progress in inorganic chemistry€¦ · spin-state transitions ..... 527 by edgar konig institut...

PROGRESS IN INORGANIC CHEMISTRY

Edited by

STEPHEN J. LIPPARD

DEPARTMENT OF CHEMISTRY MASSACHUSETTS INSTITUTE OF TECHNOLOGY CAMBRIDGE, MASSACHUSETTS

VOLUME 35

AN INTERSCIENCE@ PUBLICATION JOHN WILEY & SONS New York Chichester - Brisbane Toronto Singapore

Progress in Inorganic Chemistry

Volume 35

Advisory Board

THEODORE L. BROWN UNIVERSITY OF ILLINOIS, URBANA, ILLINOIS

JAMES P. COLLMAN STANFORD UNIVERSITY, STANFORD, CALIFORNIA

F. ALBERT COTTON TEXAS A & M UNIVERSITY, COLLEGE STATION, TEXAS

RONALD J. GILLESPIE McMASTER UNIVERSITY, HAMILTON, ONTARIO, CANADA

RICHARD H. HOLM HARVARD UNIVERSITY, CAMBRIDGE, MASSACHUSETTS

GEOFFREY WILKINSON IMPERIAL COLLEGE OF SCIENCE AND TECHNOLOGY, LONDON, ENGLAND

PROGRESS IN INORGANIC CHEMISTRY

Edited by

STEPHEN J. LIPPARD

DEPARTMENT OF CHEMISTRY MASSACHUSETTS INSTITUTE OF TECHNOLOGY CAMBRIDGE, MASSACHUSETTS

VOLUME 35

AN INTERSCIENCE@ PUBLICATION JOHN WILEY & SONS New York Chichester - Brisbane Toronto Singapore

An Interscience" Publication

Copyright0 1987 by John Wiley & Sons, Inc.

All rights reserved. Published simultaneously in Canada.

Reproduction or translation of any part of this work beyond that permitted by Section 107 or 108 of the 1976 United States Copyright Act without the permission of the copyright owner is unlawful. Requests for permission or further information should be addressed to the Permissions Department, John Wiley & Sons, Inc.

Library of Congress Catalog Card Number: 59-13035 ISBN 0-471 -84291-5

Printed in the United States of America

10 9 8 7 6 5 4 3 2 1

Contents

New Light on the Structure of Aluminosilicate Catalysts . . . . . 1 By J. M. THOMAS Department of Physical Chemistry, University of Cambridge, Cambridge, England and C. R. A. CATLOW Department of Chemistry, University of Keele, Staffordshire, England

Rational Design of Synthetic Metal Superconductors . . . . . . 51 By JACK M. WILLIAMS, HAU H. WANG, THOMAS J. EMGE, URS GEISER, MARK A. BENO, PETER c. W. LEUNG, K. DOUGLAS CARLSON, ROBERT J. THORN, and ARTHUR J. SCHULTZ Chemistry and Materials Science Divisions, Argonne National Laboratory, Argonne, Illinois and MYUNG-HWAN WHANGBO Department of Chemistry, North Carolina State University, Raleigh, North Carolina

Binding and Activation of Molecular Oxygen by Copper Complexes . . . . . . . . . . . . . . . . . . . . . 219

By KENNETH D. KARLIN Department of Chemistry, State University of New York at Albany, Albany, New York and YILMA GULTNEH Department of Chemistry, University of Pittsburgh at Johnstown, Johnstown, Pennsylvania

The Chemistry of 1,4,7-Triazacyclononane and Related Tridentate Macrocyclic Compounds . . . . . . . . . . . . . . 329

By PHALGUNI CHAUDHURI and KARL WIEGHARDT Lehrstuhl fur Anorganische Chemie der Ruhr- Universitat Bochum, Bochum, Federal Republic of Germany

V

vi CONTENTS

Butterfly Cluster Complexes of the Group VIII Transition Metals . . . . . . . . . . . . . . . . . . . . . . 437

By ENRICO SAPPA Istituto di Chimica Generale ed Inorganica, Universita di Torino, Torino, Italy and ANTONIO TIRIPICCHIO Istituto di Chimica Generale ed Inorganica, Universita di Parma, Parma, Itafy and ARTHUR J . CARTY and GERALD E. TOOGOOD The Guelph- Waterloo Centre for Graduate Work in Chemistry, Department of Chemistry, University of Waterloo, Waterloo, Ontario, Canada

Structural Changes Accompanying Continuous and Discontinuous Spin-State Transitions . . . . . . . . . . . . . . . . . . . . 527

By EDGAR KONIG Institut fur Physikalische und Theoretische Chemie, University of Erlangen-Nurnberg, Erlangen, Federal Republic of Germany

Subject Index . . . . . . . . . . . . . . . . . . . . . . . 623

Cumulative Index, Volumes 1-35 . . . . . . . . . . . . . . . 651

Progress in Inorganic Chemistry Volume 35

New Light on the Structure of Aluminosilicate Catalysts

J. M. THOMAS*

Department of Physical Chemistry University of Cambridge Lensfield Road, Cambridge, England

and

C. R. A. CATLOW

Department of Chemistry University of Keele Keele, Staffordshire, England

CONTENTS

I. INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . 2

11. FOUR DISTINCT APPROACHES TO THE STRUCTURE OF ALUMINOSILICATES . . . . . . . . . . . . . . . . . . 8

A. Neutron Diffraction and Recent Developments in X-Ray Diffraction . . 9 B. Solid-state NMR . . . . . . . . . . . . . . . . . . . . . 11 C. High-Resolution Electron Microscopy (HREM) . . . . . . . . . . 12 D. Computer-Modeling Techniques . . . . . . . . . . . . . . . 13

1. Nature and Scope of Solid-state Simulations. . . . . . . . . . 13 2. Calculation of the Properties of Crystals . . . . . . . . . . . 15 3. Calculation of Crystalline Defects . . . . . . . . . . . . . 15 4. Calculation of the Structure of Surfaces . . . . . . . . . . . 17 5. Molecular Dynamics Simulations . . . . . . . . . . . . . . 17 6. Interatomic Potentials . . . . . . . . . . . . . . . . . 18

*Present-address: Davy Faraday Research Laboratory, The Royal Institution, 21, Al- bemarle St., London, 'WIX 4BS, England.

2 J. M. THOMAS AND C. R. A. CATLOW

111. SOLVING SPECIFIC STRUCTURAL PROBLEMS . . . . . . . . . . 20

A. Examples Based Largely on Neutron-Powder-Profile Methods . . . . . 21 1. Is There Strict Alternation of Si4+ and A13+ in the Framework

of Zeolite-A? . . . . . . . . . . . . . . . . . . . . 21 2. Where Do Guest-Species Reside within the Intracrystallite

Cavities? . . . . . . . . . . . . . . . . . . . . . . 21 3. Can the Nature of the Active Site Be Identified? . . . . . . . . 27

B. Examples Based Largely on Solid-state NMR . . . . . . . . . . 28 1. Can the Composition of the Aluminosilicate Framework Be Evaluated

from the -9si MASNMR Spectrum? . . . . . . . . . . . . . 29 2. How Does One Monitor Dealurnination and Realumination of a

Zeolite Framework? . . . . . . . . . . . . . . . . . . 31 3. What of Silicon-Aluminum Ordering in Zeolites X and Y? . . . . 32

C. Examples of Some Recent Simulation Studies. . . . . . . . . . . 34

2. Structure and Energetics of Sorbed Species . . . . . . . . . . 37

E. Direct “Real-Space” Imaging of Intergrowths. . . . . . . . . . . 40

1. Cation Distributions . . . . . . . . . . . . . . . . . . 34

Solving the Structure of a Powdered Zeolitic Catalyst: A Multipronged A p p r o a c h . . . . . . . . . . . . . . . . . . . . . . . 38

F. Clays and Clay Minerals . . . . . . . . . . . . . . . . . . 41 References. . . . . . . . . . . . . . . . . . . . . . . . . 44

D.

I. INTRODUCTION

Aluminosilicates, which constitute one of the largest classes of minerals, have, from the earliest times, been used as catalysts or catalyst supports for a number of commercially important reactions. Acid-treated clays, for example, were used from the 1920s to the mid-1940s for the cracking of oils (1) and for the reforming (i.e., the isomerization) of short-chain hydrocarbons such as pentanes to octanes (2). Since the early 1960s synthetic aluminosilicates in the form of zeolites (3-5) have been the dominant catalysts in the petrochemical industry. Zeolites Y and the so-called pentasils, of which ZSMJ and ZSM-11 are the most renowned (4, 6, 7) mem- bers, are nowadays extensively used worldwide. Ultrastabilized (8) zeolite Y, the structure of which is essentially that of the rare zeolite mineral faujasite (9), is the cornerstone of present-day petroleum cracking and hydrocracking processes; annual consumption is close to 2000,000 tons. ZSM-5 is the catalyst of choice in the conversion of methanol to gasoline and benzene and ethene to ethylbenzene. It is also used in so-called de- waxing and selectofonning processes (4). Acid-washed mordenite is used as a catalyst support (for the platnium group metals) for reforming; and the aluminosilicate mineral erionite, like the pentasils ZSM-5 and ZSM- 11, also finds use as a commercial shape-selective catalyst.

THE STRUCTURE OF ALUMINOSILICATE CATALYSTS 3

During the last decade there has been a resurgence of interest in the utilization of clays, and more recently, pillared clays, as versatile catalysts for the conversion of organic species into more useful products. In Table I some of the reactions that have recently come into prominence are enum- erated (10-14).

Elucidating the structure of zeolites and clays is not, in general, an easy task. Very few of the zeolites (especially the highly siliceous ones that currently figure emminently as novel catalysts)-and hardly any of the clays-are available as single crystals, so that the classical X-ray techniques cannot be used. Even in the rare situations when good quality crystals are available, distinguishing structural A1 from Si is difficult because of the similarity in scattering strength of these two elements. However, enough background X-ray work, along with inspired powder-diffraction studies (15, 16), has been done to compile an atlas of zeolite structures, embracing both natural and synthetic ones (17). A compilation of the positions of extra-framework (exchangeable) cations in a wide range of zeolites is also available (18).

The open structures of zeolites can be envisaged as having been assembled from building blocks such as those shown in Figs. 1 and 2. Zeolite structures can then be readily classified according to the secondary building units (shown in Fig. 1) that are present (see Table 11).

The name “zeolite” (from the Greek (EW “to boil” and hi6a “stone”) was coined by Cronstedt in 1756 to describe the behavior of the newly discovered mineral stilbite. When heated, stilbite loses water rapidly and thus seems to boil. Zeolites are a class of framework silicates (other classes include feldspars and feldspathoids) that are built from comer-sharing S D - and A10:- tetrahedra and contain regular systems of intracrystalline cavities and channels of molecular dimensions. The net negative charge of

TABLE I A Selection of Organic Reactions Catalyzed by Clays

~

Cracking and hydrocracking of hydrocarbons Dehydration of alkanols, with the formation of ethers, alkenes, and naphthenes Isornerizations, alkylations, and cyclizations Conversion of primary amines to secondary ones Dirnerizations, oligomerizations, and polymerizations Oxidations and reductions Hydration, alkylation, and acylation of alkenes to form alkanols, ethers, and esters Hydrogen exchange and hydrodesulfurizations Decarboxylations and lactonizations Polycondensations (e.g., peptides from amino acids) Porphyrin formation from benzaldehyde and pyrrole

4 J . M. THOMAS AND C. R. A. CATLOW

S4 R S6 R

@ !- . .- .. . . . . . * . . . . .

a

S8 R

d TaOi' 5-1

@ , ..... ;

I . ..... . . . .

B @ .... 2

..... . . -__ 2. . . . . . . . . .

7

D 4 R

& TIOON) 4 4 1

@

D8 R

@ . . . .........

c

@

D 6 R

Figure 1. A selection of the secondary building units [double-four (D4), double-six (D6) rings; double-eight (D8), sodalite (p), cancrinite (E), gmelinite (y) cages] from which the structures of zeolites are derived. The a-cage is synonymous with the supercage in zeolites A and ZK-4.

the framework equal to the number of the constituent aluminum atoms is balanced by exchangeable cations, M"' , typically sodium, located in the channels that normally also contain water. The general oxide formula of a zeolite is

It is invariable found that y 2 x . The simplest interpretation of this ine- quality, given that each silicate and aluminate tetrahedron is linked via oxygen bridges to four other tetrahedra, is that aluminate tetrahedra cannot be neighbors in a zeolite framework, that is, that A1-0-A1 linkages are forbidden. This requirement, known as the Loewenstein rule (19), will be discussed later.

It has been traditional until very recently to regard all zeolites as having been derived from aluminosilicates, but the definition must now be ex-

Zeolite A

- 4.2A G

\

- 7 . 4 i '

Sodalite Cage

Zeolite X 8. Y

Figure 2. Illustration of how zeolites X and Y and zeolites A and ZK-4 may be pictured as having been assembled from primary (TO,) (where T is Si4- or A13') and secondary building units (cubes, D6, etc.).

TABLE I1 Classification of Some Well-Known Zeolites According to Secondary Building Units

Number of tetrahedral atoms in main Aperture of channel

Classification Zeolite channel or cage or cage (A) D4R Types A and ZK-4 D6R Chabazite

Gmelinite Faujasite Types X and Y Type ZK5 Type L

S6R Erionite Offretite Levyne Mazzite Omega Losod

Dachiordite Ferrieri te

Silicalite

5-1 Mordenite

ZSM-5

8 8

12 12 12 8

12 8

12

12 12 6

12 10 10 10 10

8

4.2 3.7 x 4.2

7.0 7.4 7.4 3.9 7.1

3.6 x 5.2 6.9

3.2 x 5.1 7.4 7.4 2.2

6.7 x 7.0 3.7 x 6.7 4.3 x 5.5 5.4 x 5.6 5.2 x 5.8

5

6 J . M. THOMAS AND C. R. A. CATLOW

tended to encompass many other tetrahedrally (T) bonded atoms, besides Si and A]. Evidently there are numerous open-framework structures, of stoichiometry TO2, all made up of corner-sharing tetrahedra, which can, in practice, be formed. As well as preparing many pure, crystalline silica variants of aluminosilicate zeolites (e.g., faujasitic silica-see Fig. 12) it is established that materials such as AIPO,, GaPO,, and FePO,, as well as those containing three or more different elements as tenants of tetrahedral sites, can be prepared (20-22). In such open structures it is possible for many of the heteroatoms (notably Fe, Cr, Ti, V, and Zn) of the zeolite to have six- or five- as well as the customary four-coordinated sites.

When only powdered samples of synthetic or naturally occurring zeolites are available, how are we to

1. Determine the atomic structure and hence identify the secondary building units?

2. Assess the distribution of the Si and A1 among the tetrahedral sites? 3. Characterize the nature of intergrowths and defects within a given

structure or between two or more related ones? 4. Pinpoint the positions of sorbed species and/or the exchangeable

cations? 5 . Identify and characterize the catalytic sites in the zeolite?

Fortunately, thanks to the relatively recent arrival of many important new structural techniques these questions can be answered, in some cases with a degree of precision that rivals, if not exceeds, that achievable with conventional, single-crystal X-ray methods.

With clays and their pillared variants (see Figs. 3 and 4 and Table 111) much progress in structural elucidation has been accomplished using the techniques that have also found value for polycrystalline zeolites, especialbj spectroscopic ones utilizing multinuclear, solid-state NMR (see below). Greater insight has also been achieved as a result of the methods now available for the laboratory syntheses of ultrapure specimens of clay, which are free from the obscuring influences of paramagnetic and other impurities. Because, in general, sheet silicates lack well-developed three-dimen- sional order, the newer techniques of neutron-powder profile and X-ray powder profile analysis using the Rietveld procedures (see Section 1I.A) is inapplicable. X-ray-diffraction and neutron-diffraction studies are nevertheless valuable for structural elucidation of clays and their intercalates, especially when supplemented by FI'IR and NMR spectroscopic analysis.

Of importance, in view of the recent febrile growth of computer science in the whole domain of silicate and aluminosilicate structural chemistry,


LAI, Mg

40.20H LSI 60

I Solvated exchangeable cations

do01

6 0 451

&O, ZOH LAI, Mg

40,20H

GSI

60

0 2 4 6 8 l O A Illlllllljl

Figure 3. Schematic illustration of the structure of montmorillonite.

are computer modeling techniques. Zeolites, in particular, have been em- inently amenable to this approach (see Sections 1I.A and III.A), but so also have the pyroxenoid silicates (general formula MSi03) consisting of corner-linked Si0:- tetrahedra. Such solids, such as wallastonite (CaSiO,), because they exhibit high-temperature stability, possess attractive properties as catalyst supports. In this regard, the aluminosilicate cordierite

- - -

Contact solution

M "+ M"' - - - -

M"' M"'

- - - Figure 4. converted into its pillared form.

Illustration of how a clay mineral (typically montmorillonite or beidellite) is


TABLE 111 Idealized Formulas for Some Selected Clays

Clay Formula

Pyrophillite" (AlJ"'(Sis)'"0&OH),

Montmorillonite" M::.nHzO(Al,-.M~)mSis)"'O,( OH),

Beideltite" M,";'.aHzO(Al,)m(Si8 -,Al,)"'O&OH),

Talcb (M&)m(Si8)e'Om(OH)4

Hectoriteb M;LaH,O(Mg, .,Li,)"Si8)"'Om(OH),

Saponite M,"LaH,O( MgS)""Si, -xAlx)'"OB( OH),

"In these clays, two-thirds of the available octahedral sites are occupied. These are termed dioctahedral.

these clays, all (three-thirds) of the octahedral sites are occupied, and they are termed trioctahedral.

(23), idealized formula of the synthetic form Mg2Si5Al4OI8, consisting of both rings and chains of linked tetrahedra, is especially important. It is the support material favored for most automobile exhaust catalysts, the active components being finely dispersed rhodium and platinum. Computer-modeling promises to disentangle the fundamental factors governing the nature of the Si, A1 ordering in this material, the ordering itself having been established by %i solid-state NMR (23).

To answer the questions posed previously we invoke the combined use of powerful techniques, some of which have not been widely deployed hitherto by inorganic chemists. We begin (Section 11) by outlining the essential features of these techniques before proceeding, in Section 111, to consider some specific examples. Finally, we briefly assess future prospects in this area of inorganic solid-state chemistry.

II. FOUR DISTINCI APPROACHES TO THE STRUCTURE OF ALUMINOSILICATES

When single-crystal X-ray crystallographic methods are inapplicable, one or more alternative techniques are employed. Provided the material under study is monophasic and well-ordered, powdered samples can be solved structurally by means of neutron scattering, which yields quantitative information comparable in quality to that obtained from X-ray crystallog- raphy. Solid-state NMR spectroscopy is another powerful technique, but it is at its best when used to determine the local environment of certain


atoms and local ordering preferences. Seldom does it provide quantitative data pertaining to bond length and bond angles. The third approach is high-resolution electron microscopy (HREM) and its allied procedures of electron diffraction and electron-induced X-ray emission, both of which greatly assist in the identification and characterization of crystallographic phases. The fourth approach is computational, involving, as outlined previously, simulation and calculation of likely structures. Many comprehen- sive reviews deal with the principles of these four approaches; only techniques for which the relevant details are less readily available will be discussed at some length here with the background principles.

A. Neutron Diffraction and Developments in X-Ray Diffraction

If aluminosilicate specimens are not of adequate dimension to be suitable for X-ray single-crystal diffractometry, they are hardly likely to be suitable for neutron diffraction single-crystal methods. But because of the pioneer- ing work of Rietveld (24), who capitalized on the fact that the peaks of elastically scattered neutrons can be represented by Gaussians or other well-defined functions, some progress can still be made. This means that, provided a model structure is available, the neutron-powder profile can be refined, to yield structural parameters. The procedure (25) is to arrive at a final refinement Rp, for the profile of the ne.utron powder diffraction pattern, in which all the peaks are of Gaussian shape:

where yi(obs) and yi(calc) are the observed and calculated intensities, respectively, at the ith position on the profile, and c is the scale factor.

With the dramatic improvements in neutron fluxes [available at IPNS, Chicago, IL, Grenoble, and the SNS (spallation source) at the Rutherford- Appleton Laboratory], many dramatic new developments can be expected from this technique. Not only is it the case that neutron diffraction experiments can be made over a wide range of temperatures and atmospheres (coupled with the sensitivity they provide towards light atoms such as hydrogen), but very precise time-of-flight measurements can be carried out. A recent study (26) (on a nonzeolitic material) shows the great potential of this approach, in that an ab initio structure determination has been demonstrated, thereby indicating that such determinations should ultimately become routinely possible for zeolites and other powdered solids.

10 J. M. THOMAS AND C . R. A. CATLOW

Because improvements in neutron fluxes have also been matched by comparable improvements in X-ray fluxes (from synchrotron sources) and because, in principle, there are many similarities between recording neutron-powder and X-ray-powder made in structure determination by X-ray diffraction. To be sure, conventional-powder X-ray diffraction is used an- alytically in distinguishing different zeolites. There is also a tradition (27) of quantitatively utilizing the peak intensities of powder X-ray pattern in the solution and refinement of zeolite structures by invoking additional, indirect items of evidence and plausible model structure. But the approach based on X-ray peak integration is now being superseded by full profile (X-ray) refinement (28, 29). This least-squares curve fitting procedure can be used either in deconvolution of the diffraction pattern or in optimizing directly the parameters that describe an approximate structural model. By supplementing the observed diffraction data with reasonable constraints that can be imposed on the (desired) structure, it is possible to refine quite complex zeolitic structures. ZSM-23, which is described in Section 1II.D falls into this category.

It is to be noted that, spurred by the success of neutron profile methods, corresponding X-ray powder methods have been dramatically resuscitated. It is now possible with ordinary laboratory X-ray sources to track the delicate structural changes that active zeolitic catalysts undergo during the course of high-temperature pretreatment or during actual use as an active catalyst (30).

Synchrotron radiation is potentially of very great value for the structural elucidation of zeolitic catalysts, as the recent work of Newsam (31, 32) elegantly reveals. Such radiation is very intense, polarized, sharply focused, and continuous over a wide range of wavelengths. In view of the “white” nature of the radiation experiments based on variations in scattering con- tract become possible with anomalous scattering. Moreover, because of their high intensity, synchrotron radiation sources make so-called energy- dispersive determinations of powder diffraction patterns a feasible prop- osition, particularly for time-resolved studied or experiments with samples under a controlled atmosphere. High intensity, as well as intrinsic resolution, confers extra advantages on this mode of recording X-ray powder diffractograms.

Furthermore, the brightness of the synchrotron source also enables conventional single-crystal diffraction measurements of very small crystals. Newsam (31) and his co-workers were able to determine the details of the framework structure of a minute specimen (1 pm3 in size) of a cancrinite crystal in this way. There is another important practical feature, associated with the use of synchrotron sources, that is likely to assume increasing importance in future years, namely, the great ease (in principle)-com-

THE STRUCIWRE OF ALUMINOSILICATE CATALYSTS 11

pared with rotating anode sources-with which position-sensitive detectors (PSDs) can be employed to dramatically reduce data collection times. It is known that some zeolites, under the influence of X radiation, undergo induced structural changes that perturb the equilibrium or dynamic positions of exchangeable ions. Use of PSD’s should greatly assist in amelio- rating this problem.

Finally, it has become (33-35) almost routine to determine the precise location of organic species and others (including water) accommodated within the intracrystallite cavities of a zeolite, as we shall illustrate in Section 111.

B. Solid-state NMR

NMR spectra cannot normally be measured in solids the same way that they are routinely obtained from liquids and solutions. The reason for this is the existence of net anisotropic interactions which, in the liquid state, are averaged by the rapid thermal motion of molecules. This is generally not the case in the solid state; although certain solids have sufficient molecular motion for NMR spectra to be obtainable without resorting to special techniques-a situation which is met when intercalated species in clays or sorbed species in zeolites execute a fair degree of thermal motion (36)-in the overwhelming majority of solids, and for every zeolite, there is little internal motion of the framework atoms. Conventional NMR for solids consequently yields broad signals up to 100 kHz wide, which conceals information of interest to the chemist. High-resolution spectra-where magnetically nonequivalent nuclei of the same spin species are resolved as individual peaks-of solids can be obtained only when the anisotropic interactions giving rise to line broadening are substantially reduced. The predominant interactions are dipolar, chemical shift anisotropy, and quadrupolar. When a solid sample is rapidly spun about an axis set at the “magic angle” with respect to the magnetic field, most of the broadening influences disappear and extremely sharp lines may be produced (37-39). The so- called magic angle (54’44’) technique yields the sharpest lines possible when the magnetically active nuclei in the sample are of spin 4. Samples with quadrupolar nuclei yield lines that have residual width at the magic angle, but this width is inversely proportional to the strength of the applied magnetic field. Solid-state NMR also suffers from two other intrinsic disadvantages: low abundance and/or sensitivity of the observed nucleus, and long spin-lattice relaxation times. Both these disadvantages can be over- come using a double-resonance technique known as cross-polarization (40). Full details of the principles of high-resolution solid-state NMR as it applies to solids of catalytic interest have been given elsewhere (39, 41-45). Al-


though many nuclei besides ?3, 27Al and ’H have been used as probes of zeolitic structure (e.g., 13C, 14N and 15N, 31P, 23Na, 1 7 0 , ’Li, and 129Xe) most of our knowledge derived from solid-state NMR of zeolites has come from experiments which have used these three nuclei either alone or in concert.

C. High-Resolution Electron Microscopy (HREM)

The principles of this technique have been adequately discussed elsewhere (46-51). The merit of HREM is that it can yield structural information, in real space and at the subnanometer level, about materials that are not amenable to structural determination by X-ray crystallographic and other conventional techniques. Using microscopes possessing lenses that have the lowest acceptable coefficients of spherical aberration, as well as lens pole-pieces that offer adequate scope for generous sample tilting about two orthogonal axes and adequate space for the insertion of detectors for emitted X rays or secondary electrons, a series of high-resolution images is recorded as a function of sample thickness and also as a function of lens defocus. These measurements are facilitated either by examining a series of thin samples (of differing thicknesses) or by selecting tapered or wedge- shaped specimens. The faithfulness of the images so recorded is assessed by comparing observed and calculated intensity distributions in two-di- mensional projections of the structure. (The electron-optical theory re- quired for such computation is reckoned to be on a secure footing). It follows, therefore, that some rudimentary knowledge of the unknown structure is a prerequisite. Structure can be refined by iterative procedures in which computation of image is carried out after each successive alteration of the atomic coordinates until, ultimately, the observed and calculated images match. For zeolite catalysts, because of their tendency to lose structural integrity under electron irradiation, it is not generally possible to “solve” structures of microcrystalline material in the same manner as it has proved possible for the much more beam-resistant bismuth-molybdate- based, selective oxidation catalysts (52, 53). The best advantage of HREM for studying zeolitic catalysts become clear when intergrowth structures, and structural defects which affect the performance of the catalysts are probed. Such intergrowths govern, in a critical fashion, the precise performance of a particular catalyst. The product distribution of aromatic and naphthenic hydrocarbons generated, for example, in the so-called methanol-to-gasoline conversion over ZSM-5 catalysts is much affected by the degree to which the structurally related ZSM-11 (see next section) is in- tergrown with the ZSM-5.


D. Computer-Modeling Techniques

As noted earlier, the availability of high-speed computers has led to the development of new techniques in solid-state chemistry. Indeed, during the last decade computer-modeling techniques have developed such that they now constitute a reliable and routine procedure for investigating the properties of perfect and defective materials. Reviews of both the meth- odology and applications of the techniques are given in Refs. 54 to 58. In this article we consider the application of the methods to silicate minerals and to aluminosilicate catalysts. The usefulness of modeling techniques in solid-state chemistry is considerable; furthermore, the results obtained to date are encouraging, although a number of problems remain, principally concerning the correct choice of interatomic potentials. This particular problem is considered after we briefly outline the general nature both of the modeling methods and their achievements. Later sections describe recent applications to framework silicates including zeolites.

1. Nature and Scope of Solid-state Simulations

The basis of the simulation technique is the specification of an interatomic potential model for the system, that is, an analytical (or possibly numerical) description of the energy as a function of atomic coordinates. For polar materials the model must include Coulomb energies, short-range terms, and ionic polarization.

Coulomb Energies. To evaluate these terms charges must be assigned to all atoms. In most studies reported to date, the fully ionic model has been used, that is, integral charges have been assigned. But for silicates there is now strong evidence that improved performance of the potentials can be obtained by using partial charges; we return to this point below.

Short-Range Energies. Included under this heading are both the re- pulsive forces which arise from overlap of atomic charge clouds and the attractive forces originating from dispersion and covalence. Generally, these are described by two-body, central-force models, expressed as a simple analytical function, of which the most widely used is the potential:

u(r) = A exp(-r/p) - C r 6 ,

where r is the internuclear distance and A , p, and C are constants. For more covalent materials, such as aluminosilicates, it would be expected


that such potentials would be inadequate because of the presence of angular dependent forces; indeed, it is known that even for ionic materials, many- body terms may be important (59, 60). Nevertheless two-body potential models have enjoyed some success in modeling silicates, although it is now clear that many-body terms are needed for accurate results.

Ionic Polarization. It is essential to include a description of the elec- tronic polarization of ions in studies of energies of defects, although these terms are less important in simulation of purely structural properties. To date, greater success has been achieved by shell-model descriptions of polarizability, in which the development of an ionic dipole is described in terms of the displacement of a massless shell, representing the valence- shell electrons, relative to a core in which all the mass is concentrated, representing the nucleus and the core electrons. Despite its crudity, the model has proved successful in describing properties of perfect and defective ionic materials. A more detailed appraisal is given in Chapter 11 of Ref. 54.

Potential models may be parameterized by empirical procedures, that is, by adjusting parameters, in a least-square-fitting routine, until the best agreement between calculated and experimental properties is achieved. This approach was used in developing the potentials for silicates discussed in Section 111. Alternatively, theoretical procedures based on electron gas or ab initio Hartree-Fock methods may be used. Details are given in the papers of Mackrodt and co-workers (61-63) and again in Chapter 11 of Ref. 54. Having developed a potential model, we may proceed to calculation of crystal structures, details of various crystalline defects, the surface structure of a solid, as well as undertaking molecular dynamics simulations.

Calculation of Crystal Structures. This process entails predicting the minimum energy configuration (i.e., cell dimensions and unit-cell coordinates) of a crystal structure. It is achieved by coupling lattice energy calculations-which themselves rest upon the use of efficient and exact summation techniques-with minimization procedures, based where possible on Newton methods, but employing conjugate gradient techniques for large complex structures. Automated computer codes PLUTO (64) and METAPOCS (65) are available for these calculations. They have various uses when applied to structures that are accurately known from diffraction studies; comparison of calculated and experimental structures is a good way of testing potential models; indeed, structure is commonly used as a source of data in the empirical parameterization of potential. More exciting

THE STRUCTLJRE OF ALUMINOSILICATE CATALYSTS 15

applications, however, concern first, the refinement of structures for which only approximate information is available from, for example, X-ray powder diffraction; second, the prediction of the effects on structure of temperature and pressure, a topic of obvious geophysical relevance; and, third, the study of new or hypothetical structures.

The reliability of energy minimization studies of crystal structures has been demonstrated by studies of Ti02 and titanates (65) and by the simulation of silicates described later in this article. Indeed, we believe that there is a major role for these techniques in the study of complex structures such as zeolites (66), especially those that have not yet proved amenable to experimental studies, and in the simulation of minerals formed at high pressure.

2. Calculation of the Properties of Crystals

Since the simulation codes calculate first and second derivatives of the lattice energy with respect to ionic coordinates, quantitative values can be obtained for other properties that depend upon these derivatives. These include elastic, dielectric, and piezoelectric constants as well as phonon dispersion curves. Again, such calculations are of value in testing and parameterizing potentials, but they may also have a predictive value concerning the effects of pressure, temperature, and chemical alterations in the solid. Moreover, the techniques could be used in screening classes of solid possessing desirable properties, such as high values of the static dielectric constant.

The success of interatomic potentials of the type described earlier in reproducing crystal properties of a wide range of properties of oxide and halide crystals is apparent from the results reviewed in Chapter 11 of Ref. 54. Application to selected aluminosilicates is reviewed in Section 111.

3. Calculation of Crystalline Defects

The calculation of the energies of formation, interaction, and migration of point defects has arguably been the most extensive and successful use of solid-state simulation studies carried out thus far. Moreover, calculations of defects have demonstrated the quantitative reliability of these techniques. There is already an extensive literature pertaining to calculations on halide and oxide materials, reviewed in Ref. 67 and in Chapter 14 of Ref. 54.

The techniques used in calculations of defects resemble those in the


Figure 5. Calculations pertaining to defects in solids are generally carried out using the so- called two-region strategy (54) (see text).

“perfect lattice” simulations regarding summation procedures. Calcula- tions of defects, however, introduce one essentially new feature, namely the need to include a detailed treatment of lattice relaxation about defects, which considerably perturb their environment. The two-region strategy shown in Fig. 5 is used to handle this problem. In concept, the procedure is simple; the defect and the surrounding region of the crystal (region I, containing typically 100-300 atoms) is treated explicitly, that is, coordinates of all ions are adjusted until the minimum energy configuration is generated. Again, Newton minimization procedures are found to be the most efficient (68). The response of the more distant regions of the crystal (region 11), where the defect forces are weaker, may in contrast be handled by pseudocontinuum methods. A favored procedure is that developed by Mott and Littleton (59), who describe the response of region I1 entirely as a dielectric response to the effective charge of the defect. Thus in dielectrically isotropic materials the polarization P(r ) at a point relative to a defect of charge q at the origin is given by

P = (qr(1 - e-’))/4nr3 (3)

where E is the static dielectric constant of the material. For dielectrically anisotropic materials more complex formulas (see Chapter 1 of Ref. 54)


are used for the polarization P, which is divided into atomistic components according to the type of potential model used.

In practice, it is found necessary to include an interface region (region IIa) between regions I and 11. Details are given in Ref. 54 (Chapter l), 69, and 70, which also describe the mathematical development of the theory of defect calculations. Automated computer codes are available for the calculations, of which the most notable are the HADES I11 (69) and CASCADE (70) codes. In addition to calculations of the energies of simple point defect, attempts have been made to study the complex modes of defect aggregation in nonstoichiometric oxides such as Fe, -,O (71), U02+, (72, 73), and TiO2_, (74, 75). The calculations have proved to be a valuable aid (58) in elucidating the complex and varied modes of defect aggregation in these materials.

4. Calculation of the Structure of Surfaces

This field, which has been developed mainly by Tasker (75, 76) and co- workers at AERE Harwell, and Mackrodt and his group (77-79), at ICI New Science Group, is one of the most active and exciting areas of progress in current simulation studies. The calculations are of two types: perfect structure simulations, which study the distortions within the surface region using energy minimization techniques, and study of surface defects, which are again based on a two-region strategy, but in this case using a hemi- spherical region I. The former class of calculation has been particularly useful in amplifying our knowledge of so-called surface rumpling effects (79, a mode of distortion based on differential displacements of different ions perpendicular to the surface. The most important application of the latter calculations has been in the study of surface segregation of defects and impurities (78, 79, SO), and indeed computational procedures seem to be the most effective way of studying this important phenomenon.

There have to date been no extensive studies of surface properties of aluminosilicates, but developments in this field are expected in the near future.

5. Molecular Dynamics Simulations

All the simulation studies discussed previously are based on a static picture, that is, no explicit account is taken of thermal motions. For many applications such an approach is quite acceptable. But in studies of high- temperature materials and/or materials in which there is exceptionally high

I8 J . M. THOMAS A N D C. R. A. CATLOW

atomic mobility (i.e., superionic or fast-ion conductors) it may become necessary to include thermal motions explicitly which is achieved by using the molecular dynamics (MD) technique, whereby kinetic energy is included explicitly in the simulation. The basis of the method, which has been used extensively in studies of liquids, is the specification of an ensemble of particles to which periodic boundary condition are applied and to which coordinates and velocities are assigned. The time evolution of the ensemble, normally a super cell in solid-state studies, is then followed by an iterative numerical solution of the equations of motion of the system. This requires the specification of a “time-step” Zit after each application of which coordinates and velocities are “updated”. In the case of the velocities, calculation of forces are needed using the specified interatomic potentials. The 6t necessarily has to be shorter than any characteristic process in the system (e.g., the atomic vibrational period); values of to

Molecular dynamics techniques have been used in the study of molten ionic materials (81). Applications to crystalline solids have concerned mainly superionic materials, such as CaF, (82-83). There are a number of potential applications to aluminosilicates, such as the study of the high-pressure perovskite structure phase of MgSi03, and to zeolite catalysts used at high temperatures.

sec are normally used in solid-state studies.

6. Interatomic Potentials

Models based on Formal Charges. Models of this nature are clearly open to criticism as covalence is known to be appreciable in silicates and aluminosilicates. Nevertheless, the ionic model forms the basis of reasonable interatomic potentials in several solids for which the model provides an inaccurate description of the electron density distribution. A more detailed discussion of this problem is given by Catlow and Stoneham (84). For this reason it has been worthwhile to develop potential models on the basis of the full, formal charges (Si4+ and 02-). In the brief review that follows we concentrate on our recent studies, although we note that there have been other studies of crystalline (85) and amorphous (86) silicates using ionic potentials.

Parker and co-workers (87, 88) explored the use of conventional Born- model rigid-ion potentials with two-body short-range forces described in Eq. 2. Their criterion for success was the reproduction of structural properties. Structures in reasonable agreement with experimental work were -

THE STRUCTURE OF ALUMINOSILICAE CATALYSTS 19

obtained for a number of ortho and meta silicates (e.g., MgzSiO,, MgSi03, CaSiO,, Al2Be3Si6OIR), although in some cases bond lengths were in error by approximately 0.1 A. Further discussion, especially of the modeling of MgZSiO4, is given below, where the effect of reducing the effective charge is also considered.

When two-body, ionic model potentials were used in the study of other classes of mineral, it was found that they failed badly. Studies of SiOz (89, 90) and of zeolites (91) revealed that such models did not yield stable structures. The problem does not seem to be associated with the use of formal ionic charges; rather it seems to be a failure of the two-body models which, while providing an acceptable description of the more closely packed silicates, cannot be used for the open framework structures. It is gratifying, however, that a relatively simple extension of the model seems to have a dramatic effect on its performance. Thus Sanders et al. (90) investigated the effect of including explicit bond-bending terms of the type

where E ( 8 ) is the bond bending energy, 8 is the 0-Si-0 bond angle and 8, is the tetrahedral angle; k, is a constant. The term confers a degree of “tetrahedrality” to the Si04 groups. Shell-model potentials in which such terms were included in additional to conventional two-body, short- range terms were very successful in reproducing not only structural, but elastic, dielectric, and lattice dynamic properties of quartz. The potentials reproduced well the structures of other polymorphs of Si02, namely cris- tobalite, tridymite, and coesite (90, 91). Further application of bond-bending potentials are presented in Section 111.

Partially Ionic Models. Despite the success enjoyed by ionic potentials described previously, it is well known that the bonding in silicates is not fully ionic. Consequently, in addition to fully ionic potential models for silicates a variety of more complex potential forms have been developed, as in the partially ionic model of Price and Parker (92). Here the ionic charges are allowed to be nonintegral, and the effect of covalent bonding between Si and 0 is described explicitly by the addition of a Morse potential, having the form

V ( r ) = D[1 - exp(p(r - re)}]’ (5 )

in which re is the equilibrium Si-0 bond length, and D and p are adjustable parameters.


Price and Parker found that the partially ionic model was generally most successful in describing the MgzSiO, system, it was able not only to reproduce the zero-pressure structural and physical properties of forsterite and the spinel polymorph of MgZSiO4, ringwoodite, but also to reproduce their pressure dependence. The partially ionic model developed by Price and Parker possessed fractional ionic charges for Mg, Si, and 0 comparable with those inferred from detailed electron density studies on forsterite. The potential reproduces the forsterite and ringwoodite cell volumes to 3 and 0.3%, respectively, and the predicted Si-0 and Mg-0 bond lengths have a root-mean-square error, when compared with observed values of only 0.004 and 0.025 A, respectively. In addition, this potential predicts the elastic constants of forsterite and ringwoodite to within 15 and 20% of the measured values.

In contrast, although the empirically derived fully ionic potentials of Miyamoto and Takida (125) produce excellent predictions of the structures of orthosilicates, they fail badly to simulate the elastic behavior of these phases, predicting elastic constants which are too large by a factor of two. We also note that the fully ionic potentials investigated by Price and Parker (92) and by Parker et al. (88) were only moderately successful in predicting the elastic behavior of forsterite and ringwoodite (with root-mean-square errors of approximately 50%), and gave means errors in the Si-0 bond lengths of approximately 0.165 A. In one respect, however, the latter potential was superior to the other two forms in that it satisfactorily reproduced the energetics of the magnesium orthosilicate system with the calculated lattice energy difference between forsterite and ringwoodite in good agreement with that inferred from thermochemistry (92). Therefore, the partially ionic, and to a lesser extent the fully ionic, transferrable potential, models orthosilicates well, and they can be used with some confidence to investigate some aspects of the physical behavior of these phases, such as diffusion, as discussed later. However, as we have seen, the usefulness of these potentials is limited by the fact that they ignore the directionality of the Si-0 bond. In Section 1II.C we will discuss the use of bond-bending potentials in modeling zeolites.

111. SOLVING SPECIFIC STRUCTURAL PROBLEMS

Armed with the information summarized in Section 11, we can now illustrate how certain specific problems in the structural chemistry of aluminosilicate, powdered catalysts have been solved. For convenience these problems are treated with reference to the main types of technique (among those outlined in Section 11) that have been employed.

progress in inorganic chemistry€¦ · spin-state transitions ..... 527 by edgar konig institut...

Documents