Catalysis Today, 2 (1966) 621-631 Elsevier Science Publishers B.V., Amsterdam-Printed
621 in The Netherlands
ALUMINA-PILLARED CLAYS AND THEIR ADSORPTIVE PROPERTIES.
M.H. STACEY, ICI plc, C&P Group, P.O. Box 8, The Heath, Runcorn, WA7 4QE. (England)
ABSTRACT This work has explored the synthesis of alumina-pillaredmontmorillonite and Laponite. The importance of stoichiometryhas been examined. A zirconia-pillaredmontmorillonite is also reported. All these products are fully characterised by X-ray diffraction and by the adsorption of nitrogen, water, n-butane, methanol, and neo-pentane. The implications for the real microstructures of these materials are discussed.
INTRODUCTION Brindley and Sempels (ref. 1) first reported in 1977 that beidellite clay could be pillared by aluminium polycations. The solution used was a partly-neutralisedaluminium chloride in which the (OH)/Al ratio was greater than 2. Products have been commercially available for several years in which this ratio is 2.5 and Vaughan and Lussier (refs. 2,3) as well as Shabtai (ref. 4) simultaneously found that these solutions are effective in yielding pillared clays from all types of smectites. Despite their availability, the nature of the polycations in such solutions has long been the subject of speculation. Very recently new structural investigationshave probed their constitution. While it is now agreed that the All,+' cation originally described in Johansson's sulphate, occurs in solutions having a neutralisation ratio of 2.2, it is also clear that a higher molecular weight species supercedes it at higher degrees of neutralisation. Currently two main views exist of the nature of the higher Mwt species. Bottero and coworkers(ref. 5) propose that the Al,,'? cations form loose aggregates of linear or platy character. On the other hand Farthing and Akitt (refs. 6.7) 0920-5861/88/$03.85 0 1988Elsevier Science Publishers B.V.
622 propose that a larger cation forms containing Al,, units. Whichever view is more correct it is undoubtedly true that commercial aluminium chlorohydrates contain at least three different species. Free Al,,+' polycations can only be present in amounts less than 10%. Consequently when cosnnercialaluminium chlorohydratewas reported to yield pillared clays having only an 8.5-91 interlayer spacing, which is in accord with the incorporation of Al,, cations, this requires explanation since at most such species constitute a very minor fraction of the solution. Additionally if Farthing and Akitt are correct in maintaining that the main component of commercial aluminium chlorohydrate is an Al,, cation then larger interlayer spacing pillared-claysmight be synthesised. The work reported here then addresses the problem of the mechanism of alumina-pillaredclay synthesis and control of interlayer spacing in the products. The interlayer spacing has been assessed both by measurement of the basal plane repeat distance by X-ray diffraction and by determination of the molecular sieving capability.
EXPERIMENTAL Materials Wyoming montmorillonitewas obtained from Production Chemicals, Stockport. It was purified by conversion to the sodium form by suspension in 1M NaCl solution. The suspension was dialysed against mains water for 2-3 days and the clay suspension decanted. The material was not separated significantly from orystobalite and feldspar impurities by this treatment and was about 70% smectite in all cases. For one preparation a sample (Eccagum) supplied by English China Clays, St Austell, Cornwall was used. This sample contained a greater.amountof montmorillon~te than the Wyoming material and had a lower cation exchange capacity (ccc). In all cases the dilute suspension was treated with the polycations as described below. Laponite, the synthetic smectite made by Laportecref. 8). was used as received. It is a Hectorite-typeclay and its' analysis is given in Table 2. This does not correspond to the published typical data but is magnesium-rich. Commercial aluminium chlorohydrate (Hoecht Locron solid: 44% Al,O,) was dissolved in deionised water. Also a commercial cationic alumina sol was investigated (Bluonic A, Wesolite Co. Wilmington, Delaware, USA; 20% w/w Al,O, aqueous dispersion). For zirconium pillars, commercial zirconyl chloride (MagnesiumElektron Ltd) was diluted with deionised water to a strength of 0.2M.
Syntheses Pillared clays were made by adding a known amount of alwninium polycation or
623 sol to a dilute (1%) suspension of clay (previouslyconverted to Na' form) and stirring until the clay coagulated. The solution was centrifuged to recover the clay and then dialysed and washed to free it from sodium chloride and other soluble materials. It was then dried at 1lO'C and finally calcined in a muffle furnace at temperature selected in the range 200 to 5OO“C. Products were submitted for elemental analysis to determine the degree of aluminium cation incorporation and to X-ray diffraction to determine the basal spacing. Analyses The cation exchange equivalent (ccc) of the clays were determined as received by stirring an accurately-weighedamount of dry clay with 1M ammonium acetate for 4 hrs, filtering off the solid, and diluting the filtrate to a known volume before determining the ions Na+. Ca++ by atomic absorption. Results are quoted as meq/lOOg crude clay. Molecular sieving. Nitrogen adsorption isotherms were determined at 77X using a volumetric Digisorb analyser. Other gases were studied at the appropriate temperaturesby either the frontal chromatographicmethod (n-butane, neo-pentane) or by gravimetry (H,O, MeOH, benzene). The nitrogen adsorption isotherms were used to characterise the microstructure of the products in terms of a micropore contribution and a mesopore contribution. In all cases (except one noted in the text) the isotherms had a type H4 hysteresis loop together with large adsorption at low pressure (see Fig. 1). as-plots of the adsorption branch all had a knee at as=l.O-1.2. The slope of the line above the knee yielded the surface area in mesopores (SW) and the intercept of this line extrapolated back to the Y-axis yielded the micropore volume. The total pore volume was derived from the maximum uptake on the desorption branch at high relative pressure. Hence the mesopore volume (Vmesof was assessed as the difference between the total and micropore volumes. Since type H4 hysteresis loops are characteristicof slit-shaped pores the meso = 2v pore width was also calculation from the formula: D meso meso/sw RESULTS Syntheses Since montmorillonite already contains aluminium ions incorporated into the octahedral sites in the sheets, when it is converted to an alumina-pillared product the aluminium content is expected to increase. But, since the montmorillonitesused were also contaminated by crystobalite and feldspars, derivation of the alumina pillar fraction in the products from the analyses is complicated. Two possible assumptions can be made.
First it can be assumed that during
the pillaring process the clay fraction was not enriched. Consequently the silica analysis of raw material and products can be used to compute the degree
624 of alumina incorporation. Alternatively, it can be assumed that the Fe,O, content of the raw material is wholly contained in the montmorillonite fraction. The Al,O,/Fe,O, ratio of raw materials and products then allows the degree of alumina incorporation to be computed. By comparing the results of these calculationswith the ccc value of the raw clay an estimate of the charge associated with each Al atom in the pillars has been obtained. In the case of Laponite, which does not contain any aluminium initially, these considerations did not apply. Neither did the zirconia-pillaredmaterial present a problem. For Wyoming montmorillonitethe ccc was 110 meq/lOOg and
if the aluminium
chlorohydrate contains only A113+7, the stoichiometricratio would be 204 mmoles Al/lOOg clay. In this work both the reactants used and the ratio of al~ini~/clay
were varied. TABLE 1 shows the stoichiometricratios used to
make alumina-pillaredmontmorillonite and the analyses of the products and the inferred Al pillar content. Similar data are given in TABLE 2 for the zirconia-pillaredmontmorillonite (using ECC montmorillonitehaving a ccc of 60 meq/lOOg) and for alumina-pillaredLaponite (having a ccc of 76 meq/lOOg). TABLE
Synthesis of Alumina-PillaredClays (from Wyoming Montmorillonite)
Pillar source Al(sol)/Clay ~ole/lOOg
<___---_--______-_HoechtLocron_______-___> Bluonic A 0
Oxide/% (dry basis) SiO, 74.45 67.30 Al,03 16.05 25.03 Fe,C, 3.60 3.58 MgC 1.43 2.60 CaO 0.45 1.25 Na,O 3.62 0.12 E,C 0.39 0.12
67.80 25.50 3.73 1.30 1.09 0.30 0.27
Product Basis Alp/clay Fe mmoles/lOOg Si
67.30 25.30 3.53 1.15 0.96 1.24 0.41
65.56 28.22 3.49 1.19 0.33 0.64 0.57
60.62 34.25 3.62 0.88 0.01 0.28 0.36
57.00 37.81 3.53 1.29 0.15 0.11 0.11
The elemental analyses showed that the interlayer cations (Na and Ca) had been virtually wholly replaced by the Al polycations (or by Zr polycation). Indeed even at low stoichiometricratios of the reactants there is a marked tendency for the product to incorporate ca 200 meq of Al atoms/lOOg clay (but , the yield of pillared montmorillonite is then lower). However at high stoichiometricratios the alumina pillar content was almost doubled. These data allow a rough estimate of the effective charge carried by the pillar cations. We must assume that the clay sheet charge is neutralised accurately by the combination with the polycations (the validity of this will be discussed later)
625 so that if no enrichment of the clay fraction has occurred, then the total charge carried by the pillar species is given by the ccc. The effective charge is therefore the ratio of the ccc and Alp/clay in the Tables. When the synthesis ratio is <120 n-molesAl/lOOg clay the charge is ca 0.53/Alp but at high ratios this drops to ca O.Z/Alp; The Laponite sample gave a charge of 0.471Al . These are to be compared with that for the All,+' cation which is 0.54/A?
The zirconia pillars had a charge ratio of 0.24/Zr .
chloride solutions the only reported polymeric cation is Zr,?OH),+e, which has a charge of 2/Zr. TABLE 2 Synthesis of Zirconia-PillaredMontmorilloniteand Alumina-PillaredLaponite
Clay source Pillar source M(sol)/clay mmole/lOOg
<---- ECC MM ---> ZrOCl, 0
Oxide/% (dry basis) 69.16 SiO, 25.22 AlaO, 0.00 Pe,O, 1.23 MgO 0.60 CaO 3.67 Ha,0 0.12 K,O Li,O ZrO, Product Basis Mp/clay Si mmoles/lOClg
54.00 17.92 3.07 1.27 0.02 0.15 0.01
<---- Laponite ----> Hoecht Locron 240 0
62.50 0.71 0.13 33.41 0.09 3.38 0.00 0.81
62.51 11.64 0.06 24.95 0.06 0.30 0.00 0.47
The analytical data therefore indicate that there is a strong tendency of negatively-chargedclay sheets to combine with polycations. However there is an indication that the combining ratio is a function of the amount of reactant used. Effect of calcination Alwninium trihydroxides decompose between 200 and 300°C and aluminium chlorohydrate behaves similarly. The situation for a pillared clay is less clear. Presumably the initial pillaring species will be fully hydroxylated/ hydrated in solution; it may lose some water of hydration when the pillared clay is formed if the silicate ions at the clay sheet surface are able to co-ordinate directly to the Al cations in the polymeric species. The ssmple in column 5 of Table 1 was therefore calcined at temperatures of 200, 300, 400, and 500% to discover if the microstructure of the material was affected by dehydration of the pillaring species.
626 TABLE 3
Effect of Calcination on Alumina-pillaredMontmorillonite
18.7 16.0 17.9 17.5 16.9
50/vat 200/vat 300 400 500
0.124 0.118 0.088 0.077
21.5 23.0 22.8 21.1
"total ml/g 0.166 0.192 0.156 0.141
39 65 60 60
The results are in Table 3. There is a progressive drop in the micropore volume above 300°C as the materials are decomposed. This is paralleled by the basal spacing which simultaneouslydecreases. The 200 and 3OO'C materials are virtually identical in properties the only difference being in the mesopore volume which may be due to the use of vacuum conditions in the case of the 200°C sample. Because of the large drop on heating at 400°C the other materials made were only calcined at 200°C. Composition and Microporosity The materials in Tables 1 and 2 after calcination at 200°C were also examined by XBJJand nitrogen adsorption. At or below the stoichiometricratio, although the interlayer cations were replaced, the pillaring process did not give products showing clear XRD evidence of uniform pillars (see Table 4). To obtain material with a clear (001) reflection at 17-188,and with high nitrogen adsorption capacity it was necessary to utilise a >50% excess of the aluminium polycation sol relative to the stoichiometricratio of 204 mmoles/lOOg clay. There appears to be a critical composition at which the pillared clay structure is generated. At the stoichiometricratio only about half the final micropore volume is formed. TABLE 4 Basal Spacing and Nitrogen Adsorption on Alumina-pillaredMontmorillonite
Pillar source Alp/Clay mmole/lOOg
Basal spacing/A "micro ml/g SW ma/g V ml/e.
<_____________----BoschtLocron___________> Bluonic A 196
29 0.102 33
29 0.104 33
Some idea of secondary microstructure could also be gained from these data, The XRU peaks for the (001) basal planes were not very sharp. This could be either due to small crystal dimensions normal to these planes or due to variability in pillar heights. If all the broadening were due to small coherent stacking height of clay sheets then it would be characteristicof material with a crystallite size of ca lOOIL. This implies that typically about 5 layers of pillars and clay sheets are present. An alternative estimate of the thickness of a crystallite is provided by the nitrogen adsorption data. The wide pore surface area (SW) yields an estimate of plate crystal thickness from the formula 4/Sw.p where p is the density of the sheets. Using an estimated density of 1.6 g/ml (allowing for the microporosity caused by pillaring) gives a crystallite thickness of ca lOOO&. Since this is much larger than that estimated from the XEB line width it seems more likely that the broadening is due to variability in pillar heights. On this basis it seems that in all cases the primary microporous aggregates of sheets and pillars are about 1OOOllthick and form a secondary structure with slit-shaped mesopores about 30-4OILwide and having a surface area of ca 20-30 ma/g, TABLE 5 Characterisationof Zirconia-PillaredMontmorilloniteand Alumina-Pillared Laponite
Clay source Pillar source Mp/clay mmoles/lOOg
Basal spacing/A 'micro m1/g SW ma/g 'total ml/g Dmeso/&
MM ---> ZrOCl, 250
<---- Laponite ----> Hoecht Locron 0 242
The use of zirconia pillars with ECC montmorilloniteand of alumina pillars with Laponite gives the results in Table 5 and Fig 1. The zirconia-pillared material is very similar to the Wyoming montmorillonitepillared with alumina but the Laponite based material has very different characteristics.It seems that alumina-pillaredLaponite must have a different stacking mode since the isotherm type is very different and can be interpreted as being transitional between meso and microporous so that the contributionsof micropores cannot be separated from those of mesopores. The pores appear to have a continuous
o Alp/MM 555 A Alp/MM 280; 200'~
h( Pl50 2 "J 2100 (i 4:: 50
oAlp/MM 555 mmole/lOOg
0.4 0.6 P/PO
1 4 Diameter/g
Fig. 1 Typical N, Isotherms at 77K
Fig. 2. Molecular Sieve Performance
distribution of widths in the micro/mesa range with an average width of 18A. A clear (001) spacing is not detectable although there is general diffracted intensity in the lo-17A region which is in general agreement with the adsorption data. Molecular sieving Capabilities of selected samples The samples selected for detailed study were 280 and 550 mmoles/lOOg Alp/MM (the former calcined at 2OO'C and 500°C), Zrp/MM, and Alp/Laponite. The results are given in Table 6 and Fig 2 where the kinetic molecular diameters used are those given by Breck (ref. 9). In all cases the micropore volume is crudely assessed at low relative pressure (O.l
Adsorptive Liquid Density g/ml Diameter/A P/PO Alp/MM=280 ZOO~C 5oooc Alp/MM=555 2oooc Alp/Lap=240 2oooc Zrp/MM=250 2oooc
Hz0 1 2.65 0.75
N, 0.8081 3.64 0.2
C,H, neo-C,H,, 0.8765 0.6135 6.2 5.85 0.1
Volume adsorbed ml(liquid)/g 0.101 <0.04 0.122 0.053 <0.03 0.076
(1) External Sorption only
CH,OH n-C,H,, 0.7914 0.5788 3.9 4.3 0.3 0.2
There are two classes of materials. Those which do not allow entry of molecules >4.5A diameter (both 280 mmole/lOOg Alp/MM samples) and those which do allow free entry up to at least neopentane. In the case of the two 280 mmole/lOOg Alp/MM materials it seems that the only effect of higher temperature calcination is to reduce the pore volume accessible to those molecules which do penetrate. The cut-off dimension seems to be unaltered. For the other three materials it seems that, since the water uptake at 75% relative humidity is much greater than the nitrogen microporosity, there is a larger mesoporosity contribution. As has already been remarked the Alp/Laponite is unique in its' total capacity and in the fact that distinction cannot be made between the contributions of micro and mesoporosity. Where the data are available the sorption characteristicsof 550 mmoles/lOOg Alp/MM and Zrp/MM are essentially identical. Tables 4 and 5 show that both the XRD and nitrogen isotherms are also remarkably similar.
DISCUSSION In considering the above results the following findings require some explanation. First the effect of stoichiometricratio on pillar incorporation. Secondly the dependence of micropore volume on this ratio even though the basal spacing is constant. The effects of higher calcination temperature are also of concern. Lastly the different pore volumes available to different sized molecules amongst the various products (especiallythe large apparent spacing in the case of Alp/Laponite). It is best to attempt this by means of a structural model of a pillared clay. The analytical data support the idea that the pillaring polycations accurately neutralise the negatively-chargedclay sheets. The simplest model of a pillared clay envisages uniformly tall pillars separating the clay sheets with a regular repeat spacing determined by the charge balance requirement. In reality the organisation may possess a significant degree of randomness so that these dimensions may be quite variable. It is useful to discuss how closely the experimental data support or deviate from such an ideal model concept. The negative sheet charge arises from isomorphous substitution in the octahedral central layer of the sheets. If electrostatic attraction determines the positioning of the pillars on a sheet there is no way that a multicharged cation can achieve equally close approach to a similar number of evenly-separatednegative charges. In Wyoming montmorillonite there is about one negative charge on every four silicate rings. If the Mg++ ions are regularly arranged in the sheets, the associated negative charges would be on a rectangular lattice with a=10.4A and b=9A.
If the pillars are indeed the Al,,
ion, which is globular with about 9-10A diameter, and if such ions are located opposite one such negative charge, the other 8 surrounding charges in the sheet
630 would be at least 5A distant. Moreover to achieve the correct overall electroneutralityonly one such ion can be incorporated for every 2.8silicate rings. This implies pillar repeat spacings of approximately 27x20.8A in the basal plane. It seems rather unlikely that such a highly regular structure would necessarily be created from a dispersion of sheets and polycations. The most direct evidence on the height of the pillars is provided by the basal spacings observed. These vary in these samples from 16.9 to 18.7A. However there is not a clear trend with stoichiometryand the line-broadening suggests that there is considerablyvariability around this average spacing. As can be seen from Table 3 the spacing drops from 18.9 to 16.9A as the calcination temperature is increased. This implies an average pillar height initially of 9.7 dropping to 7.7A at 5OOOC.
(assuming a sheet thickness of 9.2A). This is
consistent with 20% linear shrinkage on dehydration of the pillars and is similar to that expected from the loss of 1 layer of oxide ions from the four layers in the Al,, ion. The other evidence presented here is from nitrogen adsorption isotherms and from the ability of larger molecules to penetrate the structure. It is instructive to calculate the nitrogen micropore volume based on the above ideal model. Making the assumptions that the clay sheets are 9.2A thick and have a density of 2.8 g/ml and that the Al 13 pillars are 9.5A diameter spheres located at a packing density of l/28 rings, then the micropore volume cannot exceed 0.28 ml/g. There are several reasons why lower volumes might be found. First the figures in the Tables are for impure montmorillonites. If 70% pure then 0.2 ml/g would be expected. Secondly the calculation assumes 100% efficiency in space filling by the nitrogen molecules and if the pores are close to the dimensions of a single molecule this is unlikely. Thirdly if the pillar heights are variable then sometimes the sheet separation will be too small to allow the efficient packing of nitrogen molecules. The actual data found are that maximum micropore volumes for alumina-pillaredmontmorillonitesare in the region of 0.12 ml/g. This is significantly less than values reported by Fripiat et al for alumina-pillaredbeidellites but comparable to their values for alumina-pillared montmorillonites (ref. 10). Fripiat has however discovered differences between the montmorillonite and beidellite samples in the behaviour of the pillar structure on dehydration. Since the micropore volume is only about 60% of expectation then it appears that molecular packing is quite inefficient. The situation is even worse for the 5OO'C calcined products where the micropore volume drops 38% even though the pillar height only drops 20%. Consequently the filling efficiency is then only 50%. In liquid nitrogen each molecule normally occupies 57.5 IL3but in these micropores an average occupancy of ca lOOA has been found. The fact that the 290 mmole/lOOg Alp/MM samples show a cut-off of about 4.5A in size of molecule
631 which can penetrate while the other samples do not, suggests that the entrances to the pore system are smaller in this sample but that the majority of the internal pillars are significantlybigger than this. This also implies that the pillaring process is variable. Lastly the synthetic clay, Laponite, also shows variable pillaring heights using the same reagent as alumina source. Consequently overall the best structural model seems to be one in which there is considerable variability in the heights of pillars incorporated. The observation that the charge on each Alp cation decreases from 0.54 to 0.2 as the synthesis ratio is increased also supports this idea. The main implication of these findings is that it is difficult to synthesise pillared clays with accurately defined micropore dimensions. While average dimensions can certainly be varied, it is likely that each preparation will possess a spread of pore dimensions which depends in turn on the preparation conditions. For these reasons it is again unlikely that sharp cutoffs in molecular sieving capability at any chosen size are going to be readily achievable.
REFERENCES 1. G.W. Brindley and R.E.Sempels Clay Minerals 12, (1977) 229-237 2. D.E.W. Vaughan, R.J. Lussier, J.S. Magee Jr. U.S. Pat. 4,176,090 (1979) 3. D.E.W. Vaughan, R.J. Lussier, in L V Rees (Editor), Proc 5th Int. Conf. Zeolites, Naples, Italy, June 2-6, 1980. Heyden & Sons, London, 1980, pp 94-101. 4. N. Lahav, U. Shani, J. Shabtai, Clays and Clay Minerals 26, (1978). 107-115 5. J.Y. Bottero, M. Axelos, D. Tchoubar, J.M. Cases, J.J. Fripiat, and F. Fiessinger, J. Coll. Interface Sci., 117, (19871, 47-57 6. J.W. Akitt and A. Farthing, J.C.S.Dalton, 1981, 1606-8 7. J.W. Akitt and A. Farthing, J.C.S.Dalton, 1981, 1624-28. a. J.R. Mayes, Miner. Eng. Sot. Tech. Mag., 1976, 41-53 9. D.W. Breck, "Zeolite Molecular Sieves", Wiley, New York,1974 10. D. Plee, F. Borg, L. Gatineau, and J.J. Fripiat, J. Am. Chem. Sot., 107, (19851, 2362-2369