The Synthesis and Characterisation of Iron Silicate Molecular Sieves Sieves

The Synthesis and Characterisation of Iron Silicate Molecular Sieves Sieves

The Synthesis and Characterisation of Iron Silicate Molecular Sieves Sieves W J Ball, J Dwyer, A A Garforth and W J Smith Chemistry Department, UMIST,...

445KB Sizes 2 Downloads 56 Views

The Synthesis and Characterisation of Iron Silicate Molecular Sieves Sieves W J Ball, J Dwyer, A A Garforth and W J Smith Chemistry Department, UMIST, PO Box 88, Manchester UK BP Research Centre, Sunbury-on-Thames. UK Ferrisilicates, containing iron in both framework and extraframework positions were synthesised. Characterisation using MASNMR, ESR, TPDA, ESCA and SIMS gave results which were consistent with framework substitution. Alumino-silicates crystallised more rapidly than corresponding ferrisilicates and a secondary nucleation of silicalite was observed in the iron system at prolonged crystallisation times. Acid sites in ferrisilicates were weaker than those in aluminosilicates. Product distributions, in the methanol conversion, were intermediate between those for H-ZSM-5 and silicalite. Relatively more coke and C aromatics were observed with 10 ferrisilicates and with iron impregnated silicali te. In the Fischer-Tropsch synthesis ferrisi1icates were less active than iron impregnated si1icalite but activity was improved by hydrothermal treatment which dislodged framework iron. INTRODUCTION There is considerable interest in isomorphous substitution of aluminium, in zeolite frameworks, by other elements (l) and several papers describe substitution by Fe{IIl) (2)(3). Currently, however, there are few studies concerning the synthesis of ferrisilicates which give extensive characterisation and provide catalytic evaluation. In this paper we present details of synthesis and use a wide range of techniques to characterise silicates having the pentasil (ZSM-5) structure and the associated Fe (III) species. Catalytic results for methanol conversion and for the Fischer-Tropsch synthesis are discussed. EXPERIMENTAL 1 Zeolite Synthesis and Characterisation Ferrisilicates (4) were synthesised hydrothermally in monel, or PTFE-lined stainless steel autoclaves, using tetrapropyl ammonium hydroxide (TPAOH) or -bromide (TPABr), a source of silica (Ludox AS40 or Pyramid No 1 sodium silicate solution), sodium hydroxide and ferric nitrate. Corresponding alumino-silicates (pentasils) were made by substituting an aluminium source for ferric nitrate. Iron impregnated silicates were prepared from silicalite and Fe{N0 solutions. 3)3 Hydrothermal treatment involved steaming at fixed partial pressure of steam for a given period at 600 DC. A Philips diffractometer (PW 1380) was used for XRD and electron micrographs were obtained with a Philips SEM 505 and a Philips EM400T. Both instruments were fitted with analytical facilities (EDAX). Framework infrared studies utilised a Perkin Elmer 397 spectrometer or a Nicolet FT/IR system, and ESR spectra were generated with a Varian E-9 X band spectrometer (-9.5 GHz) with a 10 kHz magnetic field modulation and pitch signal (g = 2.0028) as reference. Acidity was measured by temperature programmed desorption of ammonia (TPDA) using a TPD/MS system. Magic-angle-spinning NMR w!27e obtained using a GX 400 instrument at 104.17 MHz with spinning rate 3-5 ( Al) and an FX 200 spectrometer at 39.65 MHz with spinning rate 3-5 kHz ( si ) . Surface

137

138 (SY-8-2) analysis and depth profiling were made using a VG SIMSLAB instrument, samples were etched with an argon ion beam current of 15 nA.

RESULTS AND DISCUSSION 1 Zeolite Synthesis and Catalyst Preparation 1.1 Crystallisation and source of silica A typical synthesis with starting composition, 98SiO .4Na20.10(TPA)20. (LudoxAS40 silica, TPAOH) gives a crystalline pro~uct with XRD and Fe 203.1000H20 infrared patterns associated with ZSM-5/silicalite-l.. The splitting of the reflection at 20 = 45.5 increases with increasing Si/Fe (Fig 1) which, by analogy with the Si/Al system (4,5), is consistent (in the absence of ZSM-ll) with the incorporation of Fe(III) into framework sites. Aluminium content is minimal and is constant in these zeolites. from colloidal silica consist (typically) of spherulitic aggregates of rod-like crystals, stacked radially, as revealed by SEM and TEM (Fig Differences arise when sodium silicate solution is used. For example a starting composition 98Si02.29Na2o.l0(TPA)20.Fe203.l624H20 results in significant by a-quartz ana a different morphology associated with large "twinned" coffin-lid crystals (Fig 2) and a broad crystal size distribution. These results might be explained by assuming a solution phase mechanism (6,7) for the sodium silicate system and a solid phase mechanism (7,8) for colloidal silica which would suggest a reversal for the analogous aluminium (7) and further work on this aspect is in progress. 1.2 Crystallisation of ferrisilicates compared with A1ZSM-5 The crystallisation of ferrisilicate and A1ZSM-5, both from colloidal silica, are compared in Fig 3. Aluminosilicate ZSM-5 crystallises more rapidly than the corresponding ferrisilicate. Moreover, in the ferrisilicate system a secondary nucleation of twinned silicalite crystals is observed after 72 hours, resulting in a bimodal distribution of crystal sizes (Fig 4b). No secondary nucleation is observed in the aluminium system (Fig 4a), nor in the iron system, when sodium silicate is used in place of colloidal silica. Furthermore the induction period prior to secondary nucleation is strongly dependent upon initial composition. The effect of variation in the initial composition on crystallisation of ferrisilicates (based on colloidal silica) is shown in Fig 5. Both crystallinity and crystal size decrease with increasing iron content. Crystallisation shows an optimal region for OH/Si and crystal change from single spherulites to smaller aggregates as OH/Si increases. Both effects are well documented for synthesis of the Al analogues (9,10). Further comparison between the iron and aluminium systems can be seen from Fig 6 which compares the incorporation of Al(III) and Fe(III) into framework tetrahedral positions (Fig 6a). Incorporation of Al is followed by the signal around 50 ppm in the MASNMR and incorporation of Fe by the signal at g = 4.3 in the esr spectra. Results show that aluminium is more readily and rapidly incorporated into framework positions than Fe (III) in agreement with results in Fig 3. However, the amount of tetrahedral Al is relatively constant after 24 hours, suggesting that crystallisation is complete and this accords with SEM results which show no further increase in crystal size after this period (Fig 4a).

W.J. Ball et al.

139

The increase in tetrahedral Fe(III) reflects the growth of ferrisilicate and its reduction after 24 hrs coincides with the secondary crystallisation of silicalite (this was clearly observed by SEM after a period of 70 hr s ) . Fig 6b shows the effect of composition variables Si/Fe and TPA/Si on the incorporation of tetrahedral Fe(III). For Si/Fe > 30 there is a progressive decrease in the signal at g = 4.3, as Si/Fe increases, corresponding to reduced iron content. The reduced signal at Si/Fe<30 arises from the reduced degree of crystallinity at low Si/Fe (Fig 5b). Increase in the TPA/Si results in a steady increase in the inclusion of Fe(III) into tetrahedral sites and this effect is also reflected in increased reaction rates when TPA concentrations are increased. These results clearly show that the Fe(III) systems are, up to a point, similar to the Al(III) systems when colloidal silica is used as the silica source. Differences in rates of crystallisation may be related to differences in the chemistry of Fe (III) and Al (III) in highly alkaline solutions. It is well known that Al(III) exists as monomeric Al(OH)4- in alkaline solutions, but the state of Fe(III) is less well defined, with polymeric Fe(III) species or octahedral species likely to be present. If the reaction rate is dependent on the condensation of appropriate silicates and metal hydroxide species to form zeolite nuclei it is likely that the different reaction rates for ferrisilicates and AIZSM5 are related to the different concentrations of the appropriate metal hydroxy species in solution. 2

Characterisation of Ferris~bicates Magic angle spinning NMR ( Si) for ferrisilicates and silicalite are shown in Fig 7. A shoulder around -105 ppm, clearly visible in the ferrisilicates is close to the value for Si(lAl) and is tentatively assigned to Si(lFe). It is known that a peak in this region can arise from silicons containing terminal hydroxyls and in aluminosilicates this can be examined by using cross polarisation (11). However, in the ferrisilicate, initial cross polarisation experiments did not resolve this point presumably because of the much ~~ster relaxation of protons in the presence of iron species which diminished all Si signals in the CP mode. This point is currently being ~estigated. However, it was observed that the width at half height of the Si signal increases with increasing iron content in the ferrisilicates in agreement with highly dispersed Fe(III) in intimate with the Si as might be expected from isomorphous substitution of Fe NMR spectra are considerably broadened by the presence of paramagnetic iron but a constant small amount of tetrahedral aluminium (impurity) can be detected in the ferrisilicates of low iron content. Surface analysis using SIMS demonstrates that the Fe(III), in ferrisilicate is homogeneously distributed throughout, whereas in iron impregnated silicalite the outer surface is enriched in iron. This result is consistent with, but does not prove that, the Fe(III) is in the framework. Extensive steaming causes some pore blocking, presumably due to dislodged iron species. ESR spectra of ferrisilicates (Fig 8) reveal two main signals, a sharp signal at g = 4.3 assigned to tetrahedral Fe(III) (13,14), and a broad signal at g = 2.0 assigned to octahedrally coordinated Fe (III) in non-framework positions. These signals show different temperature dependence. The much larger increase in the intensity of the signal at g = 4.3, with reduction in temperature, is consistent with the presence of Fe(III) as a matrix element intimately associated with the zeolite structure, as expected for isomorphously substituted Fe(III)· (Fig 8). On calcination in air 550°C for 16 hr-s ) the intensity of the g = 4.3 signal decreases, which is compatible with the removal of some framework Fe (III) during calcination. Removal of framework Al(III) species on calcination of aluminosilicates is known to occur (15). The esr spectra of silicalite and of impregnated silicali te (1% loading) are also shown in Fig 8. In both of these cases a small signal at g = 4.3 corresponds to a residual iron impurity. It is noticeable that, after impregnation, there is no increase in this residual signal. These spectra show clear differences between iron incorporation during synthesis and iron dispersed post synthesis. In Fig 9 are shown esr spectra of an iron silicate successively extracted with

140 (8Y-8-2)

oxalic acid. An initial decrease in intensity of the g = 2.0 signal from extraction of some non-framework (probably neutral) iron species. The g = 2.0 signal corresponds to Fe(III) species in less accessible or held positions (probably charged species in cationic sites). In agreement with the assignment of the g = 4.3 signal, to framework Fe(III), its intensity is not reduced by the Following extensive steaming of ferrisilicates a broad ESR signal, similar to that assigned (16) to highly dispersed oxidic iron species is observed. Results of some preliminary experiments on dehydration and reduction with carbon monoxide of ferrisilicates are shown in Fig 9. The intensity of the g = 2.0 signal is decreased, on dehydration, due to the self reduction of octahedral non-framework iron species (18). Reduction with CO (pulsed at 400°C) leaves the residual g = 2.0 and the g = 4.3 signal unaffected, indicating that the former signal arises from octahedral species that are stable to reduction with CO. These are likely to be charged cationic species. As expected Fe (III) present in the zeolite framework is not readily reduced by CO. Stability of framework Fe(III), to reduction, is suggested from MO calculations (20). Dehydration of the ferrisilicate causes a change in lineshape of the g = 4.3 signal indicating a change in symmetry for the Fe(III) assigned to framework positions. Rehydration restores the original lineshape so that this process is reversible. At present the details of this symmetry change are not clear, although it could arise from loosely coordinated water molecules or from (reversible) changes in the lattice or ion positions on dehydration. However, the comparison with an iron impregnated (1%) silicali te is instructive. Dehydration of impregnated silicali te (Fig 9) leaves the signal centred at g = 2.7 unaffected. Reduction with CO causes the colour of the impregnated sample to change from brick-red to grey and a very intense esr signal develops (Fig 9) which we assign to reduction of a-Fe to 203 produce reduced superparamagnetic species. The acidic properties of the ferrisilicate were also studied using TPD. Both H20 and NH were monitored. A TPD profile of an ammonium form ferrisilicate (Fig 3 lOa) shows that these materials are relatively hydrophillic. A peak corresponding to desorbing NH indicates an ion exchange capacity and the presence of framework 3 Fe(III)04 units. The TPD profiles of ammonium forms of silicalite and iron impregnated silicali te are featureless. This is consistent with the absence of (significant amounts) of framework Fe(III)04 units. The TPDA profiles from acid forms of a ferrisilicate and an analogous AlZSM5 are shown in Fig lOb. The two peaks present in the profiles correspond to physi- (low temp) and chemi- (high temp) sorbed ammonia. Although the two profiles are similar the chemi-sorbed ammonia is removed at a lower temperature in the ferrisilicate (280°C) compared to the AlZSM5 (320°C) in keeping with the expected lower acidity of the ferrisilicates. These results, MASMNR, ESR, TPDA, SIMS and the changes on dehydration reduction and steaming are consistent with substitution of a considerable proportion of the Fe(III) species into the framework. 3

Catalytic Properties of Ferrisilicates Typical results for methanol conversion on ferrisilicates and comparisons with silicalite, iron impregnated silicalite and HZSM5 are given in Table 1. Selectivities to light gases, olefins and aromatics are as expected. Iron impregnation does not markedly change the distribution from that for silicalite but there is more coke and slightly more lower alkanes and aromatics when iron is present. This may arise from more facile hydrogen transfer in the presence of iron species. Selectivities for the ferrisilicate are, as expected for framework substi tution producing acid sites somewhat weaker than those of HZSM5, intermediate between those for silicalite and HZSM5. Qui te good yields of aromatics are obtained at 450°C. Product distributions wi thin the aromatics differ, however, in that the ferrisilicate, silicalite and iron impregnated silicalite show a higher proportion of the large C aromatics (Table 1). This, lO presumably, implies relatively more reaction at unrestrained sites (possibly surface sites). The enhanced C aromatic proportion in ferrisilicates and iron lO impregnated silicalite may be associated with the presence of non-framework iron

W.J. Ball et al.

141

species, initially present in the impregnated sample, and presumably generated in the ferrisilicate by in situ steaming resulting from product water in the methanol conversion (19). Some preliminary results for the Fischer-Tropsch synthesis are given in Table 2 where it is seen that the ferrisilicates are not very active in this reaction, compared with the impregnated catalyst. However, following hydrothermal treatment, which dislodges framework Fe (III) and produces highly dispersed iron oxide species, a much more active catalyst is generated. CONCLUSION Pentasil structures in which iron is isomorphously substituted for aluminium, in the framework, can be synthesised. Crystal sizes and morphology depend upon synthesis conditions. Some non-framework iron is generally obtained in the zeoli tes, a part of which is readily removed. Catalytic properties depend upon the acid sites generated by framework iron which are weaker than those in corresponding aluminosilicates, and on involvement of oxidic iron species. Catalytic activity can be modified by hydrothermal treatment which dislodges iron from framework positions. ACKNOWLEDGMENT We thank BP International for grant of an EMRA (AAG) and for financial support (WJS) • REFERENCES (1) E Moretti, S Contessa, M Padovan, Chim Ind, 67, 21, 1985. (2) H W Kouenhowen, and W H J Stork, USP 4,208,305, 1980. (3) P Ratnasamy, R B Borade, S Sivasanker, V P Shiralker, S G Hedge, Proc of Int Symp on Zeolite Catalysts, Siofok (Hungary), 137, 1985. (4) D M Bibby, L P Aldridge, N E Mi1eston, J Catal, 72, 373, 1981. (5) S S Pollack, J WAdkins, E L Wetzel and D Newbury, Zeolites, Vol 4, April, 181, 1984. (6) S P Zhdanov, Adv Chern Ser, No 1010, 20, 1971. (7) E G Derouane, S Detremmerie, Z Gabelica and N Blom, Appl Catal, 1, 101, 1984. (8) B D McNicol, G T Pott and K R Loos, J Phys Chern, 76, 2288, 1972. (9) K J Chao, T STasi, M S Chen and I Wang, J Chern Soc, Faraday Trans 1, 3, 547, 198!. (10) S B Kulkarni, N P Shiralker, A N Kotasthane, R B Borade and P Ratnasamy, Zeolites, Vol 2, October, 313, 1982. (11) G Engelhardt, V Lohse, A Samoson, M Magi, M Tarmak and E Lippman Zeolites, Vol 2, January, 59, 1982. (12) A G Ashton, S Batmanian, D M Clark, J Dwyer, F R Fitch, A Hinchliffe and F J Machado, "Catalysis by Acids and Bases" (Ed B Imelik et all Elsevier, p101, 1985. (13) B D McNicol and G T Pott, J Catal, 25, 223, 1972. (14) E G Derouane, M Mestdagh and L Vie1voye, J Catal, 33, 1969, 1974. (15) C V McDaniel and P K Maher, "Zeolite Chemistry and Catalysis" ACS Monograph 171, 285, 1976. (16) L E Iton, R B Beal and D T Hodu1, J Mol Cata1, 21, 151, 1983. (17) W J Smith, PhD Thesis to be submitted 1986. (18) J Novakova, L Kubelkova, B Wichteva, T Juska and Z Do1ejsekiz, Zeolites Vol 2, January, 17, 1982. (19) A G Ashton, S Batmanian, J Dwyer, I S Elliot and F R Fitch, J Mol Catal, 34, 73, 1986. (20) S Beran, P Jiru and B Wichter1ora, Zeolites, Vol 12, October, 252, 1982.

142 (5Y-8-2)

~Sll callt1a

M -

c100 o 80 Of == 60

:i=

Ferrisilicates

Si/Fe=83 --.-/ ' - - =38 ~

=25

i40 ~ 20

°

00

46 2El/o 44 Fig.1 XRD Patterns at 2e.44-46°

Fig.' Crystallisation of Al (0) or Fe silicates(using colloidal silica. 985i0 ~Na20-10(TPA)20-M20,-10 H20).

«»

3

Fig.4 SEM micrographs of (a)AIZSM5 (b)terrisilicate revealing secondary nucleation

Fig.2 Ferrisi1icate prepared from colloidal silica (a) 5EM,(b) TEM (c) prepared from sodium silicate

W,J. Ball et aI,

OH/Si

0

1·0

~

,h100 'I: = ~ 80

(a)

t

143

60:; :I:

5~ ~

4~

'iii c

b' 60

GI

2

40

20

00

00

20 40 60

20 40 60 80 100

:5

1

time/hours

80 100

Si/Fe

Fig.5 Effects of OB/Si and SiIFe 0 T=1?5 C, TPA/Si.O.2, H 100-";";:~';"";'' ' ' ;;'' ;;'' '~;'' ' ;'~;'' ;;;''; Ca) 4 days Cb) 2 days. 20/Si=10,

~II 80

en ~60

Ferrisilicates Si/Fe

44

'j! 40

~

83

relative widths (Yz height)

(2'8)

(1'5)

..

~

\

20 0

20 40 60 80 100

Si/Fe

Fig.6 (a) Incorporation of MCIII) into tetrahedral framework sites during synthesis Cb) Effect of TPA/Si and SiIFe on incorporation of FeCIII).

100 ppm (1'0)

I

50

,

100

silicalite I

150 ppm

(b)

I

I

I

Fig.? 29Si MASNMR of Ferrisilicates 120 240 360 (a), Cb) and silicalite Cc) Fig.10 Ca) TPD of an NH~-ferrisilicate (b) TPDA from BferrisilicateC-) and BAIZSM5(--)

144 (SY-8-2)