-N bond.

However, we note that the interaction energy of octamethonium is significantly lower than heptamethonium and also lower than that for similarly sized templates. The packing energy is also only marginally favorable. Thus in this case the relationship between nonbonding energy and templating ability is not as clearcut as for the other systems considered. Although the packing calculations described for ZSM-23 adequately explain the relative templating ability of the bisquaternary amines, it is important to note that we have neglected the effect that other solvents and other molecular species may have on the packing arrangements. We have also assumed that the templates will pack as closely together as possible (as in the non-one-dimensional systems considered previously). However, we must also consider the possibility that the templates may be further spaced, either with or without the presence of other species. We find that it is possible to space the bis-quaternary amines by a single unit cell length with only a small reduction in the packing energy. Thus, care must be taken when attempting to simulate the templating ability of linear templates in one-dimensional systems. High-quality experimental structural evidence of the location of templates in such

4K5’ 4. Conclusions Our calculations have identified two complementary criteria which must be satisfied for an organic molecule to template successfully a zeolite. First, the favorable nonbonding interactions between the template and the framework must be maximized. Second, the template molecules must be able to pack efficiently within the framework. In particular we have shown that the nonbonding interactions between the template and the framework during the synthesis of microporous materials can have a major influence on the product formed. We have found that these interactions control the efficacy of the template and that the ability of the template to pack effectively in the host is also of crucial importance. We have demonstrated quantitatively the action of templating, an action previously considered from empirical observation to be simply a “good fit” between the template and the framework. Calculations can correctly predict the positions of the template, thus making the technique a powerful aide in the solution of structures in which template molecules are still occluded in the structure. We have further shown the ability of such calculations to “predict” suitable templates for a given host. Therefore when considering the synthesis of new materials we can design templates so that they do not favor the formation of previously synthesized structures. The methodologies described in this paper, although containing many simplifications, can rapidly determine the relative templating efficacy of an organic species within a large range of known frameworks in order to determine if it is likely to form a known or new structure. Although at present we have limited our work to calculations on frameworks of known structure, these results indicate that calculations on hypothetical structures will prove useful for the design and synthesis of new materials. Acknowledgment. D.W.L. is grateful to the EPSRC and Unilever Plc for supporting this work. References and Notes (1) Barrer, R. M. Hydrothermal Chemistry of Zeolites; Academic Press: London, 1982.

Lewis

11202 J. Phys. Chem., Vol. 99, No. 28, 1995 ( 2 ) Lok. B. M.: Cannan, T. R.; Messina, C. A. Zeolites 1983, 3. 282. (3) Barrer, R. M.; Denny. P. J. J. Chem. Soc. London 1961, 971. (4) h i . T.: Morinaga, N.: Ishihara, T.: Kanie. T.: Takgami, Y. J. Catal. 1983. 79. 176. ( 5 ) Dwyer, F. G.; Chu, P. J. Catal. 1979. 59. 263. (6) Davis. M. E.: Lobo. R. F. Chem. Mater. 1992. 4. 759. (7) Gies. H.: Marler. B. Zeolites 1992. 12. 42. (8) Lawton. S . L.: Rohrbaugh, W. J. Science 1990. 247. 1319. (9) Schmitt. K. D.: Kennedy. G. J. Zeolites 1994, 14. 635. (10) Catlow. C. R. A,: George. A. R.; Freeman, C. M.. submitted. (11) van Koningsveld. H.: van Bekkum. H.: Jansen. J. C. Acta Ctytallogr. B 1987. 43. 127. (12) Price. G. D.: Pluth. J. J.: Smith. J. V.: Araki. T.: Bennett. J. M. Nature 1981. 292. 818. ( 13) Chao. K.-J.: Lin. J.-C.: Wang, Y.: Lee. G. H. Zeolites 1986. 6. 35. (14) McCusker. L. B. Mater. Sci. Forum 1993. 133-136. 423. ( 15) Baerlocher. C.: McCusker. L. B.: Chiappetta. R. Microporous Mater. 1994, 2. 269. (16) McCusker. L. B. In Zeolites and Related Microporous Materials: State of the Art 1994. Proceedings of the 10th International Zeolite Association: Weitkamp. J.. Karge. H. G.. Pfeifer H.. Holderich. W., Eds.: Elsevier: Amsterdam. 1994: Vol. 84. p 341. (17) Newsam. J. M.: Yang, C. Z.: King, H. E. J.: Jones. R. H.: Xie. D. J. Phys. Chem. 1984. 88. 1635. (18) Fitch. A. N.: Jobic. H.: Renouprez. A. J. Phys. Chem. 1986, 90. 1311. ( 19) Engelhardt. G.: Michel. D. High Resolution Solid State NMR of Silicares and Zeolites: Wiley: New York. 1987. (20) Brun. T. 0.: Curtiss. L. A.: Iton, L. E.: Kleb. L. E.: Newsam. J. M.: Beyerlein. R. A,: Vaughan. D. E. W. J. Am. Chem. Soc. 1987. 109. 4118. (21) Newsam. J. M.: Silbemagel, B. G.: Mechior. M. T.:Brun. T. 0.: Trouw. F. In Inclusion Phenomena and Molecule Recognition: Atwood. J.. Ed.: Plenum Press: New York. 1990: p 325. ( 2 2 ) Zones, S. I.: Yuen. L. T.: Nakagawa. Y.: van Nordstrand. R. A.: Toto. S. D. In 9th International Zeolites Conference: Buutenvirj-Heinemann: Montreal. 1992 (23) Wright, P. A,: Jones. R. H.: Natarajan. S.: Bell. R. G.; Chen. J.: Hursthouse. M. B.: Thomas. J. M. J. Chem. Soc.. Chem. Commun. 1993. 1861. (24) Modelling of Structure and Reactivin. in Zeolites: Catlow. C. R. A,. Ed.: Academic Press: London, 1992. ( 2 5 ) Freeman. C. M.: Catlow. C. R. A,: Thomas. J. M.: Brode. S. Chem. Phys. Lett. 1991. 186. 137. (26) Pickett. S. D.: Nowak. A. K.; Thomas, J. M.: Cheetham. A. K. Mol. Sim. 1989. 2. 353. (27) Henson. N.: Cheetham. A. K.: Redondo. A,: Levine. S. M.: Newsam. J. M. In Zeolites and Related Microporous Materials: State of the Art 1994. Proceedings of the 10th International Zeolite Association: Weitkamp. J.. Karge. H. G.. Pfeifer. H.. Holderich. W.. Eds.: Elsevier: Amsterdam. 1994: Vol. 84. p 2059.

et al.

(28) Nowak, A. K.: Cheetham. A. K.: Pickett. S. D.; Ramdas. S. Mol. Sim. 1987, I , 67. (29) Titiloye. J. 0.:Parker, S. C.: Stone. F. S.: Catlow. C. R. A. J. Phvs. Chem. 1991. 95. 4038. (30)Hernandez. E.; Catlow. C. R. A. Pruc. R. Soc. London, Ser. A-Math. Phys. Sci. 1995, 448. (31) Santilli. D. S.: Hanis. T. V.: Zones. S. I. Microporous Mater. 1993. 1. 329. (32) Harris. T. V.: Zones. S. I. In Zeolites and Related Microporous Materials: State of the Art 1994. Proceedings of the 10th lntemational Zeolite Association Meeting; Weitkamp. J.. Karge. H. G.. Pfeifer, H.. Holderich, W. Eds.: Elsevier: Amsterdam, 1994: Vol. 84, p 29. (33) de Vos Burchart. E.: Jansen. J. C.: van de Graaf, B.: van Bekkum. H. Zeolites 1993. 13. 216. (34) Kirkpatrick. S.: Gelatt. C. D.: Vecchi. M. P. Science 1983. 220. 671. (35) Stevens. A. P.; Boyett. R. E.: Cox. P. A. Manuscript in preparation. (36) Catalysis. Biosym Technologies Inc: San Diego. 1993. (37) Discover. Biosym Technologies Inc: San Diego. 1993. (38) Maple. J.: Dinur. U.:Hagler. A. T. Proc. Natl. Acad. Sci. 1988. 85. 5350. 139) Hill. J.-R.: Sauer. J. J. Phys. Chem. 1994. 98. 1238. (40) Dauber-Osguthorpe, P.: Roberts. V. A,: Osguthorpe. D. J.: Wolff. J.: Genest. M.: Hagler. A. T. Proteins: Struct.. Function Genetics 1988. 4 . 21. (41) Patarin. J.: Soulard. M.: Kessler. H.: Guth. J. L.: Diot. M. Thermochim. Acta 1989. 146. 21. (42) Nowak. A. K.: Cheetham. A. K.: Pickett. S. D.: Ramdas. S. Mol. Sim. 1987. 1. 67. (43) Searchcompare. Biosym Technologies Inc: San Diego. 1993. (44)Gale. J. D.: Cheetham. A. K.: Jackson. R. A.: Catlow. C. R. A,: Thomas, J. M. Adv. Mater. 1990. 2. 487. (45) Schroder. K.-P.: Sauer. J.: Leslie. M.: Catlow. C. R. A. Zeolites 1992. 12. 20. (46) Schroder. K.-P.; Sauer. J.: Leslie. M.: Catlow. C. R. A.: Thomas. J. M. Chem. Phvs. Lett. 1992. 188. 320. (47) Lewis, D. W.; Catlow. C. R. A,: Sankar. G.: Can. S. W. J. Phys. Chem. 1995. 99. 2377. (48) Szostak. R. S. Handbook of Molecular Sieves: Van Nostrand Reinhold: New York, 1992. (49) Dodwell. G. W.: Denkewicz. R. P.: Sand. L. B. Zeolites 1985. 5. 153. (50) Casci. J. L. In 7th International Conference on Zeolites: Elsevier: Tokyo. 1986: p 215. (51) Schlenker. J. L.: Rohrbaugh. W. J.: Chu. P.: Valyocsik. E. W.: Kokotailo. G. T. Zeolites 1985. 5. 355-358. JP9506898