J. Phys. Chem. B 2005, 109, 21539-21548
21539
Structure-Directing Role of Molecules Containing Benzyl Rings in the Synthesis of a Large-Pore Aluminophosphate Molecular Sieve: An Experimental and Computational Study Luis Go´ mez-Hortigu1ela,*,†,‡ Joaquı´n Pe´ rez-Pariente,† Furio Cora` ,‡ C. Richard A. Catlow,‡ and Teresa Blasco§ Instituto de Cata´ lisis y Petroleoquı´mica, C/Marie Curie 2, 28049 Cantoblanco, Madrid, Spain, DaVy Faraday Research Laboratory, The Royal Institution of Great Britain, 21 Albemarle Street, London W1S 4BS, United Kingdom, and Instituto de Tecnologı´a Quı´mica, UPV-CSIC, AVenida de los Naranjos s/n 46022 Valencia, Spain ReceiVed: April 13, 2005; In Final Form: July 20, 2005
We describe the synthesis of AlPO-5 and SAPO-5 materials (AFI topology) using five different tertiary amines or quaternary ammonium ions containing one or two benzyl rings as structure-directing agents (SDAs). All of the molecules successfully direct the crystallization of AlPO-5; however, only the most efficient templates are able to crystallize SAPO-5. The observed differences in template efficiency can be rationalized in terms of the interaction energy between these molecules and the AFI framework. In ranking the template molecules, we notice that a well-defined molecular shape enhances the templating ability, but molecules that are too rigid are not able to adapt to the AlPO framework, yielding an inferior templating ability. Results of atomiclevel modeling show that templates with one benzyl ring self-assemble in the main AFI channel by forming dimers with the benzyl rings parallel to each other; templates with two benzyl rings assemble instead into longer chains in which the benzyl ring of one molecule faces the ring of the subsequent one. Both mono- and dibenzyl templates show a high space-filling ability in AFI. Kinetic and thermodynamic factors that might affect the structure-directing activity of the molecules are examined.
Introduction Since the discovery of microporous aluminophosphates (AlPOs) by Wilson et al. in 1982,1 the synthesis of these materials has been widely studied, yielding a diversity of structural types comparable to that of aluminosilicate-based zeolites.2 Known microporous AlPO structures include polymorphs that are common to both SiO2 and AlPO4 compositions, but also structures that have no zeolitic counterpart. These microporous oxides are synthesized through hydrothermal methods and in the presence of water-soluble organic molecules in the synthesis gels.3-10 These molecules are usually referred to as structure-directing agents (SDAs); the most commonly used are amines and quaternary ammonium ions. The role of these organic molecules has been described as a “template effect”, referring to the close relationship between the shape of the SDAs and the frameworks they direct. The template effect involves the participation of the organic molecules in the organization of the inorganic tetrahedral units into a particular topology, thus providing the initial building blocks for further crystallization of the structure. These organic molecules are encapsulated inside the zeotypic framework as it crystallizes, but they retain their chemical identity and thus stabilize a particular framework type via nonbonded interactions between organic and inorganic components. However, the nature of the relationship between the SDA and the zeolite it forms is not yet clearly understood.11 Thus, the study of the template-framework interaction is a * To whom correspondence should
[email protected]. † Instituto de Cata ´ lisis y Petroleoquı´mica. ‡ The Royal Institution of Great Britain. § Instituto de Tecnologı´a Quı´mica.
be
addressed.
E-mail:
major issue in the science of molecular sieves, as controlling this feature would enable the synthesis of new topologies as well as the control of crystal size, crystal morphology, and the location of heteroatoms, if present. The structure-directing effect is not as specific as one might expect. In fact, one molecule can direct the crystallization of several different microporous structures depending on the synthesis conditions, and one structure can be templated by different molecules.5 The ability of an organic molecule to direct the crystallization of a certain structure type depends strongly on the nonbonded interactions established between organic and inorganic components,12-15 but the structural and conformational flexibility of the molecule also has to be taken into account.9 The specificity of a template molecule to direct just one microporous structure is likely to be influenced by the internal structural degrees of freedom of the molecule. Very flexible molecules (with many soft internal degrees of freedom) are able to adopt many different conformations, so the probability for a molecule to stay in that conformation that leads to the required zeolitic phase is small; thus, the specificity between the molecule and the structure formed is expected to be small. Rigid molecules, which are stable in just one conformation, are instead expected to be more specific in directing one particular microporous structure type whose geometric structure is usually closely related to that of the conformation of the molecule. However, because of the molecular rigidity, it is quite difficult to find the proper synthesis conditions for these molecules to work in the synthesis of a microporous material. The interaction of the SDA with heteroatoms can condition the synthesis too. AlPO frameworks are known to accept a large variety of dopant types, including both redox- and acid-active
10.1021/jp0519215 CCC: $30.25 © 2005 American Chemical Society Published on Web 10/21/2005
21540 J. Phys. Chem. B, Vol. 109, No. 46, 2005
Go´mez-Hortigu¨ela et al.
Figure 1. Template molecules viewed in the minimum-energy conformations from density functional theory calculations. The single bonds, which are free to rotate, are highlighted by arrows. (1) Benzylpyrrolidine (C11H15N). (2) Benzylpiperidine (C12H17N). (3) Benzylhexamethylenimine (C13H19N). (4) Dibenzylpiperazine (C16H22N2). (5) Dibenzyldimethylammonium (C16H20N).
TABLE 1: Characterization of the Organic Molecules amine benzylpyrrolidine benzylpiperidine benzylhexamethylenimine dibenzylpiperazine dibenzyldimethylammonium chloride a
distillation conditions 382 K (16 mmHg) 390 K (17 mmHg) 418 K (30 mmHg)
C
chemical analysis (wt %)a H N
81.29 81.99 81.42 82.29 81.98 82.54 80.74 81.20 71.98 73.40
9.88 9.32 10.01 9.71 11.22 10.05 10.76 8.27 8.76 7.70
8.72 8.70 8.00 8.00 7.46 7.41 10.67 10.53 5.35 5.35
1
H NMR chemical shifts (ppm)
1.7-1.8, 2.4-2.6 (pyrrolidine), 3.7 (NCH2Ph), 7.1-7.3 (Ph) 1.3-1.6, 2.2-2.4 (piperidine), 3.43(NCH2Ph), 7.15-7.25(Ph) 1.6, 2.6 (hexamethylenimine), 3.6 (CH2N), 7.3-7.4 (Ph) 2.5 (piperazine), 3.5 (NCH2Ph), 7.2-7.3 (Ph)
Theoretical values in bold.
species.16 In the present work, we examine the incorporation of silicon, to yield silicoaluminophosphates (SAPOs). The topic of interest is whether the silicon substitution in AlPOs is affected by the nature of the structure-directing agent (SDA) used in the crystallization. This feature is of the greatest importance in catalysis, as the successful application of these materials as selective catalysts in a variety of processes relies on the catalyst acidity, which is modulated through the control of the silicon substitution mechanism. Examining the silicon environment in the same AlPO framework, but synthesized using different SDAs, might provide useful information in this respect. On this basis, our aim has been to design new and efficient organic template molecules for the synthesis of large-pore aluminophosphates (AlPOs) and silicoaluminophosphates (SAPOs). We have looked for a compromise between flexibility and rigidity in the template molecules: rigid moieties in the molecule would provide the desired specificity between the SDA and the structure type directed, and flexible structural components would make the molecules effective under a wider range of synthesis conditions. With these factors in mind, we have selected several tertiary amines and ammonium cations containing at least one aromatic ring and one methylene (N-CH2-C) bridge (see Figure 1) as SDAs for the synthesis of aluminophosphates. The rationale for this choice resides in the planar and rigid nature of the phenyl ring, which confers a well-defined shape to the SDA molecule. The presence of the methylene bridge between the nitrogen atom and the aromatic ring provides, instead, flexibility to the molecules; the single bonds in the (NCH2-C) moiety provide soft torsional relaxation modes that enable the molecules to adopt different conformations, making them more likely to work as templates in the synthesis of
microporous materials. Finally, some of the molecules also contain a nitrogenated ring, which increases not only the rigidity of the molecule but also its size. The use of elongated and large amines, with a C/N ratio greater than 11, would favor the synthesis of channel-based structures over cage-based ones. Combining the structural elements detailed above, we have synthesized five different organic molecules, namely, benzylpyrrolidine, benzylpiperidine, benzylhexamethylenimine, dibenzylpiperazine, and dibenzyldimethylammonium hydroxide (shown in Figure 1), and tried them as SDAs for the synthesis of crystalline microporous aluminophosphates. We have already studied the structure-directing ability of benzylpyrrolidine and its fluorinated derivatives in the synthesis of microporous AlPO-5 in previous work.17,18 Here, we present and compare results for the different aromatic molecules. Experimental Details The synthesis of each tertiary amine was carried out by reacting the corresponding secondary amine (pyrrolidine, piperidine, hexamethylenimine, and piperazine) (Sigma-Aldrich) with benzyl chloride (Sigma-Aldrich) in ethanol (363 K, 24 h) in the presence of potassium carbonate. Then, the products were extracted with chloroform, and the solvent was removed. Finally, the liquid amines (benzylpyrrolidine, benzylpiperidine, benzylhexamethylenimine) were purified by vacuum distillation, and the solid one, dibenzylpiperazine, was purified by recrystallization from ethanol. The purity of the amines was assessed by thin-layer chromatography (hexane/ethyl acetate as the solvent), chemical analysis, and 1H NMR spectroscopy (Table 1). In this way, benzylpyrrolidine, benzylpiperidine, benzylhexamethylenimine, and dibenzylpiperazine were obtained.
SDAs in the Synthesis of AlPO-5 and SAPO-5 Materials The synthesis of the ammonium cation was carried out by adding dimethylbenzylamine (Aldrich) to a cooled solution of benzyl chloride in ethanol, and after a few minutes, the reaction was carried out under the same conditions as before (363 K, 24 h). Then, the solvent was removed, and the resulting solid (dibenzyldimethylammonium chloride) was washed several times with diethyl ether and then filtered. The collected solid was characterized by chemical analysis (Table 1) and 13C MAS NMR spectroscopy. The chloride salt was converted into the corresponding hydroxide by ion exchange with an Amberlyst IRN78 resin (exc cap ) 4 mequiv/g, Supelco). The solution of the hydroxide obtained was then titrated with 1 N HCl (Panreac) and phenolphthalein (Aldrich). The synthesis gels were prepared with the following molar composition: 1R:P2O5:Al2O3:ySiO2:40 H2O, where R represents the corresponding organic molecule and y was 0 for AlPO compositions and 0.5 for SAPO compositions. The gels were prepared by adding the aluminum source, pseudoboehmite (Catapal Pural SB, 75.3 wt % Al2O3), to a solution of phosphoric acid (Riedel-de Haen, 85 wt %) and water and stirring for 1 h in a closed receptacle (having a hole for the stirrer). The corresponding amine was then added, and the stirring was maintained for 2 h. In the dibenzyldimethylammonium case, where the molecule was added as an aqueous solution, the amount of water included with the addition of this solution had to be subtracted from the total amount of water initially added. For the preparation of SAPO gels, the silicon source (tetraethyl orthosilicate, TEOS, Merck) was previously hydrolyzed in the presence of water and the organic molecule (one-half of the necessary amounts) in a different vessel until all of the ethanol produced by the hydrolysis of TEOS was evaporated; then, this solution was added to the synthesis gels in the last step (after the addition of the remaining organic reactant) and stirred for 2 more hours. The pH of the resulting gels was between 3.2 and 4.3. The gels were introduced into 60-mL Teflon-lined stainless steel autoclaves and heated statically at 423 K for 1-3 days. The resulting solids were separated by filtration; washed with acetone, ethanol, and water; and dried at 333 K overnight. The solid products were characterized by XRD (Seifert XRD 3000P diffractometer, Cu KR radiation), thermal analysis (Perkin-Elmer TGA7 instrument, heating rate ) 10 °C/min, air flow ) 30 mL/min), chemical analysis (Perkin-Elmer 2400 CHN analyzer; ICP Plasma for silicon), and SEM/EDS (JEOL JM6400 instrument operating at 20 kV). Solid-state nuclear magnetic resonance (NMR) spectra were recorded with a Bruker AV 400 WB spectrometer, using a BL7 probe for 29Si and the 1H to 31P cross-polarization/magic-angle spinning (CP/MAS) experiment and a BL4 probe for 31P and 27Al. 29Si spectra were acquired using pulses of 3.3 µs to flip the magnetization by 3π/8 rad and a recycle delay of 240 s, while the samples were span at the magic angle at a rate of 5-5.5 kHz. The 27Al spectra were measured using π/12 rad pulses of 1 µs and delays of 1 s between two consecutive pulses. For 31P, π/2 rad pulses of 4.25 µs and recycle delays of 80 s were used. Both 31P and 27Al spectra were recorded while the samples were spun at c.a. 11 kHz. The 1H to 31P CP/MAS experiment was carried out using a π/2 rad pulse for proton of 5 µs, a contact time of 1 ms, and a recycle delay of 3 s, with the sample spinning at 7 kHz. Computational Details A computational study was carried out to complement the experimental work and rationalize the ability of the aromatic
J. Phys. Chem. B, Vol. 109, No. 46, 2005 21541 molecules to template the synthesis of the AFI structure. The same protocol we previously developed for the benzylpyrrolidine SDA18 was used to calculate locations and interaction energies of the five SDA molecules discussed here. The molecular structures and the interaction energies of the SDAs with the framework were described with the CVFF force field;19 van der Waals and electrostatic interactions were explicitly included in all calculations. The geometry of the AFI structure was optimized with the GULP code20 and the potential from Gale and Henson,21 and it remained fixed during the calculations. The atomic charges for the neutral template molecules (benzylpyrrolidine, benzylpiperidine, benzylhexamethylenimine, and dibenzylpiperazine) were calculated by the charge-equilibration method, and the charges of the framework atoms were fixed to -1.2, 1.4, and 3.4 for oxygen, aluminum, and phosphorus, respectively. For the dibenzyldimethylammonium studies, the net molecular charge of +1 had to be compensated by the framework; in the true solid, this would be achieved by the presence of defects in the framework, such as Si/P replacements in SAPOs, cation vacancies, or coordinated hydroxyl ions. The nature of these defects is not uniquely defined and cannot yet be resolved experimentally; because the calculated interaction energy of the SDA with the framework is affected by the type, concentration, and configuration of the charge-balancing defects, this problem does not have a unique computational solution. To make the calculations feasible and unbiased, we have employed a charge-balance model that makes reference to the “uniform charge background” method,22,23 which is widely employed in solid-state science to study charged defects whose countercharge in the solid is ill-defined. In this model, charged defects are compensated by a uniform distribution of charge across the whole structure. However, because the charge balance for the positive SDAs in SAPO materials is likely to be provided by Si/P replacements, we have used a modified version of the uniform charge background model, in which the negative charge of the framework is equally divided among all P sites; this is achieved by decreasing the charge of each framework P ion to 3.3167, until a charge-neutral structure is obtained. Other studies of template locations in zeotypes have neglected the net charge on the template and the electrostatic interaction between molecule and framework; under this assumption, the calculated interaction energies for amines and ammonium ions are directly comparable. This is not the case when electrostatic terms are included. In this case, results are dependent on the net charge of the SDA and the method chosen to charge-balance it in the framework. The interaction energies between the SDA and the framework were calculated by subtracting the energy of the isolated molecules from the energy of the interacting system (framework + SDA molecules). Because of the different methodology used for the calculations of the dibenzyldimethylammonium cation, the interaction energy of this molecule will not be comparable to those of the other molecules. The arrangement of benzylpyrrolidine inside the AFI channel has already been studied computationally by us in an earlier work.18 That study showed that the SDA molecules form stable dimers in which the benzyl groups of the two molecules stack parallel each other. The computational work is now extended to the other four SDAs described earlier. For each SDA, the computational protocol consisted of a succession of the following steps: (1) Monte Carlo docking of one SDA molecule in the AFI channel was performed. (2) Monte Carlo docking of a second molecule was then performed, followed by molecular dynamics (NVT) and energy
21542 J. Phys. Chem. B, Vol. 109, No. 46, 2005
Go´mez-Hortigu¨ela et al.
Figure 2. XRD patterns of AlPO-5 samples (heated for 3 days) synthesized using benzylpyrrolidine (A-bnpyrr-3), benzylhexamethylenimine (A-bnhexam-3), dibenzyldimethylammonium (A-dbdm-3), and dibenzylpiperazine (A-dbnpipz-3) as templates. The pattern of the sample synthesized with benzylpiperidine is the same as that of sample A-bnhexm-3.
minimization operations. Dimer formation is sometimes difficult to achieve in statistical Monte Carlo docking; we therefore also employed preformed SDA dimers as building units for the docking operation. The final energies obtained using mono- and dimeric SDAs were compared, and the most favorable was employed. This choice ensured that pairing through the benzyl rings, if stable, was achieved. (3) Subsequent molecules (or dimers) were then stacked in the AFI channel, without imposing periodic boundary conditions on the solution. This step was studied with a series of molecular dynamics (MD) and energy minimization calculations and yielded the preferred density of SDA molecules in the AFI channel, which we also refer to as the “theoretical” density. (4) The template-framework interaction energy was studied for a model of the solid based on periodic boundary conditions. A density of SDA molecules as close as possible to that determined at step 3 above was used, and molecules were allowed to relax and locate in the most stable arrangement by simulated annealing. Molecular dynamics simulations of the SDA molecules in vacuum were carried out using an NVT ensemble at 423 K (synthesis temperature), to assess their molecular flexibility. Finally, the mobility of SDA molecules within the AFI structure was studied by running 100 ps of MD simulations of the molecules loaded inside the structure (under the same conditions as in step 4) at 423 K. Results Henceforth, the samples will be denoted as bnpyrr (benzylpyrrolidine), bnpiper (benzylpiperidine), bnhexam (benzyl-
TABLE 2: Elemental Analysis of the Solid Samples sample
C
H
N
C/Na
organicb (wt %)
A-bnpyrrol-3 A-bnpiper-3 A-bnhexam-3 A-dbnpipz-3 A-dbdm-3 S-bnpyrrol-3 S-dbdm-3
7.57 9.58 9.70 22.07 9.20 16.83 9.55
1.21 1.11 1.15 2.17 2.31 -
0.71 0.91 0.87 2.81 0.71 1.69 0.67
12.6 (11) 12.2 (12) 13.0 (13) 9.1 (9) 15.0 (16) 11.6 (11) 16.5 (16)
9.2 (1.0) 11.7 (1.2) 11.8 (1.1) (0.9)c 10.9 (0.9) 11.2 (0.9)
a Theoretical values in parentheses. b Molecules per unit cell in parentheses. c Calculated from TGA analysis.
hexamethylenimine), dbnpipz (dibenzylpiperazine), and dbdm (dibenzyldimethylammonim) referring to the organic molecule used as the SDA in the synthesis, preceded by A- (for AlPO materials) or S- (for SAPO materials) and followed by -1 (heated for 1 day) or -3 (heated for 3 days). AlPO. All organic molecules studied here directed the crystallization of AlPO-5, a large-pore 12-membered ring material, without notable differences, although additional weak reflections of the AlPO-H3 structure, with the APC topology, appear in the diffraction pattern of the samples synthesized with dibenzyl molecules (Figure 2). The organic molecules cannot be accommodated within the pores of the APC structure, with an opening of 8 MR, and therefore, all of the SDA molecules in the solid must be incorporated into the AFI structure. The chemical analysis of the crystalline products (Table 2) suggests that the template molecules retain their chemical integrity inside the AlPO-5 channels, as the C/N ratios are very similar to those of the corresponding free molecules. The pore
SDAs in the Synthesis of AlPO-5 and SAPO-5 Materials
J. Phys. Chem. B, Vol. 109, No. 46, 2005 21543
Figure 3. SEM images of AlPO solids obtained with (1) A-bnhexam-3, (2) A-dbnpipz-3, and (3) A-bnpyrr-3.
filling of the AFI structure by the template species, i.e., the organic content in the samples, can be calculated from the chemical analysis and the total weight loss known from thermogravimetric (TG) analysis. The number of molecules within the pores of the AlPO-5 materials depends on the size of the organic agent; in this way, we estimated slightly more than one molecule of monobenzyl and slightly less than one molecule of the larger dibenzyl molecules to be present per unit cell. A remarkable influence of the template molecules on the crystal size and morphology was observed (Figure 3). Very large spherical aggregates showing a peculiar morphology, formed by the stacking of small layered crystals of a few micrometers in width, were found in the sample synthesized from benzylhexamethylenimine. An entirely different result was observed for benzylpyrrolidine. In this case, two different classes of aggregates could be distinguished: spherulitic, formed by very tiny elongated crystals, and some others formed by very thin platelet-like crystals. The material obtained with dibenzylpiperazine was composed of small spherical aggregates formed by small prismatic crystals. This observation provides clear evidence that the particular molecular structure of the template molecule has an influence on the nucleation and crystal growth of the AFI structure. SAPO. The XRD patterns of the silicoaluminophosphate solids after 3 days of heating are shown in Figure 4. The nature of the resulting crystalline phases formed from the siliconcontaining gels depends on the organic SDA used, in contrast to the findings for the undoped aluminophosphates; SAPO-5 crystallizes as the main phase only with benzylpyrrolidine and dibenzyldimethylammonium templates, the latter giving a higher crystallinity, although, in both cases, this is accompanied by a certain amount of the AlPO-H3-type material. The latter phase, with only traces of SAPO-5, was obtained when benzylpiperidine and benzylhexamethylenimine were used as SDAs, and just AlPO-H3 was obtained when dibenzylpiperazine was used. The crystallization of this dense phase, which cannot accommodate organic molecules within its small channels, can be taken as evidence of the lower directing ability of the latter organic molecule toward the AFI topology in the presence of silicon in the synthesis gel. Incorporation of silicon involves a distortion of the AlPO framework; therefore, the synthesis of SAPO-5 is a more critical test than the synthesis of AlPO-5 in assessing the structure-directing abilities of the organic molecules. The chemical analyses and 13C NMR spectra of the SAPO-5 samples
(benzylpyrrolidine and dibenzyldimethylammonium) indicate the integrity of the molecules inside the inorganic framework (Figure 5). TG analysis showed a pronounced weight loss between 350 and 440 K, attributed to the desorption of structural water (Figure 6), which confirmed the presence of minor amounts of AlPO-H3 in the solid. The SAPO-5 solids obtained with benzylpyrrolidine and dibenzyldimethylammonium as SDAs were also characterized by MAS NMR spectroscopy. Figure 7 shows the 31P MAS NMR spectra of the two samples, which display two peaks at -25.8 and -23.8 ppm produced by the presence of the AlPO-H3type material24 and a broader higher-field signal with a maximum at ca. -30 ppm characteristic of phosphorus atoms in the AFI structure type. These peaks are superimposed on a very broad band, probably associated with amorphous material, of low intensity for sample S-dbdm-3 but quite prominent in the spectrum of sample S-bnpyrr-3. The latter sample also shows a peak at -19 ppm. When the 31P spectrum was recorded under cross-polarization conditions from protons at short contact time (1 ms), the relative intensity of the low-field region, especially the broad band, increased, indicating the proximity of protons. This result suggests the presence of defect sites of O3POH type in the amorphous material present in the solid. The origin of the resonance at -19 ppm is not clear to us. Additional 31P NMR signals at -16 and -20 ppm accompanying AlPO-5 have previously been attributed to the presence of some byproducts in the sample.25 The 27Al NMR spectra are presented in Figure 8. The spectra of the two SAPO-5 samples (S-bnpyrr-3 and S-dbdm-3) consist of a main signal typical of tetrahedral 27Al at ca. 38 ppm with two components: a peak of pentacoordinated 27Al at 8 ppm and a band between 0 and -20 ppm with the second-order quadrupolar shape of octahedral 27Al. The AFI structure contains only one crystallographic Al site, so the appearance of two peaks in the tetrahedral region can be explained by the presence of the AlPO-H3-type phases. The signal at 8 ppm is usually attributed to the presence of unreacted pseudoboehmite or to framework aluminum coordinated to one water molecule. Finally, the octahedral aluminum signal at around -20 ppm is attributed to Al(OP)4(H2O)2 ions in the AlPO-H3-type material and probably also in the AFI structure type. 29Si MAS NMR spectroscopy was applied to investigate the incorporation of silicon atoms into the aluminophosphate frameworks. This might occur via two possible mechanisms: (i) the substitution of one phosphorus by one silicon atom, giving
21544 J. Phys. Chem. B, Vol. 109, No. 46, 2005
Go´mez-Hortigu¨ela et al.
Figure 4. XRD patterns of SAPO samples (heated for 3 days) synthesized using benzylpyrrolidine (S-bnpyrr-3), benzylhexamethylenimine (Sbnhexam-3), dibenzyldimethylammonium (S-dbdm-3), and dibenzylpiperazine (S-dbnpipz-3) as templates. The pattern of the sample synthesized with benzylpiperidine is the same as that of sample S-bnhexm-3.
Figure 6. TG analysis of SAPO samples S-dbnpipz-3 (dashed line) and S-dbdm-3 (solid line). The intense weight loss at around 400 K evidences the presence of the AlPO-H3 material. Figure 5. 13C CP/MAS NMR spectra of (top) S-bpyrr-3 and (bottom) S-dbdm-3 samples. * denotes rotation bands.
Si(4Al) sites and the appearance of a 29Si NMR signal at -89 ppm and (ii) the simultaneous substitution of neighboring aluminum and phosphorus atoms to form silica islands resulting in the formation of Si(4Si,0Al) environments, and the appearance of a 29Si band at around -110 ppm. Depending on the size of the siliceous domains, it is sometimes possible to distinguish different signals in the range between -89 and -110
ppm from Si(1Si,3Al), Si(2Si,2Al), and Si(3Si,1Al) environments generated at the border of the silica islands.26 The 29Si MAS NMR spectra of samples S-dbdm-3 and S-bnpyrr-3, presented in Figure 9, consist of a broad, featureless, and asymmetric band between -80 and -120 ppm that involves all of the possible silicon environments described above. The positions of the peak maxima in the 29Si spectra, at -102 and -108 ppm for S-dbdm-3 and S-bnbpyrr-3, respectively, as well as the asymmetric shape of the 29Si signal, indicate that the formation of silica islands predominates, although the presence
SDAs in the Synthesis of AlPO-5 and SAPO-5 Materials
Figure 7. 31P MAS NMR spectra of S-bnpyrr-3 (dashed line) and S-dbdm-3 (solid line) samples.
Figure 8. 27Al MAS NMR spectra of S-bnpyrr-3 (dashed line) and S-dbdm-3 (solid line) samples.
Figure 9. 29Si MAS NMR spectra of S-bnpyrr-3 (dashed line) and S-dbdm-3 (solid line) samples.
of other Si(nSi) environments is noticeable. The low-field shift of the maximum for sample S-dbdm-3 suggests that the dibenzyldimethylammonium molecule leads to the formation of slightly smaller silicon islands, of greater interest in the potential application of these materials as acid catalysts. Computational Results. Docking calculations show that the most stable arrangement for benzylpiperidine and benzylhexamethylenimine molecules involves dimers with facing benzyl rings (shown in Figure 10) and is similar to that found for benzylpyrrolidine.18 This configuration maximizes the packing efficiency. The stacking between consecutive dimers was also
J. Phys. Chem. B, Vol. 109, No. 46, 2005 21545 studied by molecular dynamics by inserting three dimers (six molecules) into the channel, yielding an optimal packing of two dimers per each three unit cells of AlPO-5, corresponding to a theoretical density of 1.33 SDA molecules per unit cell for each monobenzyl SDAs, i.e., benzylpyrrolidine, benzylpiperidine, and benzylhexamethylenimine. Let us now examine the dibenzyl SDAs; these molecules are oriented along the AFI channels with the two benzyl rings on opposite sides of the central CH2-N-CH2 or piperazine bridge. Both dibenzyl SDA molecules have one exposed benzyl ring on each side, similar to the benzyl end of the monobenzyl SDAs. It is not surprising, therefore, that interaction with the adjacent SDA molecules leads to a strong pairing of their benzyl rings, and in this case, the process is better described as the formation of a template chain (see Figure 11 for the dbdm molecule). The theoretical packing was studied by inserting three dibenzyl SDA molecules into the AFI channel, close to each other but without overlapping atoms, and letting the systems relax by MD calculations. At the end of the simulation, the packing value was calculated by measuring the distance between consecutive equivalent nitrogen atoms from the central molecule. These values were found to be 10.3 and 8.7 Å for dibenzylpiperazine and dibenzyldimethylammonium, respectively, which correspond to densities of 0.84 and 0.99 molecules per unit cell when normalized to the c parameter of the AFI unit cell (8.6 Å). The results described above indicate that a strong overlap between the benzyl rings in adjacent molecules drives a stable self-assembly process for each of the SDAs investigated. This supramolecular pairing of the benzyl rings plays a crucial role in yielding the large 12-MR pores of the AFI structure. Once we had determined the theoretical density values for all the molecules, we calculated their interaction energies with the AFI framework. This computational step had to be performed under periodic boundary conditions (PBC), to ensure a correct description of all possible interactions, including those between consecutive SDA molecules. A PBC model of the solid required us to employ an integer number of molecules and framework unit cells along the direction of the channel (z in the case of AFI). To define the periodicity required for the systems, we had to express the theoretical SDA density calculated above as a ratio of integer numbers. Thus, four molecules of the monobenzyl SDAs were inserted into a 1 × 1 × 3 supercell of AFI, to yield a density of 1.33 molecules per unit cell. For the dibenzylpiperazine molecule, we constructed a 1 × 1 × 6 supercell and inserted five molecules, which corresponds to a packing value of 0.83 molecules per unit cell, very similar to the value of 0.84 previously calculated. Finally, four molecules of dibenzyldimethylammonium were inserted in a 1 × 1 × 4 supercell, at a density of 1.0 molecules per unit cell. The calculated interaction energies, including both van der Waals and Coulomb terms, were -68.6, -64.2, -60.7, and -60.6 kcal/mol per unit cell of AlPO-5 for benzylpyrrolidine, benzylpiperidine, benzylhexamethylenimine, and dibenzylpiperazine, respectively (Table 3). The total interaction energy for dibenzyldimethylammonium cation was -147.1 kcal/mol, of which -69.0 kcal/mol was due to the short-range interaction with the framework. The interaction energy is normalized here to the number of unit cells in the solid, and not to the number of template molecules, as we have shown previously that the relevant parameter to compare different SDAs is the density of interaction energy.18 Other slightly different packing densities were also tried for the dibenzyl molecules. For dibenzylpiperazine, in addition to the 0.83 packing value previously calculated, SDA densities of 1.00 (four molecules in four unit cells), 0.86
21546 J. Phys. Chem. B, Vol. 109, No. 46, 2005
Go´mez-Hortigu¨ela et al.
Figure 10. Locations of benzylpyrrolidine (top), benzylpiperidine (middle), and benzylhexamethylenimine (bottom) inside the AFI channel. Left: View perpendicular to c axis, showing the channel filling by the molecules. Right: View along the c axis, showing the formation of the dimers. The circles highlight the regions of short-range repulsion between adjacent dimers.
TABLE 3. Calculated Interaction Energiesa per Unit Cell (kcal/mol) of the Molecules Inside the AFI Structure molecule
packing (molecules per unit cell) 1.33 1.00 0.86 0.83 0.80
bnpyrrol -68.6 bnpiper -64.2 bnhexam -60.7 dbnpipz -47.0 -60.9 -60.6 -60.0 dbdm -147.1 -132.4 -
organic content per unit cellb 16.0 17.3 18.6 17.2 17.0
a Most stable packing in bold. b Organic content (C + N atoms) of the most stable packing.
cell (Table 3), a sharp decrease from the value of -147.1 kcal/ mol at the packing density of 1.0. Discussion Figure 11. Two possible arrangements of dbdm molecules inside the AFI channels. Top: Initial arrangement of the molecules before the simulated annealing calculation (molecules are in the same conformation as in vacuum). Bottom: Most stable arrangement, obtained after the simulated annealing calculation, with benzyl rings belonging to the same molecule forming a 90° angle.
(six molecules in seven unit cells), and 0.8 (four molecules in five unit cells) molecules per unit cell were also tried. The new interaction energies were found to be -47.0, -60.9, and -60.0 kcal/mol per unit cell, respectively (Table 3). For dibenzyldimethylammonium, a lower packing of 0.86 was tried (six molecules were loaded in seven unit cells), and the new total interaction energy was calculated as -132.4 kcal/mol per unit
Although some AlPO-H3 is formed as an extra phase, all of the aromatic molecules examined here direct the gel chemistry toward the crystallization of the AFI structure for the AlPO composition. AlPO-H3 is a dense aluminophosphate containing openings of 8 MR where the organic molecules cannot be accommodated. The appearance of this phase shows an inferior templating ability of the molecule to direct the synthesis of the open-framework AFI structure. According to our experimental results, the structure-directing abilities of the five molecules are differentiated to a greater extent in the synthesis of SAPO-5 than in the synthesis of undoped AlPO-5, suggesting that the introduction of silicon atoms in the material is associated with a destabilization of the
SDAs in the Synthesis of AlPO-5 and SAPO-5 Materials framework. This result is consistent with the difficult synthesis of the silica polymorph SSZ-24, which is isostructural with AlPO-5. Only the most efficient template molecules, i.e., benzylpyrrolidine and dibenzyldimethylammonium, are able to direct the crystallization of SAPO-5. It is now well established that the capability of a molecule to direct the synthesis of a certain microporous material can be rationalized in terms of their interaction energy.12-15,18 The locations of the three monobenzylamines studied inside the AFI channels, known from the computational study, are presented in Figure 10. These three amines locate in very similar ways inside the AFI channels, forming dimers with the saturated ring perpendicular to the channel direction (Figure 10, right). The number of template molecules per unit cell is 1.33 for each amine, equivalent to 1 dimer for every 1.5 unit cells. However, because the number of atoms in the three molecules are not the same, the organic contents per unit cell of the microporous material are also different, being highest for the benzylhexamethylenimine SDA (Table 3) that contains a seven-membered saturated ring. The van der Waals interaction of the SDA with the framework is expected to increase with the number of atoms in the SDA per unit cell, yielding an order of interaction energy of benzylhexamethylenimine > benzylpiperidine > benzylpyrrolidine. However, the calculated interaction energies show the opposite trend. Thus, the template ability of the molecules depends not only on the density of organic material inside the pores, but also on the “quality” of the fitting of the SDA within the framework structure. Because the saturated ring of the monobenzylamines is perpendicular to the channel direction, the results show that five-membered rings of benzylpyrrolidine exhibit a better fit than the larger six- and seven-membered rings of benzylpiperidine and benzylhexamethylenimine. The packing energy, defined as the interaction energy between adjacent SDA molecules, was found to decrease from benzylpyrrolidine (-11.6 kcal/mol per unit cell) to benzylpiperidine (-6.7 kcal/mol per unit cell) and to benzylhexamethylenimine (-1.0 kcal/mol per unit cell). This feature affects the calculated interaction energies, which decrease in the order of benzylpyrrolidine > benzylpiperidine > benzylhexamethylenimine. To minimize the shortrange repulsion with the framework, therefore, the large sixand seven-membered organic rings rearrange in the channel and make the molecular association into dimers much weaker. These energy results explain the lower efficiency of benzylpiperidine and benzylhexamethylenimine, compared to benzylpyrrolidine, in directing the synthesis of SAPO-5, which is due to a poorer packing and fitting of the larger molecules within the AFI structure. The arrangement of the dibenzyl molecules inside the AFI structure also occurs by assembling benzyl rings of consecutive molecules parallel to each other. The calculated interaction energy of dibenzylpiperazine with the framework is the lowest among the SDAs studied, in agreement with the poor structuredirecting ability observed experimentally. In addition to this thermodynamic factor, the poor efficiency of dibenzylpiperazine in the synthesis might also have a kinetic explanation, related to the molecular size. The situation in which one molecule of dibenzylpiperazine occupies one unit cell is unstable (the interaction energy is -47.0 kcal/mol). The stable densities correspond to packing values smaller than 1, i.e., 0.86, 0.83, and 0.8 (-60.9, -60.6, and -60.0 kcal/mol) (Table 3). The dimensions of the dibenzylpiperazine molecule are therefore not suitable to fit in a commensurate way in the AFI unit cell, a feature that might hinder the kinetic nucleation of long-range ordered crystalline structures.
J. Phys. Chem. B, Vol. 109, No. 46, 2005 21547
Figure 12. Evolution of the z component of the moment of inertia of in-vacuum molecules versus time of simulation: (top) dibenzylpiperazine, (bottom) dibenzyldimethylammonium.
Finally, experimental work demonstrates that dibenzyldimethylammonium is quite efficient as an SDA for this structure. We cannot compare its interaction energy with those of the amines because of the different computational conditions. However, the experimental results clearly indicate that this molecule is a good template. This molecule has soft rotational degrees of freedom due to the presence of the single-bonded -CH2-N-CH2- unit in its middle, which provides four consecutive single bonds that are free to rotate without strong constraints. The molecule is therefore allowed to adopt different conformations inside the channel, and thus, it can better adjust to the channel geometry. In this case, the theoretical packing found was 1.0 molecule per unit cell. A lower packing value makes the system unstable. Thus, this molecule fits efficiently within the unit cell of the AFI-type structure, which makes the dibenzyldimethylammonium molecule a good template for its synthesis. Despite the fact that the most stable conformation of the molecule in vacuum is with benzyl rings roughly parallel to each other (Figure 1 and Figure 11, top), its most stable conformation inside the channels is in the form of chains with the two benzyl rings almost perpendicular to each other (the angles between the two benzyl rings of the same molecules are 80.6°, 88.0°, 88.1°, and 80.8°) (Figure 11, bottom). An additional set of calculations was carried out to complete the characterization of the two dibenzyl molecules as SDA. Molecular dynamics simulations (100 ps) of the two molecules in vacuum at the synthesis temperature (423 K) were run, to study the conformational flexibility of the molecules. A suitable way to describe the behavior consists of calculating the moment of inertia of the two molecules as a function of the simulation time; a change in molecular shape will be reflected in a different moment of inertia, and the molecular flexibility can be linked to the amplitude of variations in the moment of inertia. Figure 12 shows the evolution of the z component of the moment of inertia during the simulation. The z moment of inertia is the relevant one in our study because the channel-constrained space of AFI is oriented along the c axis. A different behavior of the two molecules is clearly observed. Whereas the z moment of inertia of dibenzyldimethylammonium is not markedly modified, that of dibenzylpiperazine fluctuates significantly during the simulation. As mentioned previously, the dibenzyldimethylammonium molecule has soft rotational degrees of freedom because of the presence of the -CH2-N-CH2- unit, providing four consecutive single bonds that are free to rotate. This -CH2N-CH2- chain is nearly aligned within the channel direction,
21548 J. Phys. Chem. B, Vol. 109, No. 46, 2005 which results in a stable moment of inertia along the z axis for the whole molecule; therefore, this molecule always stays in a conformation suitable to fit within the AFI channels. The rotation sites of dibenzylpiperazine, in contrast, are not aligned with the channel direction, and so, the fluctuation in the z moment of inertia is much higher. As a consequence, the latter molecule has a lower probability of adopting the necessary conformation (aligned with the channel direction) required for efficient structure direction, and it is a worse template for the synthesis of AFI. Finally, we performed 100 ps of MD simulations at 423 K for the dbdm and bpyrr systems to investigate the SDA mobility inside the AFI channel; these calculations showed that the supramolecular assemblies of SDA molecules are able to rotate about the c axis; however, at full packing density the displacement of the SDA molecules along the channel direction is restricted. The occurrence of such displacement is likely to require the presence of packing defects in the SDA assembly inside the micropores, such as a missing molecule or dimer. The set of results discussed above indicates that an efficient SDA for a zeolite synthesis requires the combination of favorable thermodynamic and kinetic properties. The former can be rationalized by calculating the interaction energy of the SDA with the framework; the latter by examining the degree of commensuration between SDA and framework and the variation of size among the conformations that the SDA molecule can adopt under the synthesis conditions. These requirements are simultaneously satisfied by dibenzyldimethylammonium and benzylpyrrolidine SDAs and, to a lesser extent, by benzylpiperidine and benzylhexamethylenimine, but not by dibenzylpiperazine molecules. The molecular moment of inertia, as well as its fluctuation in a MD simulation in vacuum, provides a valid descriptor for the kinetic suitability of a molecule to act selectively as an SDA in zeolite synthesis. Conclusions Elongated aromatic molecules containing one or two benzyl rings are efficient structure-directing agents for structures based on 12-MR channels. They form supramolecular assemblies inside the micropores by coupling the benzyl rings of consecutive molecules, thus fitting very efficiently within those channels. It is therefore reasonable to think of these types of aromatic molecules as selective structure-directing agents of frameworks composed of 12-MR channels. In the case of the AFI-structured solids examined here, the structure-directing ability of the organic molecules is differentiated experimentally to a greater extent in the synthesis of SAPO-5 than pure AlPO-5 materials. The type of SDA molecule employed in the synthesis influences not only the type and crystallinity of the microporous product, but also its morphology and the dopant (Si here) distribution in the solid. We have shown that dibenzyldimethylammonium forms silicon islands slightly smaller than those formed by benzylpyrrolidine. The ability of a template molecule to direct the crystallization of a certain microporous structure correlates with the interaction energy calculated with modeling. Its value does not depend solely on the space filling, i.e., the amount of organic material inside the structure; the efficiency of that space filling, meaning the quality of the interactions with the framework, and also the constraints of the molecule inside the structure have to be taken into account. Furthermore, a commensurate ratio between the
Go´mez-Hortigu¨ela et al. dimensions of the template molecule and those of the unit cell of the microporous material appears to enhance the ability to form long-range ordered crystalline lattices. A proper combination of structural rigidity to yield welldefined molecular shapes and a certain degree of freedom to adopt different conformers is necessary to design new efficient template molecules. The conformational flexibility of a molecule can be easily rationalized in terms of its moment of inertia (its z component in our study) during a simulation in vacuum. This fast calculation yields a suitable descriptor for the molecular flexibility and can provide an idea of the specificity of a molecule in directing a microporous structure. Acknowledgment. L.G.-H. acknowledges the Spanish Ministry of Education for a Ph.D. grant; F.C. thanks EPSRC for an Advanced Research Fellowship. The financial support of CICYT (Project MAT 2003-07769-C02-02) is acknowledged. We thank Accelrys for provision of software. We are grateful to Centro de Tecnologı´a Informa´tica for providing their computers for running the calculations. References and Notes (1) Wilson, S. T.; Lok, B. M.; Flanigen, E. M. U.S. Patent 4,310,440, 1982. (2) http://www.iza-structure.org/databases/. Accessed March, 2005. (3) Wagner, P.; Nakagawa, Y.; Lee, G. S.; Davis, M. E.; Elomari, S.; Medrud R. C.; Zones, S. I. J. Am. Chem. Soc. 2000, 122, 263. (4) Flanigen, E. M.; Patton, R. L.; Wilson, S. T. Stud. Surf. Sci. Catal. 1988, 37, 13. (5) Lok, B. M.; Cannan, T. R.; Messina, C. A. Zeolites 1983, 3, 282. (6) Davis, M. E.; Lobo, R. F. Chem. Mater. 1992, 4, 756. (7) Zones, S. I.; Nakagawa, Y.; Lee, G. S.; Chen, C. Y.; Yuen, L. T. Microporous Mesoporous Mater. 1998, 21, 199. (8) Zones, S. I.; Chen, C.-Y.; Finger, L. W.; Medrud, R. C.; Kivi, C. L.; Crozier, P. A.; Chan, I. Y.; Harris, T. V.; Beck, L. W. Chem. Eur. J. 1998, 4, 1312. (9) Kubota, Y.; Helmkamp, M. M.; Zones, S. I.; Davis, M. E. Microporous Mesoporous Mater. 1996, 6, 213. (10) Rollmann, L. D.; Schlenker, J. L.; Lawton, S. L.; Kennedy, C. L.; Kennedy, G. J.; Doren, D. J. J. Phys. Chem. B 1999, 103, 7175. (11) Lobo, R. F.; Zones, S. I.; Davis, M. E. J. Inclusion Phenom. Mol. Recognit. Chem. 1995, 21, 47. (12) Lewis, D. W.; Freeman, C. M.; Catlow, C. R. A. J. Phys. Chem. 1995, 99, 11194. (13) Sastre, G.; Cantin, A.; Dı´az-Caban˜as, M. J.; Corma, A. Chem. Mater. 2005, 17, 545. (14) Sastre, G.; Leiva, S.; Sabater, M. J.; Gime´nez, I.; Rey, F.; Valencia, S.; Corma, A. J. Phys. Chem. B 2003, 107, 5432. (15) Nakagawa, Y.; Lee, G. S.; Harris, T. V.; Yuen, L. T.; Zones, S. I. Microporous Mesoporous Mater. 1998, 22, 69. (16) Cora`, F.; Alfredsson, M.; Barker, C. M.; Bell, R. G.; Foster, M. D.; Saadoune, I.; Simperler, A.; Catlow, C. R. A. J. Solid State Chem. 2003, 176, 496. (17) Go´mez-Hortigu¨ela, L.; Pe´rez-Pariente, J.; Blasco. T. Microporous Mesoporous Mater. 2005, 78, 189. (18) Go´mez-Hortigu¨ela, L.; Cora`, F.; Catlow, C. R. A.; Pe´rez-Pariente, J. J. Am. Chem. Soc. 2004, 126, 12097. (19) Dauger-Osguthorpe, P.; Roberts, V. A.; Osguthorpe, D. J.; Wolff, J.; Genest, M.; Hagler, A. T. Proteins: Struct., Funct., Genet. 1988, 4, 21. (20) Gale, J. D. J. Chem. Soc., Faraday Trans. 1997, 93, 629. (21) Gale, J. D.; Henson, N. J. J. Chem. Soc., Faraday Trans. 1994, 90, 3175. (22) De Vita, A.; Gillan, M. J.; Lin, J. S.; Payne, M. C.; Stich, I.; Clarke, L. J. Phys. ReV. B 1992, 46, 12964. (23) Leslie, M.; Gillan, M. J. J. Phys. C 1985, 18, 973. (24) Duncan, B.; Sto¨cker, M.; Gwinup, D.; Szostak, R.; Vinje, K. Bull. Soc. Chim. Fr. 1992, 129, 98. (25) Gougeon, R. D.; Brouwer, E. B.; Bodart, P. R.; Delmotte, L.; Marichal, C.; Che´zeau, J.-M.; Harris, R. K. J. Phys. Chem. B 2001, 105, 12249. (26) Montoya-Urbina, M.; Cardoso, D.; Pe´rez-Pariente, J.; Sastre, E.; Blasco, T.; Forne´s, V. J. Catal. 1998, 173, 501.
