J. Phys. Chem. B 1997, 101, 6749-6752
6749
Computer Simulation Study of Aluminum Incorporation in the Microporous Titanosilicate ETS-10 Maria E. Grillo* and Jose´ Carrazza INTEVEP, S. A., Research and Technological Support Center of Petro´ leos de Venezuela, P.O. Box 76343, Caracas 1070A, Venezuela ReceiVed: NoVember 19, 1996; In Final Form: May 5, 1997X
The preferred framework site for Si substitution by Al in the ETS-10 structure has been modeled by both lattice energy minimization (LEM) and semiempirical quantum chemical (QC) calculations on model clusters. Both approaches suggest that Al substitution in the ETS-10 structure will follow an Al, Ti avoidance rule, consistent with conclusions drawn from analysis of solid-state 29Si and 27Al NMR spectra of Al-substituted ETS-10 reported elsewhere. The Al-substituted ETS-10 structure (ETAS-10) involving a Al-O-Si-O-Ti (T1) linkage was successfully minimized, showing only minor deviations from the optimized ETS-10 structure. In contrast, constant pressure minimizations of ETAS-10 substituted at T2 sites lead to the cleavage of the Ti-O-Al linkage. Indeed, QC calculations indicate that, compared to Al-O-Si-O-Ti linkages, Al-OTi direct linkages result in enhanced electronic repulsion between neighboring negative charges on Al and Ti sites and higher lattice strains. This suggests that both electronic and strain factors are influencing the observed preferential Al siting.
1. Introduction ETS-10 is a member of a new class of materials, containing octahedral Ti(IV) and tetrahedral Si in a microporous crystalline array.1,2 Its structure was recently solved by Anderson et al.3 It consists of a random stacking of orthogonally arrayed 12ring channels that could be described in terms of the intergrowth of two ordered polymorphs (A and B). Because of this interconnecting network of wide pore straight channels, and a large extraframework cation density (equivalent to a zeolite with Si/Al ratio ) 2.5), ETS-10 is a potentially useful material for applications as a catalyst and/or a sorbent. Furthermore, it has been reported that Si can be isomorphously substituted by Al in the ETS-10 structure, but only on sites that avoid Ti-O-Al linkages.4,5 We have chosen a theoretical approach to evaluate properties of this new material, employing calculation strategies previously used in the study of zeolites. Our first step was to model the position of the extra-framework cations, compensating the -2 charge on the TiO6 octahedra.6 Knowledge of the position and energy of these sites within the structure is an important requirement for modeling additional structural properties and determing ion-exchange properties. The predicted relative Na+ and K+ stabilities within the structure were consistent with our ion-exchange experiments.7 Furthermore, the inclusion of counterions at the optimized positions in the ETS-10 lattice significantly improved the matching between experimental and simulated X-ray diffraction patterns,8 confirming the modeled cation site positions. In the present work, the preferential Al siting between substitution in the two possible Si chemical environments in ETS-10, Si(4Si,0Ti) and Si(3Si,1Ti), is investigated (see Figure 1). The main objective is to identify the reasons for the Al substitution exclusion rule observed for titanosilicate structures. The energy preferred framework site for Si substitution by Al into the ETS-10 structure has been modeled by applying two different computational approaches. Namely, lattice simulations * Corresponding author. X Abstract published in AdVance ACS Abstracts, June 15, 1997.
S1089-5647(96)03860-6 CCC: $14.00
Figure 1. Schematic diagram of the ETS-10 structure. The Ti atoms are shown as dark spheres in the framework and Si and O atoms as the light ones. The nonframework K+ ions are represented by the large black spheres. The three black atoms indicated by arrows in the structure represent the Al T sites modeled.
based on classical periodic potential energy functions (force fields) and semiempirical quantum chemical (QC) calculations on model clusters. In recent studies, both approaches have been utilized to explain the Al site occupancy in zeolites. Therefore, the present study adopts both methods to investigate the energetics of Si substitution by Al into the two possible Si environments of ETS-10. 2. Methodology 2.1. Lattice Simulations. Lattice energy minimizations on Al-substituted K-ETS-10 structures (K-ETAS-10) were performed using MSI’s lattice simulation program Discover.13 The lattice energy of the K-ETAS-10 structures is calculated and © 1997 American Chemical Society
6750 J. Phys. Chem. B, Vol. 101, No. 34, 1997
Grillo and Carrazza
TABLE 1: Interatomic Potential Parameters Useda
a
b
c
d
i
Ai
Bi
qi
O Si Ti K Na Al
388 611.3727 103.8039 6 916.3989 12 886.4561 67 423.6364 278.3910
0.1893 0.007 3262.4997 0.0001 0.0005 1690.1496
-1.2 +2.4 +1.6 +1.0 +1.0 +1.4
Lennard-Jones:
qiqj Aij Bij + 12 - 6 . rij rij rij The off-diagonal potential parameters take the form Aij ) x(Ai×Aj), Bij ) x(Bi×Bj). b Refers to the atomic partial charge. c kcal mol-1 Å12, ref 13. d kcal mol-1 Å13, ref 13. V(rij) )
optimized utilizing an interatomic potential energy function which accounts for the nonbonding interactions between ions. The short-range effects are calculated with a Lennard-Jones (12-6) potential to an energy cutoff of 13.5 Å. The electrostatic long-range interactions are calculated exactly by the Ewald method.14 The potential parameters and partial charges used are presented in Table 1. The Al substitutions were carried out on a fully relaxed K-ETS-10 structure. As part of its optimization process, the nonframework cation sites for K-ETS-10 and Na-ETS-10 were modeled.6 An optimization strategy, combining a Monte Carlo packing (MCP) procedure developed by MSI15 with the lattice energy minimization (LEM) technique, was employed. In the present work, the negative charge created by the Al incorporation is neutralized with an additional extraframework K+ ion, introduced in the model structure using again the MCP procedure. Although the synthesized material contains a mixture of Na+ and K+ ions, this approximation simplifies the problem of considering the different permutations of Na+ and K+ around a given T site. This is a reasonable model, since the aim is to explore the Al substitution in the two different chemical environments within the structure. Nevertheless, one should be aware that the present results serve as a qualitative guide for Al site preference, because in general Al-site energies are dependent on the type of the counterion. The minimum energy position of the additional K+ ion in the K-ETAS-10 structure was determined. Thereby, for each of the Al-substituted structures considered, 30 different lowenergy trial configurations of this additional K+ ion in the K-ETAS-10 lattice were generated, according to an energy threshold for the short-range interactions. These structures were then optimized by means of LE minimizations. The optimization strategy used consists of an initial minimization of the K-ETAS-10 lattice energy with respect only to the 33 K+ counterion coordinates. This is followed by an optimization, in which these structures are fully relaxed with respect to the whole framework plus counterion atomic positions at constant volume. Finally, a full constant pressure minimization of each of the considered ETAS-10 structures is carried out allowing for variations in the cell volume (cell parameters). In the present calculations the unit cell contains 304 framework atoms and 33 charge-compensating K+ cations. 2.2. Semi-Empirical Quantum Chemical Calculations. Restricted Hartree-Fock cluster calculations were performed in the present work within Zerner’s INDO model (intermediate neglect of differential overlap) using MSI’s package Zindo.16 The constrained minimization of the model clusters was performed within an approximate Hessian matrix scheme using the BFGS (Broyden, Fletcher, Goldfarb, and Shanno) updating algorithm. Zerner et al. have reported INDO/1 geometry
optimizations up to second-row transition metal species in good agreement with experimental data and high level ab initio calculations, see ref 17 and references cited therein. The energy preference of the Al insertion positions studied are estimated using a methodology based on an averaged sum of the Al replacement energies of all the rings connected to a given T site in the structure,18 which has been succesfully applied to the modeling of the relative Al sites stabilities in a variety of zeolite structures (e.g., mordenite, ferrierite, and ZSM-5 zeolites). This approach has been developed for the limiting case of Al incorporation into siliceous zeolite structures. According to Blanco et al., the Al replacement energy ETk ij for a given k-member ring i, connected to a given Tj site is evaluated as the energy change of a process in which a Si atom is substituted by an Al atom in a tetrahedral site, k k k ETij ) EaTij - EsTij
(1)
where EaTk ij is the ring energy when the Tj site is occupied by Al, and EsTk ij is the energy when the site is occupied by Si. The total substitution energy (ETj), is thus given by the sum over all partial ring substitution energies ETk ij associated to a given Tj site, weighted by the number of Si T sites in the ring, (1/Pi),
E Tj )
∑i ETijkPi
(2)
For the present study, the methodology originally proposed by Blanco et al. has been modified to account for the presence of tetrahedral and octahedral sites in the structure. Multiplying ETk ij by Pi as in eq 2 would imply averaging the ring Al insertion energy by nonequivalent Si T sites involving both SiO-Ti and Si-O-Si types of linkage as well as by octahedral Ti sites. This is avoided by calculating the Al-insertion energy (ETj) as the sum of the nonnormalized partial ring energies (ETk ij) averaged by the number of rings NTRj directly linked to a given site Tj,
ETj )
1 NTj R
∑i ETijk
(3)
The present approach also differs from the original methodology of Blanco et al.18 in that the ring model clusters are allowed to relax during the SCF calculations and that the charge-balancing cations are included as part of the clusters. However, summing energy contributions of independent relaxed ring clusters (eq 3) represents a zeroth-order approximation to the aluminum substitution energy. In fact, the resulting direction of the oxygen relaxation to accommodate the Al ion will depend on the ring connectivity of neighboring T sites in the periodic structure. The lengthening of the Al-O bond is thus compensated by the compression of nearest neighbors Si-O bonds in the lattice. Therefore, at this level of approximation, only qualitative trends of the relative Al site preference is to be expected. For the ETS-10 structure, the three- and seven-member rings directly connected to the Si T sites under study are shown in Figure 2. T1 and T2 are the tetrahedral sites corresponding to the Si chemical environments (4Si, 0Ti) and (3Si, 1Ti) in ETS10, respectively (Figure 1). Considering the ring connectivity of ETS-10 in eq 3 and omitting the i-counter index, the difference in Al substitutional energies into T1 and T2 sites is then given by
Study of Aluminum Incorporation
J. Phys. Chem. B, Vol. 101, No. 34, 1997 6751 TABLE 2: Cluster Calculations Result for Al Insertion at Si T Sites in K-ETS-10 ring
k
NSia
NTib
ETk 1
ETk 2
A B
3 7
2 6
1 1
583.5
966.3 771.8
Number of Si-tetrahedra. b Number of Ti-octahedra. NRTj is the number of rings connected to the Tj site. c Insertion energy of the k-member ring, connected to the Tj site, in kcal mol-1. a
Figure 2. Schematic diagram of the three- and seven-member ring model clusters associated to the T1 and T2 tetrahedral sites considered. The Ti atoms are shown as unlabeled dark spheres in the framework, and Si and O atoms as the light ones. The nonframework K+ ions are omitted for simplicity.
1 1 7 12 ET2 - ET1 ) (E3T2 + 2E5T2 + 2E12 T2 + ET2) - (2ET1 + 6 6 2E5T1 + 2E7T1) (4) Since siliceous ring structures connected to T1 and T2 present the same Al insertion energies, this expression can be written as follows:
1 ET2 - ET1 ) (E3T2 + E7T2 - 2E7T1) 6
(5)
where ET3 2, ET7 2 and ET7 1 are the partial Al insertion energies for the seven- and three-member rings, respectively. 3. Results and Discussion 3.1. Lattice Energy Minimizations. The preferential siting of an Al ion into the ETS-10 structure is investigated comparing the lattice energies of ETAS-10 involving either Al-O-SiO-Ti (T1) or Al-O-Ti (T2) linkages. Of the four crystallographically distinct T2 sites reported by Anderson et al.,3,19 we have only considered two sites (A and B in Figure 1), since their chemical and structural environments are very similar, and therefore, we do not expect substantial differences in substitution energies among them. The present LEM calculations indicate that the T1 position is energetically favored for Al incorporation, compared to that of T2. At constant volume, the ETAS-10 structure substituted at the T1 tetrahedra was optimized to a minimum lattice energy of -113 033.5 kcal mol-1. In contrast, structures substituted at T2 type sites could only be partially optimized with respect to the nonframework cation positions. Further attempts to relax the whole atomic coordinates at constant volume failed to converge to an energy minimum. Furthermore, in the case of energy minimizations at constant pressure, ETAS-10 substituted at T2 sites leads to the cleavage of the Ti-O-Al linkages, while the substituted at T1 converges to a structure similar to the one obtained at constant volume, of lattice energy -113 263.2 kcal mol-1. These results could be explained based on the different ring connectivity of both sites. As opposed to site T1, T2 is connected to a three-member ring (Figure 1). For this ring, a small O-Si-O angle of 107° deviated from the ideal tetrahedral value was obtained, as well as a strained Si-O-Si angle of 132.3°, differing by about 8° from that found in the 12-ring. Therefore, T2 is critically affected by both the additional stress
in the already strained small three-ring caused by the introduction of a large Al ion, as well as the enhancement in electronic repulsion due to neighboring negative charges on Al and Ti sites. These resulting differences in the optimization behavior of the structures substituted at T1 and T2 sites may in turn be related to the observed Al, Ti avoidance in ETAS-10 samples by 29Si and 27Al MAS NMR spectra.4,5 Average distances of 1.80 and 1.60 Å for the Al-O and Si-O bonds have been obtained for the minimized ETAS-10 structure containing Al in the T1 site. The Al-O-Si bond angle in this structure was obtained to be 140° in average, compared to 156° for neighboring Si-O-Si linkages, with this latter value is about 3.3% higher than the one obtained for the siliceous ETS-10 structure. 3.2. Cluster Calculations. In the present cluster calculations, unsaturated valences at the terminal O atoms were completed with H atoms at a distance of 1.03 Å. The position of oxygen and hydrogen atoms in the peripheral OH bonds were kept fixed during the structure optimizations in order to constrain the environment of the ring as if it were part of the framework structure. The negative charges introduced by Ti and Al ions in the ring cluster models were compensated by K+ ions. The starting cluster geometries and charge-compensating cation positions were taken from the Discover-converged K-ETS-10 configuration. The position of the ring atoms (Si, Al, O, and/or Ti) and charge-compensating K+ ions were then allowed to vary in all model clusters, in order to obtain minimum constraint configurations. The optimized cluster configurations do not deviate significantly from the original geometries. The partial substitution energies (ETk ij) for the rings appearing in eq 5 are listed in Table 2. The partial Al ring substitution energy becomes more repulsive with decreasing cluster size. Upon Al insertion in the seven-member ring, the partial ring substitution energy at T2 (directly connected to a Ti octahedra) is 188.2 kcal mol-1 more repulsive than that obtained at T1 (T site linked to another Si), see Table 2. This energy difference can be ascribed to an increase in electronic repulsion, due to neighboring negative charges on the Al and Ti sites as evidence by the obtained Mulliken partial charges (MPC) in the INDO calculations. For instance, in the seven-member ring, a negatively charged oxygen atom (MPC ) -0.459) links the Al-tetrahedral (AlO(OH)2), MPC ) -0.434) and Ti-octahedral (TiO(OH)4, MPC ) -0.205) units for the T2-type of substitution. In contrast to the T1-substituted seven-ring, these units are separated by a positively charged Si(OH)2 unit (MPC ) 0.081). A similar energy difference is obtained between Al insertion next to Ti (T2 site) in a three-member vs a sevenmember ring (Table 2), which might be associated to the additional strain caused by the insertion of a larger Al ion into the small ring. The calculated enhancement in both electronic repulsion and lattice strain on Al insertion at T2 leads to a difference in substitutional energy into T1 and T2 sites (ET2 - ET1) of about 95 kcal mol-1. This result suggests a preference for Al insertion into sites connected to tetrahedral Si (T1). The striking energy gap obtained for Si substitution by Al at the two different Si
6752 J. Phys. Chem. B, Vol. 101, No. 34, 1997 environments could be interpreted so as Al incorporation will be in practice favored into sites connected to Si tetrahedra, avoiding connection to Ti octahedra, and thus justifying the experimentally observed Al, Ti avoidance rule for ETS-10. 4. Conclusions The present study makes use of two methodologies for modeling Al siting in the two different Si chemical environments of the titanosilicate structure ETS-10. Both methods suggest an Al, Ti avoidance in the ETS-10 structure and support conclusions drawn from 29Si and 27Al MAS NMR experiments. Furthermore, semiempirical cluster calculations on partially optimized clusters seem to indicate that both electronic repulsion between neighboring charges and lattice strain effects influence Al siting and result in Al, Ti avoidance in the ETS-10 structure. These calculations suggest that for ETAS-10 these factors are both influencial in determining the Al site stabilities. This result illustrates the importance of relaxing the model clusters used, in order to account for strain effects. Acknowledgment. The authors would like to acknowledge M. W. Anderson for useful discussions and thank INTEVEP S. A. for permission to publish this work. References and Notes (1) Kuznicki, S. M. U.S. Patent 4 853 202, 1989. (2) Kuznicki, S. M.; Thrush, A. K. European Patent 0405978A1, 1990.
Grillo and Carrazza (3) Anderson, M. W.; Terasaki, O.; Ohsuna, T.; Phillippou, A.; Mackay, S. P.; Ferreira, A.; Rocha, J.; Lidin, S. Nature 1994, 367, 347. (4) Anderson, M. W.; Philippou, A.; Lin, Z.; Ferreira, A.; Rocha, Angew. Chem., Int. Ed. Engl. 1995, 34, 1003. (5) Anderson, M. W.; Rocha, J.; Lin, Z.; Philippou, A.; Orion, I.; Ferreira, A. Microporous Mat. 1996, 6, 195. (6) Grillo, M. E.; Carrazza, J. J. Phys. Chem. 1996, 100, 12261. (7) Grillo, M. E.; Lujano, J.; Carrazza, J. In Proceedings of the 11th International Conference on Zeolites, August, 1996; Chon, H., Ihm, S.-K., Uh, Y. S., Eds.; Elsevier: Amsterdam, 1997; p 2323. (8) Freites, A.; Grillo, M. E.; Anderson, M. W.; Carrazza, J. Manuscript in preparation, 1997. (9) Schro¨der, K. P.; Sauer, J. In Proceedings of the 9th International Conference on Zeolites, August, 1992; von Ballmoos, R., et al., Eds.; Butterworth-Heinemann: Boston, 1993; p 687. (10) Schro¨der, K. P.; Sauer, J.; Leslie, M.; Catlow, C. R. A. Zeolites 1992, 12, 20. (11) Derouane, E. G.; Fripiat, J. G. Zeolites 1985, 5, 165. (12) Fripiat, J. G.; Galet, P.; Delhalle, J.; Andre, J. M.; Nagy, J. M.; Derouane, E. G. J. Phys. Chem. 1985, 89, 1932. (13) DiscoVer Molecular Simulations Program, Version 94.1; Molecular Simulations, Inc.: San Diego, CA, 1995. (14) Ewald, P. P. Ann. Phys. 1921, 64, 253. (15) Catalysis Package, Version 2.3.7; Molecular Simulations, Inc.: San Diego, CA, 1995. (16) Zindo Package, Version 950; Molecular Simulations, Inc.: San Diego, CA, 1996. (17) Anderson, W. P.; Cundari, W. P.; Zerner, M. C. Int. J. Quantum Chem. 1991, 39, 31. (18) Blanco, F.; Urbina-Villalba, G.; Ramirez de Agudelo, M. M. Molecular Simulations 1995, 14, 165. (19) Anderson, M. W.; Terasaki, O.; Ohsuna, T.; O’Malley, P. J.; Philippou, A.; MacKay, S. P.; Ferreira, A.; Rocha, J.; Lidin, S. Philos. Mag. B 1995, 71, 813.
