The local structure of aluminium sites in zeolites

The long range ordering and thus the average structure of crystalline zeolites can be determined by various diffraction and spectroscopic techniques.

It has, however, proved difficult to establish the local structure surrounding these aluminium sites by diffraction methods.

The most useful information has come from theoretical studies (M. Brandle et al., J. Chem. Phys., 1998, 109, 10379; U. Eichler et al., J. Phys. Chem. B, 1997, 101, 10035) which suggest that the Al–O distance associated with the proton is longer than other Al–O interatomic distances.

Employing in situ X-ray absorption fine structure spectroscopy (EXAFS) of the aluminium edge at 1565.6 eV, we report individual bond lengths angles for the local aluminium environment of neutral and acidic zeolites.

For two acidic zeolites we find that there is indeed one Al–O distance that is significantly longer than those in a neutral material.

We also show that for the average T-atom positions our EXAFS results are consistent with X-ray diffraction measurements, to an accuracy of ca. 0.01 Å.

Changes in bond angles show how the zeolite structure distorts to accommodate Brønsted acidity.


Structural studies of zeolites indicate that they are highly ordered materials with pore networks in up to three dimensions.

Yet there is local disorder in these frameworks, that appears wherever an aluminium atom is present.

This disorder is not evident from the results of techniques such as electron microscopy or X-ray diffraction.

Yet these local disorders have immense chemical significance, and so need to be understood.

In particular, the structural changes associated with the introduction of Brønsted acidity are of interest.

One reason for the introduction of local disorder with aluminium is its generally greater size, compared to the silicon atom.

In quartz (SiO2) and in aluminium-free ZSM-5 and faujasite, the Si–O interatomic distances are very constant at ca. 1.605 ± 0.002 Å.3,4

In tetrahedral environments such as berlinite (AlPO4) and in the crystalline aluminophosphates the Al–O interatomic distance is longer, in the range 1.70–1.80 Å.5

It is therefore to be expected that Al–O distances in zeolites are longer than those found when silicon occupies a similar site.

Indeed it is well-known that increasing the aluminium content of faujasite increases the unit cell size.6

For each aluminium zeolite framework atom, there is a need for an additional positive charge, provided by charge balancing cations, which also introduce local disorder.

The location of the cation depends on a number of factors, including its size and charge, and the state of hydration of the zeolite.

The cation is normally considered to be part of the local structure near to aluminium and a proton is typically represented as Al–OH–Si.

Good theoretical calculations suggest that the Al–O distance in Al–OH is longer again than Al–O–Si distances not associated with the acid site.2

Structural determination methods provide little detailed information on the local environment surrounding aluminium in zeolites.

Aluminium and silicon are very close in X-ray scattering power, so XRD gives only the weighted average of the Si–O and Al–O interatomic distances present at each distinct site.

Some attempts to probe structural variations have been made by MASNMR spectroscopy.

In faujasite (FAU) the presence of more than one aluminium site has been suggested by 27Al MASNMR.7

27Al NMR has also been used to estimate the aluminium–hydrogen distance in Al–O(H) entities.8

More recently in a very interesting combination between MASNMR studies and ab initio calculations, Al–O bond lengths have been inferred.9

The most interesting and detailed picture of the likely local structure around aluminium in acidic zeolites, however, probably comes from theoretical studies.1

These suggest that the Al–O distance associated with the acidic proton is significantly longer than the other Al–O distances, by up to 0.2 Å.2,9

Extended X-ray absorption fine structure (EXAFS) of the aluminium K edge at ca. 1560 eV should be a fruitful way to probe the local structure around aluminium in zeolites.

Although the EXAFS information is less accurate than the averages from XRD, its ability to probe the local structure around a selected element is key, as is the (relative) ease of in situ measurement.

In an early X-ray absorption studies of the aluminium environment in zeolites, Koningsberger and Miller10 studied three hydrated faujasite samples, Na–Y, NH4–Y and H–Y.

In each case they concluded that each aluminium had four oxygen neighbours, with the four Al–O bonds of the same length, 1.62 Å for Na–Y, 1.64 Å for NH4–Y and a longer value of 1.70 Å for their acidic sample, H–Y.

Hydration is expected to lead to a more ordered tetrahedral site, so the observation of constant Al–O distances is not unexpected.

The Koningsberger group has also been very successful in extracting information from studies of the aluminium X-ray absorption near edge structure (XANES) in zeolites.11,12

XANES however does not directly yield interatomic distances.

In a preliminary study13 we have reported some structural data on H-ZSM-5 and Na-FAU.

Subsequent to this study, van Bokhoven and Prins have reported a preliminary study of HY and Na–Y and the interaction of the faujasite with hydrocarbons.14

The two studies are in general agreement.

In this work we have extended our studies to H-FAU and also include NMR and infrared evidence to attest to the integrity of our materials.

We have analysed our results using the multiple scattering capabilities of the ab initio code EXCURV98, which allows the Al–O–Si angles to be studied, and which also allows determination of second nearest neighbour distances—information that is not available from the empirical analysis approach of van Bokhoven and Prins.

We present results first for the non-acidic zeolite Na-FAU, followed by those for the weakly acidic H-FAU and then for the strongly acidic H-ZSM-5.

Importantly, we show that our SOXAFS results are consistent with X-ray diffraction studies of similar materials and we also consider how the variation in acidity between H-FAU and H-ZSM-5 is mirrored in their structures.

Results and discussion

Fig. 1a shows experimental and best fit calculated EXAFS spectra for Na-FAU (Si/Al ratio = 2.5).

The experimental measurements were made in vacuo and in situ, after the sample had been heated to 640 K for 1 h, to remove residual water and cooled to 423 K.

Eight spectra were collected over 6 h, and averaged.

The quality of fit obtained is very good, with the R factor of 40.9 largely reflecting the noise level of the experiment.

The parameters used in the calculation are listed in Table 1, and a representation of the local structure is given in Fig. 2.

Although the range of the spectrum is limited by the presence of the Si K edge at 1839 eV, we are able to draw statistically significant conclusions15 about aluminium–oxygen and aluminium–silicon nearest neighbour distances, and about the associated angles.

EXAFS indicates that each aluminium atom in our Na-FAU sample has exactly four oxygen nearest neighbours.

Further confirmation of this is provided by analysis of the XANES spectra, using the approach of van Bokhoven et al.,12 which shows the absence of detectable quantities of octahedral aluminium.

The Al–O interatomic distances in Na-FAU are all very similar, at 1.74 ± 0.01 Å.

Our fit also includes four Al–O–Si distances, with an average length of 3.12 Å.

Only a single crystallographically distinct T site, where T represents aluminium or silicon is available for the FAU framework.

Note however, that due to perturbation of the bridging protons with neighbouring oxygen atoms two distinct OH frequencies and NMR signals can be found.

These are attributed to OH groups pointing into supercages and β cages, respectively.16

Each T–O bond has a slightly different T–O distance, with an average of 1.74 Å, so agreement with our EXAFS conclusions is good.

Lastly we find that introduction of a sodium atom gives a significant improvement in the fit.

The angular information obtained shows that the tetrahedral environment of aluminium in Na-FAU is rather regular.

The average O–Al–O angle is 109.3°, compared to the ideal tetrahedral value of 109°.

However, as we will see, the standard deviation of the six O–Al–O angles in the tetrahedron provides a better measure of structural distortion.

The standard deviation of these angles for Na-FAU is small, at 6°.

We have also studied the acidic materials H-FAU and H-ZSM-5, again making in situ measurements after dehydration in vacuo, and collecting data over 11 h (H-FAU) and 18 h (H-ZSM-5).

The EXAFS spectra are shown in Figs. 1b and c, and the parameters used in the calculations are in Table 1.

Again there are exactly four oxygen nearest neighbours, and the XANES spectra show no evidence of the presence of octahedral aluminium.

For each zeolite, three similar but not identical Al–O distances are found, together with one Al–O distance which is significantly longer.

For H-ZSM5 the three Al–O distances are a little shorter than in Na-FAU, with an average length of 1.70 Å, while the longer distance is 1.98 ± 0.01 Å.

Although hydrogen is a poor scatterer of X-rays and so makes no contribution to the EXAFS spectra, we take our cue from the theorists, and associate this longer bond with the Al–O(H) distance.2,9

This elongated distance has also been suggested for Fe–silicalite the iron containing analogue to ZSM-.517

The aluminium–silicon distances in the EXAFS calculations are encouragingly close to the crystallographic values.

In contrast to the earlier results for hydrated samples,10,11 we observe significant differences in the different bonds of aluminium to the surrounding four oxygen atoms.

Mainly on the basis of NMR evidence, it was suggested that distortion of tetrahedral aluminium occurs in protonic zeolites upon dehydration.18

Both dehydrated Brønsted acidic zeolites (H-ZSM-5 and H-FAU) exhibited a single lengthened Al–O bond, well in line with the distortion suggested by Fajula et al.18,19

Interesting changes are also noted in the angles.

For H-ZSM-5 there is a decrease in the three (H)O–Al–O angles, from the average value of 109° to 97 ± 6° (standard deviation) and an increase in the other three O–Al–O angles, to 119 ± 6°.

These changes provide an insight into how the zeolite structure adapts to the longer Al–O(H) distance in the acidic materials.

Fig. 2 shows that the change is accommodated by movement both of the acidic oxygen and the aluminium atom.

The distance of the other oxygen atoms from aluminium changes little, but the aluminium atom moves much closer to the plane that they inhabit.

To test their significance, all of the above EXAFS derived conclusions have been subjected to detailed statistical analysis (see ESI).

We consider it important to find ways to test for consistency between the local structural information provided by EXAFS and the results obtained from X-ray diffraction.

This can be done by recognising that silicon–oxygen interatomic distances are almost invariant in different materials, as is to be expected for a strong covalent bond.

In α-quartz the average Si–O distance is 1.609 Å, in highly siliceous FAU it is 1.606 Å, and in silicalite it is 1.594 Å.

For comparison with EXAFS results, accurate to ±0.01 Å, we assume that the Si–O bond length is invariant on the introduction of aluminium, at 1.602 Å and 1.594 Å for FAU and ZSM-5, respectively.

We can then calculate that average T–O distance that we expect from an X-ray diffraction experiment from:(T–O)average/Å = [Si–Osilic.* Φ/(Φ + 1)] + Σ [0.25.di / (Φ + 1)],where Φ is the Si/Al ratio of the zeolite, Si–Osilic is the Si–O distance of the siliceous ZSM-5 and FAU, respectively and the summation is over the four Al–O distances, di, obtained from EXAFS.

Table 2 shows the results of applying this formula to a number of materials, and assuming that the Al–O distances stay constant as the Si/Al ratio changes.

The agreement between the X-ray diffraction results and the calculations is remarkably good, and suggests that the EXAFS results may be accepted with a high degree of confidence.

It is tempting to speculate that the increase in length of this bond may be related to acid strength.

Brønsted acid strength in zeolites is associated with the barrier for removal of the acidic proton.

In general, methods measure the ease of this transfer by the energy released upon interaction with bases.

This energy depends, besides other factors, on the charge density at the proton as well as the charge distribution and total structure associated with the oxygen bonded to the proton and the nearby aluminium atom.

The charge distribution in turn is influenced by structural parameters as well as the aluminium content of the zeolite.

What is very clear from our data is that the Al–O(H) bond length varies with the zeolite structure.

The shortest bond length (1.75 Å) of the elongated bond is found for Na-FAU and the longest for H-ZSM-5 (1.98 Å) H-FAU has an intermediate bond length of 1.87 Å.

Na-FAU shows at most very weak Lewis acidity, and the environment around each aluminium atom is an almost perfect tetrahedral.

It is generally accepted that H-ZSM-5 shows significantly higher acid strength than H-FAU.

Calorimetric measurements for ammonia adsorption show an adsorption energy constant with loading at ca. 150 kJ mol−1,20 whereas for H-FAU21 values in the range 110–130 kJ mol−1 are reported.

Increasing the Al–O bond length will result in variation of the charge distribution between the oxygen, proton and aluminium atoms which in turn is expected to influence the acidity of the protonic zeolite.

In addition the distortion of the tetrahedron increases for the H-ZSM-5 compared to H-FAU, resulting in the highest net energy gain by relaxation upon removal of the proton.

This effect is of course also influenced by the long range order, structure type and flexibility of the zeolite structure.

It is important to note that other properties such as the bond angles, and confinement effects are also expected to play a significant role in the acidity of zeolite materials.

For example, it was suggested that the interaction of bases with the framework oxygen play an important role for the interaction energies observed.22

Other structural parameters, sorbate–sorbate interactions and ordering also contribute to the interaction energies of e.g., alkanes23 with the solid acid.

Material and methods

The zeolites were obtained from Sűd Chemie, Munich (ZSM-5) and Zeolyst (FAU).

EXAFS experiments were performed on Station 3.4 at the Daresbury SRS, using a YB66 monochromator.

Thin disc samples were held in a cell designed by van der Eerden et al.,24 in which the sample could be heated to >500 °C.

Within the beamline hardware the cell achieved a vacuum of ca. 10−5 mbar.

FAU samples were heated at about 5 K min−1 in vacuum to 650 K, while ZSM-5 was heated to 725 K.

Samples were kept at the activation temperature for 1 h and cooled to 423 K.

Spectra were then collected for the appropriate time at 423 K.

EXAFS data analysis used the standard suite of Daresbury programs, including EXCURV98.

Statistical significances and error bars were determined by standard methods.25

Our data analysis approach and phase shifts were validated by studies of an AlPO4 material of known structure, in which the Al–O environment is a regular tetrahedron.

Since EXAFS averages over the whole range of various aluminium sites, it is very important to demonstrate the structural integrity of the zeolite samples studied.

In particular, if octahedral aluminium were present in the samples it would be difficult to separate the contributions of different sites to the EXAFS spectrum, and the structural conclusions could be compromised.

We have used two experimental methods to examine the extent to which the aluminium in our key H-ZSM-5 sample is in a tetrahedral environment, infrared spectroscopy and 27Al magic angle spinning nuclear magnetic resonance (MASNMR).

Fig. 3 shows an infrared spectrum of H-ZSM-5 after activation in a vacuum for an hour at 823 K.

Bands at 3716 cm−1 and at 3600 cm−1 are attributed to terminal SiOH and bridging OH bands, respectively.

Fig. 4 shows the same sample after adsorption of ammonia at 423 K and 10−2 mbar pressure.

N–H stretching vibration bands can be observed in the region 3350–2800 cm−1, characteristic of ammonia adsorbed on acidic zeolites.

More informative is the N–H deformation region, where there is an intense band at 1454 cm−1, attributed to the formation of an ammonium ion, by interaction of ammonia with Brønsted acid sites.26

The material shows very little Lewis acidity, as no significant band at ca. 1620 cm−1 is observed.

Lewis acidity in this material would be indicative of extra-framework aluminium, probably in an octahedral environment.

The absence of any additional band in the region of 3600–3700 cm−1 indicates the absence of extra-framework AlOOH species.

Fig. 5 shows the 27Al MASNMR spectrum of the same H-ZSM-5 zeolite material.

Aluminium MASNMR spectra of dehydrated zeolite H-ZSM-5 and H-FAU samples are known to exhibit quadrupolar line broadening to such an extent that the signal may become invisible.19

This artefact can be eliminated by hydration, so vapour water was adsorbed on our sample in order to ensure that all of the aluminium present was detected.27

The major band in Fig. 5 is observed at a chemical shift of 54.5 ppm, and is due to tetrahedral aluminium.

The band at 0.15 ppm is attributed to octahedral aluminium, but comprises <3% of the total intensity.

Fig. 6 shows the infrared spectra for H-FAU together with the spectrum of the same material in contact with 10−1 mbar ammonia.

The three bands observed above 3500 cm−1 are typical, and are assigned to terminal OH groups (3735 cm−1), high frequency bridging OH groups (3640 cm−1) and low frequency bridging OH groups (3534 cm−1).

On ammonia adsorption (Fig. 6) the main new band observed is at 1425 cm−1, typical of Brønsted acidity; the bands corresponding to bridging hydroxyl groups disappear, and that due to terminal OH groups is unperturbed.

There is no indication of Lewis acidity, which would result from the presence of octahedral alumina in our H-FAU sample.

Fig. 7 shows the 27Al MASNMR spectrum for H-FAU.

The large band at 58.3 ppm, corresponding to ca. 86% of the aluminium in our sample, is attributed to aluminium in a tetrahedral environment.

The band at −1 ppm is attributed to octahedral aluminium, but we believe that it results from a reversible transformation of part of the tetrahedral framework aluminium into an octahedral structure28 upon hydration of the zeolite.

In clean phases and in contact with ammonia, we do not observe any evidence for the formation of extra-framework phases.

We have included a detailed description and statistical analysis of the EXAFS fitting and analysis procedure as well as a colour copy of Fig. 2 as ESI.