2010:077
M A S T E R ' S T H E S I S
Alternative seeding methods for growing MFI zeolite films
Romain Destruel
Luleå University of Technology Master Thesis, Continuation Courses Chemical and Biochemical Engineering Department of Chemical Engineering and Geosciences
Division of Biochemical and Chemical Engineering
2010:077 - ISSN: 1653-0187 - ISRN: LTU-PB-EX--10/077--SE
2
Abstract
Zeolite membranes can be used in particular for gas separation. For this purpose the aim of this work was to focus on the seeding step of the synthesis of zeolite films and improve it in order to obtain thinner films. Indeed thinner films give rise to a lower pressure drop, an essential point in gas or liquid separation.
For this purpose, several different seeding methods have been studied. The first method consisted in growing a membrane on a polished alumina disc. The results were quite good since we obtained a more uniform deposition compared to unpolished alumina support. But the polished sample was too much polished and the layer with fine porosity was not present anymore.
In this work, we also studied seeding and the different steps of growing a film on silicon wafer. An interesting approach to obtain a dense deposition of seeds was to do multi-seeding, i.e. repeating the seeding process several times. It worked but the drawback is that the thickness of deposition increases with the seeds density.
Moreover, the observation of several films after various synthesis times showed that nanoparticles are likely to take part in the growth mechanism.
The last part of the report is about electrophoretic deposition. Despite all our efforts, it never worked as expected. But on the other hand the dip-coating method seems promising and might give very interesting results.
3
Table of content
Introduction ... 4
1 Silicate-1 zeolites growth mechanism ... 5
2 Experimental Part ... 12
2.1 Synthesis solutions ... 12
2.2 Films on alumina supports ... 12
2.2.1 Experimental procedure ... 12
2.2.2 SEM observations ... 13
2.2.3 Polishing of waxed alumina supports ... 15
2.3 Films on carbon films... 18
2.3.1 Experimental procedure ... 18
2.3.2 SEM observations ... 18
2.4 Films on silicon wafers ... 20
2.4.1 Experimental procedure ... 20
2.4.2 Seeding step ... 20
2.4.3 Growth of the film ... 31
2.5 Seeding by Electrophoretic Deposition (EPD) ... 37
2.5.1 Literature review about zeolite seeding by EPD ... 37
2.5.2 Experimental procedure and results ... 39
Conclusion ... 51
Bibliography ... 52
4
Introduction
Zeolites are crystalline microporous aluminosilicate materials often used in industrial applications thanks to their unique properties. Continuously improving, they offer a strong potential in catalysis, adsorption, ion-exchange, membrane separation, microelectronics …
In the laboratory of the Division of Chemical Engineering, silicalite-1 films are grown on alumina discs as support, after a seeding step. They can be used for gas separation for instance. For this utilization, a very important point is to have the smallest pressure drop as possible. To do that, the separation membrane should be as thin as possible.
This work focuses on the seeding step and studies several possible methods of seeding. Indeed, improving the seeding step will help to have a better membrane. The aim is to have a seeds deposition as dense and as thin as possible. Ideally the best would be to get a very dense monolayer of seeds. We want the deposition to be thin because a thin membrane implies a thin seeds deposition. And it should be dense so that the membrane growth requires less hydrothermal treatment time to be uniform and continue because less hydrothermal treatment time means thinner membrane. Besides, a dense seeding would certainly imply fewer defects in the membrane structure.
5
1 Silicate-1 zeolites growth mechanism
Having a better understanding of the zeolite crystallization mechanism could accelerate momentous technological developments but despite many experimental studies on nucleation and growth of zeolites, the mechanism still remains elusive.
Silicalites-1 are nanoporous pure silica zeolites, with a MFI framework, and have been extensively studied as a model for the growth mechanism. Indeed silicalite-1 has been the first zeolite synthesized from a clear solution, which enable much easier (and in-situ) analysis of the zeolite synthesis than gels. Besides, the synthesis mixture is relatively simple: tetraethylorthosilicate (TEOS), silica source, tetrapropylammonium hydroxide (TPAOH), template agent, and water.
Since then, this model has been studied with many different technologies such as X-ray and neutron scattering, transmission electron microscopy (TEM), infrared spectroscopy (IR), nuclear magnetic resonance (NMR), dynamic light scattering (DLS), atomic force microscopy (AFM)…
Yet zeolite formation is still not well understood. All researchers agreed on the presence of small nanoparticles, called precursors, in the clear solution prior to the formation of crystalline zeolites.
Those subcolloidal particles have been observed by a wide variety of techniques and their size is about 3 nm. But both their structure and their role in the zeolites formation are still in debate and several different models have been proposed. One of the difficulties in the observation of zeolite growth is that it happens at a length scale between the accessible length scale of NMR and diffraction techniques (1).
The first point of disagreement among researchers is the nature of those nanoparticles. Various structure models have been proposed, ranging from amorphous silicate particles to highly defined framework similar to the crystalline zeolite-1 one (MFI).
A group from Leuven has extensively studied the mechanism of silicalite-1 crystallisation. They used
29Si liquid NMR, X-ray scattering, TEM and in situ IR spectroscopy to analyse the nature of the precursor species in solution (2) (3) (4) (5). They conclude from their results that, under ambient conditions, the polycondensation of the silicon leads to the formation of species containing 33 Si atoms and containing already the MFI framework.
With further condensation, the linking of 12 of these units results in the formation of so-called nanoslabs. Their size is 1.3x4x4 nm. They have been identified as the nanoparticles of the clear
solution and are crystalline (Figure 1). After a rise of the clear solution temperature, further aggregation of those nanoslabs leads to the formation of zeolites crystals (Figure 2).
Figure 1. Individual crystallised blocks (4)
6
Figure 2. Proposed mechanism of (A) nanoslabs formation and (B) silicalite-1 crystal growth by aggregation of nanoslabs (2)
However, some other findings are in disagreement with this mechanism, especially because, according to some other studies, the precursor nanoparticles are not crystalline but amorphous.
Tsapatis, Vlachos and their co-workers have reinvestigated the formation of the nanoparticles using the same protocol than the Leuven team and making a careful characterization of the nanoparticles using various techniques (6). They conclude that the existence of MFI nanoslabs is not supported by experimental results and other plausible structures can explain it.
Others studies using X-ray and FTIR (7) (8) also report there is no evidence of crystallinity in the precursor nanoparticles. The first indication of it arrives later in the formation process, when a second population of particles appears. This leads to the second main point of disagreement among researchers: the role of those nanoparticles in the nucleation and growth mechanism.
Currently, there are 3 main possible models proposed and discussed (4). One model assumes nanoparticles are amorphous and act as a source of monomeric or oligomeric silica species which feed the growth of larger particles by their dissolution (Ostwald ripening process) (Figure 3, pathway A). The two other models assume the nanoparticles are structured but differ on the growing process:
in one, the nanoparticles are added to a growing crystal (Figure 3, pathway B), in the other there are successive aggregations of nanoparticles into larger and larger crystalline units (Figure 3, Pathway C).
Actually there are also models which assume amorphous nanoparticles but with a growth mechanism by aggregation.
7
Figure 3. Schematic illustration of the three models of Silicalite-1 formation (4)
In 1996, Schoeman et al. studied the initial stages of the silicate-1 crystallisation with dynamic light scattering (DLS) and cryo-transmission electron microscopy (cryo-TEM) (9). Before the hydrothermal treatment, they observed the presence of a single particle population with an average diameter of 3.8 nm. Then they analysed samples at different times after the starting of the hydrothermal treatment. The light scattering analysis of the first sample, after 1.5h, indicates that the solution still contains a single population but with an average diameter of 5 nm. Two possibilities can leads to such augmentation. The first one is, because of the temperature increase and thus of the solubility increase, the smaller particles disappear and the larger grow by an Ostwald ripening mechanism. The second possibility is the apparition of a second population of particles which is not yet distinguished.
Indeed, the resolution in the light scattering technique is such that two particles populations can be distinguished only if one is at least twice as large as the other one. So in that case the average size of the two populations would be 5 nm. However, the analyses of the following sample showed that this 5 nm population remains present during all the hydrothermal treatment. So the presence of a second growing population is not likely. But the consequences of the invasive nature of the sampling are not well known and it is possible that the cooling has an effect on particle size, which would explain such results.
The second sample, after 2.5 h, showed a bimodal particle size distribution (PSD) with a peak at 5 nm and another one at 26 nm. The following samples also showed a bimodal PSD with still a peak at 5 nm and the second one at an increasing size. Figure 4 show the results of the different samples. This second population is supposed to be the growing silicalite-1 crystals.
This is confirmed by an IR analysis which shows the presence of MFI-structured material from 4.5 h. The crystal size increase as a function of time is linear and corresponds to about 18.5 nm/h.
However, the authors of this study were unable to clearly show the role of the subcolloidal particles. They are a source of nutrient for the
Figure 4. Average size of the particles as a function of time (9)
8 growing crystals but it is not clear if it is via their dissolution or via their aggregation on the crystal surface.
In a latter study in 1998 (10), Schoeman proposed a model where part of the subcolloidal particles dissolves and provide nutrient reservoir for the crystals growth. His model was based on an estimation of the interaction potential between the subcolloidal and colloidal particles using the extended DLVO theory (Derjaguin-Landau and Verwey-Overbeek). Schoeman assumed a spherical particle shape and his calculations resulted in very high energy barriers for the approach of particles larger than 7 nm. This rules out a growth by aggregation.
Another, and more recent (2005), study (1) showed with molecular modelling that in very early stages in silica polymerization, silica clusters are predicted to be amorphous and they appear to grow through Ostwald ripening. A cluster-cluster aggregation is predicted to be an unlikely growth mechanism because of the strong negative surface charge on silica clusters under alkaline conditions.
A different growth mechanism was proposed by Tsapatis, Kumar et al. in 2006 (11). It is summarised in Figure 5. In this mechanism, the precursor particle A evolves through m intermediates (Bi) with structure increasingly similar to silicalite-1 nuclei (C1). Species Bi, although not yet silicalite-1 nuclei, can contribute to crystal growth, as well as other growing crystals, via an aggregation mechanism.
Their study showed the presence of amorphous precursor nanoparticles in the solution and then the apparition of a second larger particles population at least 2 hours after the beginning of the hydrothermal treatment. HRTEM results indicate that the secondary particles (Ci in Figure 5) are crystals of silicalite-1. They are typically faceted and have the characteristic hexagonal prismatic shape (with some protrusions of about 5nm). From the TEM images they obtained, researchers conclude that the crystals exhibit aggregate-like morphologies: slight misorientations in the crystallites, precursors sized protrusions, … suggest a crystal growth by oriented aggregations of particles Bi and Ci. Besides, the crystals become more compact and faceted in the latter stages of growth. This observation suggests a growth by monomer addition in the end of the mechanism.
In 2008, the same authors went further in their study using small angle X-ray scattering (SAXS) and cryo-TEM (8). As previously, they noticed the apparition of a second population of larger aggregate- like particles (30-50 nm). But this times SAXS
showed the particles Ci are actually predominantly amorphous. This new finding could not be obtained in their previous work employing conventional HRTEM because dialysis for sample preparation apparently dissolves the amorphous aggregates. In fact C1 is amorphous and the cristallinity appears progressively via intra-aggregate rearrangements
Figure 5. Schematic silicalite-1 growth mechanism (11)
Figure 6. Improved growth mechanism (8)
9 during growth. With this new element, they provided a slightly improved mechanism (Figure 6).
In 1999, the team from Leuven used the DLVO theory previously seen with Schoeman and used it to support their aggregation mechanism via nanoslabs and tablets (3), which is precisely the mechanism Schoeman had ruled out with the very same theory. This team differentiated three different particles populations and followed their evolution thanks to X-ray scattering (XRS). Their results are shown in Figure 7. In the first hours upon heating, the initial
amount of nanoslabs (precursor particles) decreases exponentially in favour of the intermediates (11nm).
With time, this second population goes through a maximum which correspond to the apparition of the third population of large particles. As soon as these particles were detected, Bragg scattering was observed, indicating the crystalline nature of these entities.
When Schoeman used the DLVO theory, he assumed the particles were spherical and this led to too high energy barriers for an aggregation mechanism. But, using their previous results, the team from Leuven used the DLVO theory for nanoslab shaped particles, which gave different conclusions. Due to their shape, the particles have in this case different properties for each of the three directions. The energy barriers in the b and c directions are small, which enables the formation of tablets by aggregation of nanoslabs (cf. Figure 2). On the contrary, the barrier in the a direction is considerable and no further aggregation at room temperature takes place. A rise of the temperature (hydrothermal treatment) allows further aggregation and thus the formation of intermediate and large particles.
A last interesting example is the study carried out in 2008 by Tokay et al. (7). It is a recent and extensive study of the silicalite-1 nucleation and growth processes from clear solutions. They used a lot of different techniques to follow the evolution of solutions for several compositions. From a qualitative point of view, all the compositions gave the same results. First, if we follow the evolution of particles effective diameter, we can notice a
sudden jump in size (Figure 8). This jump occurs for each composition but at different times. It appears to mark the beginning of the constant linear growth rate. This time also coincide with the exo-endo thermal switch time of the reaction, previously reported by Yang et al. (12). They also followed the evolution of the particle size distribution as a function of time measured by back-scattering (Figure 9). Very small particles (about 2.5 nm) dominate the reaction mixture before the hydrothermal treatment and after 10 min. But we can also notice that the
intensity weighted distribution, which magnifies large populations, shows the presence of a second population very early. It probably consists of few larger particles (around 20 nm).
Figure 7. Evolution of particle populations in normalized scattering intensity. Nanoslabs (nb), intermediates (in), large particles (lp) and Bragg crystalline material (c) are represented.
Figure 8. Effective particles diameter as a function of time
10 After 2 hours of synthesis, only this second
population seems to be present in the solution mixture. Although it is not shown in Figure 9, after the sudden increase in diameter, the second population also disappeared leaving only a third one. However it is unlikely that the two first populations really disappeared since DLS may be almost blind to the presence of small particles in the presence of larger ones beyond a certain concentration. Other studies have shown the two first populations remain in the synthesis mixture. An important thing to notice is the broadening in the particle size distribution of the second population with respect to that of the first one. We can see the same phenomenon between the second and the third population (average size: 70 nm) (Figure 10).
Those broadenings of the particle size distribution seems to be an indication of a growth mechanism by aggregation.
X-ray diffractograms of samples after different synthesis times can be seen in Figure 11. We can see the first samples seem to be amorphous while the first indications of crystallinity are detected in the 20 h sample, i.e. just after the sudden increase in particles diameter. Thus, this jump also corresponds to the nucleation time of silicalite-1 crystals. From these results it appears that the third population, growing at a constant rate, is not fully crystalline. Its crystallinity increases as it grows, and even after the end of
Figure 9 Variation of the intensity, volume, and number- weighted particle size distributions as a function of time at the beginning of the synthesis: (a) 0 h, (b) 10 min, (c) 2 h.
Figure 10. Number weighted particle size distribution before (14 h, 18 h) and after (20 h, 22 h) the sudden jump in particles size
Figure 11. X-ray diffractograms for different synthesis times.
11 the growth. The authors reported that these last two points could be experimental artefacts even if it is unlikely. TEM analysis also showed that the first indications of crystallinity were evident in the 20 h sample.
Monitoring of pH and zeta potential supports an aggregation growth mechanism too. For example, prior to nucleation, a sharp increase in surface negative charge took place. This can be explained by a significant surface area loss due to aggregation which leads to an increased surface charge density.
Finally, one last interesting result is that the density of the second population particles increases prior to nucleation and that the particles density continues to increase even after the end of the growth. The continuing increase in crystallinity and bulk density of the particles after size growth has stopped implied that monomer addition from solution also took place.
In summary, from this study, it appears there is a time of sudden jump in effective particles diameter.
This also corresponds to the exo-endo thermal switch and to the nucleation of silicalite-1 crystals.
Nucleation was accompanied by the aggregation of smaller particles, as indicated by the broadening of the particles size distribution and the pH and zeta potential variations. The particle population participating in nucleation was seen to be formed at early times of hydrothermal treatment via another aggregation of the initial precursor nanoparticles and to change in composition during the induction period prior to nucleation.
To conclude, the nucleation and growth mechanisms are still in debate. Techniques evolve and many research teams reconsidered their previous work to correct it and add some new findings. Many mistakes were due to sampling as well as to the limits of the observation techniques used, which led to some misinterpretations of the results. This partially explains the diversity of the proposed models. Besides, operating conditions (composition, synthesis temperature, …) vary between each team.
However, it seems most of the studies are moving toward amorphous precursor nanoparticles and a growth mechanism by aggregation. It seems very likely that actually each mechanism can be present and dominant at different synthesis time (and maybe for different operating conditions), and the most accurate description of zeolite growth
may be a combination of them: an existing crystal might grow by incorporation of subcolloidal particles (3 nm), small aggregates (15-20 nm) or by monomer/oligomer addition from the solution, with an equilibrium between the three species, as shown in Figure 12.
Figure 12. Possible mechanisms during growth of a silicalite-1 crystal (17).
12
2 Experimental Part 2.1 Synthesis solutions
In order to grow silicalite-1 zeolite films, a seeding method was used. To do so, two solutions are needed: the colloidal dispersion of zeolite and the synthesis solution, in which the membrane can grow.
The seeds are synthesized by a mixture of tetraethylorthosilicate (TEOS), the silica source, tetrapropylammonium hydroxide (TPAOH), the template agent, and water, the same chemicals as for the synthesis solution but in different proportions.
In order to make the synthesis solution, the chemicals are mixed and shaken for 24h.
The seeding solution needs more time. When the chemicals are mixed, they have to be blend for 48h. The mixture is then left in an oven at 50°C for 2 months. Once the seeds are formed, they have to be purified. To do so, we used centrifugation and redispersion in distilled water four times.
Thereafter, the dry content is measured and final solutions are made to the desired concentration.
The pH of the mixture has to be at least 10, in order to avoid the coalescence of the seeds. The final seed dispersion is stored in the fridge.
For seeding, a cationic polymer is also used, in order to attract the seeds, negatively charged, towards the support. This polymer solution contains 0.4% of a cationic polymer in water at pH 8.
2.2 Films on alumina supports
The membranes are classically obtained by growing the zeolite film on disc-shaped alumina supports.
Those supports have two layers: one layer with coarse porosity, the other with fine porisity. The seeds are deposited on the latter.
2.2.1 Experimental procedure
In order to grow silicalite-1 membranes on alumina discs, the supports need to be washed first in acetone, then in ethanol. Once they are dry, a mixture of PMMA (CM205, Polykemi) dissolved in acetone (1:3.75 by weight) is added on the top of the discs. When PMMA is dry, the supports are immersed at 150°C in molten wax under low pressure so that the air contained in the alumina is removed and the wax can diffuse inside. After this step, the supports are rinsed in acetone for one week in order to remove the PMMA.
The wax is used to prevent the diffusion of silicalite-1 seeds and synthesis solution inside the substrate. A previously deposited layer of PMMA enables to keep the top of the disc free of wax during impregnation and can be finally removed in acetone prior to seeding. This first step is referred
13 to as masking the substrate. To achieve seeding, since silicate-1 seeds are negatively charged at pH = 10, we need to make the supports positively charged. To do so, the cationic polymer solution mentioned above is used. Alumina discs are immersed in this solution for ten minutes.
After being rinsed with 0.1 M NH3, the supports can now be immersed another ten minutes in the seeding solution (dry content 1%, pH = 10), then re-rinsed with NH3. Then they are immediately hydrothermally treated in the synthesis solution seen above, in an oil bath at 100 °C for 30h.
Finally the discs are rinsed again in NH3 and calcined at 500°C to remove the wax inside the supports.
Now the discs are covered with the zeolite membrane and can be observed by SEM (Scanning Electron Microscopy).
2.2.2 SEM observations
Figure 13. SEM images of silicalite-1 membrane on an alumina disc
Gold layer
Silica support Zeolite membrane
Nanoparticles
14 The obtained membrane seems to be of acceptable quality (Figure 13): the zeolite film is continuous and quite uniform on the disc surface. Its thickness is about 0.5 μm. On the forth image (Figure 13, d), nanoparticles can be seen on the membrane. It could be the precursor particles involved in the silicalite-1 growth mechanism.
Yet, a lot of impurities can be found in the alumina disc (Figure 14). It looks like chewing gum between alumina grains.
In order to identify those impurities, we masked a support and observed it with the SEM just before the seeding step. The results were similar: the same substance is between the alumina grains in some places. Thus the problem certainly comes from the masking stage and not from the following ones.
An elemental analysis shows that it is composed mainly of carbon (C) and a bit of sodium (Na). These impurities could be PMMA or wax leftovers after calcination or could also actually come from our gloves when the sample is manipulated for SEM observation.
Figure 14. SEM images of impurities in the silica disc
Fine layer
Coarse layer Impurities
15 2.2.3 Polishing of waxed alumina supports
To avoid the use of PMMA and obtain a smooth surface, another approach was to immerse entirely the discs in molten wax without any PMMA cover and then to polish the top of the disc to remove the extra wax. But during the polishing stage, the blue liquid polishing agent was absorbed by the substrate. This means the disc was not completely impregnated with wax.
To overcome this problem, we tried to modify a bit the procedure. Instead of immersing the substrates in molten wax and then reduce the pressure inside the oven, they were placed on cold and solid wax and then the oven was heated under vacuum. It could be expected that the air trapped in the supports would thereby be removed more easily. This way, once the wax is melted, the empty pores start to be immersed in and better impregnation can be expected.
When this second disc was polished, the liquid polishing agent was still absorbed, but slightly less than in the first case. The waxed and polished disc was then seeded and observed by SEM.
The first observation which can be made (Figure 15) is that, even with a careful polishing, the disc surface is not perfectly smooth: there are still holes between the alumina grains. Besides, the fine layer of the disc disappeared: too much material was removed by polishing.
On a flat grain (Figure 16), the seeding is uniform but not very dense. It is very similar with the seeding obtained on silicon wafers (cf. 2.4). In holes between the alumina grains of the substrate, seeding is much denser (Figure 17).
Figure 15. Polished alumina disc (cross section and top)
16
Figure 16. Seeding on a flat alumina grain
Figure 17. Denser seeding between alumina grains
17 This sample can be compared with seeding on a classic masked alumina support (Figure 18).
The seeds distribution is much more uniform on the polished disc, which is expected to be better to grow a membrane from it. It is difficult to compare the density because in the second case, the seeds are gathered in groves in between the grains.
But these results seem promising. The next step should be to polish very carefully in order to keep the fine alumina layer. It is not easy because it has to be polished enough to remove the wax on the top and exhibit the alumina layer, but not too much to keep the very thin fine layer in place.
Figure 18. Comparison between a polished and a classic masked support.
18
2.3 Films on carbon films
2.3.1 Experimental procedure
We tried to grow zeolite membranes on carbon film supported by a copper grid. The aim was to make self-supported membrane after the calcination of the carbon film, and to see through the membrane by forming a membrane only on one side of the carbon film, and this turned to be the most difficult in practice. As for alumina discs, the carbon films first needed to be positively charged with our cationic polymer. To modify only the top side of the carbon film, we let the grid float on the liquid surface, top side down. One of the difficulties was then to retrieve this very small floating grid.
We didn’t want to use tweezers to avoid damaging the film. So we did it by suction with a pipette.
Then the grid was rinsed with 0.1 M NH3, put on the seeding solution surface and rinsed again. Finally it was hydrothermally treated in the synthesis solution for 36h in a 100°C oil bath. For this step, the grid was completely immersed but it should not be a problem since only the carbon side was seeded.
2.3.2 SEM observations
Two grids were used: one was just seeded while the other was hydrothermally treated.
For both samples, despite our precautions, the carbon film was severely damaged (Figure 19).
But pieces of film could still be found in some areas. It was observed by STEM to investigate the internal structure (Figure 20).The film of zeolite is thick, dense and zeolites are well merged together.
It is continuous and almost uniform.
Figure 19. The carbon film is damaged
19
Figure 20. Zeolites membrane on carbon film and copper grid (Top : SEM / Bottom : STEM).
20
2.4 Films on silicon wafers
To focus more on the membrane growth and avoid wasting time with the masking of alumina discs, we used silicon wafers as supports (around 1 cm2 in area). The influence of synthesis time was studied but we focused primarily on the seeding step.
2.4.1 Experimental procedure
First, wafers were washed. This was done in three steps, with water rinsing between each one:
- Ultrasonification in acetone for 5 minutes.
- Boiling in a mixture of H2O, H2O2 (30%) and NH3 (25%) for 5 minutes.
- Boiling in a mixture of H2O, H2O2 (30%) and HCl (37%) for 5 minutes.
It is wise to do everything in the same beaker so it is cleaned as well and can also be used for the following steps.
As previously explained, wafers were immersed in the cationic polymer solution (1h) and in the colloidal dispersion of seeds (1h), with 0.1 M NH3 rinsing after each step.
Finally the wafers were hydrothermally treated in the synthesis solution and calcined at 600°C.
2.4.2 Seeding step
In order to obtain a good zeolite film, seeding is very important. We want to have a uniform and very dense distribution of seeds on the wafer surface. It can be expected that this would give a membrane with fewer defects. It could also allow to have a shorter hydrothermal treatment to obtain a uniform and continuous film but also thinner.
21 As can be seen in Figure 21, with these conditions, seeding is quite uniform. But coverage is not very dense globally. Besides, the seeds seem to be distributed into small groups.
Figure 21. SEM images of a seeded silicon wafer
22 At the edge of the wafer, a very dense and thick layer of seeds was observed (Figure 22). This is certainly due to a drop of liquid which dried there.
In order to improve the results, we tried to perform multi-seeding. Hereafter we need to distinguish between five different samples:
- One seeding step (common sample) - Two seeding steps
- Three seeding steps
- Three seeding steps but with only one immersion in the cationic polymer(in the beginning) - Ten seeding steps
As expected, one seeding step produced exactly the same results as seen above, i.e. uniform but not very dense distribution.
Figure 22. SEM images of the dense wall of seeds
23 Figure 23 shows that the seeds are still uniformly distributed on the surface and coverage is denser than with only one seeding step. But there are still a lot of bare areas and it is now more evident that the seeds are gathered in small groups, as if they were attracted by each other.
Figure 23. SEM images of two successive seedings
24 With three seedings (Figure 24), the results are exactly the same: the distribution of seeds is denser but also thicker and there are still bare areas.
The two, three and ten seeding steps were made with immersion in the cationic polymer followed by immersion in the seeding solution each time (1 hour each). But we also tried to do three seeding steps with only the first immersion in the polymer solution (Erreur ! Source du renvoi introuvable.).
Between the three immersions in the seeding solution, only fast rinsing with 0.1 M NH3 was carried out. What was obtained was quite different from the regular three seedings just above. It is denser than only one seeding step but much less than with three “polymer+seeds” steps. It is more comparable to two “polymer+seeds” steps (Figure 25) but not exactly the same: it is a bit less dense but thicker. This results show clearly that the cationic polymer is very important, especially to attract the seeds on the surface of the silicon wafer and give a denser distribution.
Figure 24. SEM images of three successive seedings
25
Figure 25. Comparison between 1, 2, 3 and 3 (with just one polymer immersion) seedings
26 With ten successive seeding steps, we finally have a dense and continuous distribution (Figure 26).
There are not bare areas anymore, but the seed layer is quite thick. It is good to have a continuous distribution and now we could try to densify the seed layer by hydrothermal treatment. But we would prefer to produce something both continuous and very thin, like a very dense monolayer of seeds. That is why we tried later another seeding method: the electrophoretic deposition (EPD), (c.f.
2.5).
Figure 26. SEM images of ten seedings
Cross-section
27 At this stage of the study, we realized that the values given by the pH-meter were not reliable. The solutions were supposed to have a pH value of about 10, but pH strips indicated values around 7 or 8. Thus, other seedings on silicon wafers were made with, this time, solutions which really were at a pH value of 10.
As can be seen on Figure 27, seeding with this new pH value is denser than previously. Thus, it can be concluded that the pH value is of course a critical parameter. On this sample there were also areas with very dense and thick seeds deposition (Figure 28).
Figure 27. Comparison between a seeding solution at pH = 10 (left) and pH = 8 (right)
28 To evaluate the influence of the seeding time, we prepared a sample with one seeding step but the wafer was immersed just 10 minutes in the polymer and 10 min in the seeding solution instead of one hour and one hour. The result can be seen on Figure 29.
Figure 28. Areas with different densities.
Figure 29. Left: immersed 10 min / Right : immersed 1 hour.
Multilayers
Monolayer
29 The sample immersed during 2 x 10 min is slightly less dense than with the 2 x 1 hour treatment.
Thus, time is a parameter which influences seeding, but maybe one hour is not necessary and 30 min would be enough to obtain the same result.
We also made another sample with three successive seedings with the new pH value (Figure 30).
Once again, it can be seen that seeding is denser with the new pH value. There are almost no more bare areas with only three successive seedings.
Finally, we also tried to immerse a wafer ten times in the polymer solution for 30 min and then 1 hour in the seeding solution. Indeed, if one seeding does not give a very dense distribution of seeds it could be because the seeds distribution is not dense on the polymer or because the wafer is not completely coated with the polymer.
There were some quite thick areas with seeds embedded in the polymer (Figure 32). But most of the wafer was similar to one or two seedings. In Figure 31, we can see seeding is a bit denser than one seeding but in fact there is almost as many holes in both and they have almost the same size: the main difference is that seeds are more grouped in small heaps in this sample. Thus, ten immersions in the polymer do not improve seeding significantly.
Figure 30. Three seedings with right pH value
30
Figure 31. Comparison between a classic one seeding (left) and the sample with 10 immersions in the polymer (right).
31 2.4.3 Growth of the film
In order to follow the evolution of the membrane growth, several samples were made with different times for hydrothermal treatment.
For the first sample, it was stopped after only half an hour. This time is too short to obtain a membrane, but it can be seen on the pictures (Figure 33) that the seeds are beginning to grow and they even begin to merge in some places. But in other places, the phenomenon does not seem to have started. So the process does not initiate everywhere at the same time.
Figure 32. Seeds embedded in a polymer.
Figure 33. Early stage of film growth (30 min), on different places of the wafer. 32
33 Another wafer was hydrothermally treated for 3 hours (Figure 34). Seeds continued to evolve and this time growth had occured everywhere. It still cannot be said there is a continuous membrane.
After 8 hours of hydrothermal treatment, we finally obtained something which looked like a continuous membrane. The sectional view (Figure 35) shows the film is already quite thick: around 200 nm. The membrane is continuous and uniform. But it can be seen it is still growing because of all the nanoparticles present on the surface of the film (Figure 36).
Those nanoparticles seem to take part in the growth process.
Figure 34. Sample after 3h of hydrothermal treatment.
Figure 35. Sectional view of the 8h sample.
Gold layer
34 Besides, the membrane is not perfectly continuous on the entire wafer surface. There are still some bare areas. It was probably areas which were not covered yet before the hydrothermal treatment (Figure 37).
Figure 36. Continuous membrane after 8h of hydrothermal treatment.
Figure 37. Uncovered areas on the 8h sample.
35 Another wafer underwent hydrothermal treatment for approximately 130h. The obtained membrane is perfectly continuous and very thick as shown by the cross section (Figure 38). Thickness is around 1.1 μm.
On those pictures we can see texture in the grains.
Figure 38. Cross section of the 130h sample.
Texture
Figure 39. Top view of the 130h sample.
36 In Figure 40, it can be seen that the grains are formed by successive layers.
Figure 40. 130h sample: growth by successive layers.
37
2.5 Seeding by Electrophoretic Deposition (EPD)
Electrophoretic deposition has the potential to produce dense and very thin layer of seeds. As EPD is a quite recent method for zeolite seeding, a literature review has been done on the subject.
2.5.1 Literature review about zeolite seeding by EPD
One of the main applications of silicalite-1 zeolites is to make membranes for gas separation. In this purpose, it is interesting to make membranes as thin as possible.
Zeolites membranes have been widely studied and have been synthesized on many different supports with different techniques. For example, zeolites membranes can be synthesized by direct treatment of the support with a zeolite precursor solution but this method is difficult to control and needs a considerable film thickness in order to have continuous films. Another method is to deposit seeds on the substrate followed by a hydrothermal treatment in a synthesis solution (which is the method used in previous parts). This method is faster because the nucleation step is skipped.
However, in both cases, the film thickness and the zeolite particles size are difficult to control. In order to improve the process, electrophoretic deposition (EPD) could be used for seeding the support. With EPD, colloidal particles suspended in a liquid medium migrate under the influence of an electric field and are deposited on an electrode. It can be used for a wide variety of particles as long as they are charged and in suspension. This process is already widely used in the industry for coating. Its main advantages are fine uniformity of the deposition, easy manipulation and convenient control process. It should be noted that the process yields only results in a powder compact and therefore should be followed by a densification step if a fully dense material is needed.
Ke et al. used this technique to make zeolite coated carbon fibers and then hollow zeolite fibers after the calcination of the carbon support (13). They used a carbon fiber as the cathode and a platinum container served as the anode. The process was studied for different voltages (1-3 V), pH and times in order to optimize each of these parameters. Suspensions for EPD experiments were obtained by dispersing the nanosilicalite-1 in distilled water to a concentration of ca. 1.5 wt%.
A very important parameter for EPD is the zeta potential of the particles. If too small, the particles will coalesce and aggregate in the solution instead of depositing on the electrode. The zeta potential of the silicalite-1 particles depends on the pH of the solution. At pH 6, the potential was +4.5 mV and particles coalesced. At pH = 4.5, the potential was + 21.1 mV, largely uncoated areas could be noticed on the carbon substrate. Finally the team obtained the best result for a pH of 2.5.
A second important parameter is of course voltage. The same group tried to apply 1, 2 or 3 V during 2 x 10 min. As it can be expected, they noticed the deposit thickness increased with increased voltage. With 1 V, there were many bare areas on the carbon fibers. At 2 V, the fibers were fully coated with nanozeolites. When 3 V was applied, there were some thick areas but in the same time some bare area. It is believed that this is due to the beginning of water electrolysis, which produces hydrogen bubbles at the cathode. Finally the best voltage was thus 2 V.
38 As shown in Figure 41, the zeta potential of zeolites
is strongly dependant of the pH of the solution. The EPD could be made at either acid or basic pH and thus seeds would go towards respectively the cathode or the anode. That is why other teams worked with basic pH for EPD of zeolites ((14) and (15)).
The drawback of electrophoretic deposition in water is that the water electrolysis prevents the use of high voltages which would enable a faster deposition.
That is why Shan et al. worked on EPD of zeolites in non-aqueous medium (14). Solutions of silicalite-1
seeds (about 100 nm diameter) were centrifugated and redispersed in isopropyl alcohol (IPA) or in acetyl-acetone (Acac). All the suspensions were adjusted to a concentration of about 15 mg/mL. A platinum crucible was used as both the container and the cathode. A stainless steel grid was used as the anode for the deposition. With these non-aqueous medium, higher voltage and thus shorter time could be used, e.g. 100 V during 5 sec.
The medium used in EPD is an important factor, and more especially its ability to donate/accept protons because it affects the protonation state of the Si-OH group on the surface of zeolites which affects the charge of the zeolites in suspension. IPA is a proton donor while Acac, as a Brönsted base, can accept protons. As a result it was found that silicalite-1 could form only a dense monolayered coating (100 nm thick) in IPA but formed a thick film (several μm) in Acac. Moreover, increasing the EPD time increased the film thickness in the latter but did not change anything in the former. This is because silicalite-1 particles carry more negative charges in Acac than in IPA. Indeed, Acac accepts the protons of Si-OH, while IPA gives protons to Si-O-. This is also supported by a zeta potential measurement: -24.4, -50.0 and – 76.1 mV in respectively IPA, water (pH = 8) and Acac.
Electrophoretic deposition seems to be a simple and efficient way to obtain dense layer of seeds, and can be followed by a relatively short hydrothermal treatment in order to obtain a good zeolite film. The membrane thickness can be controlled by varying the synthesis solution concentration, the electrical potential, the voltage and the time.
Figure 41. Zeta potential of a silicalite solution as a function of the pH (16)
39 2.5.2 Experimental procedure and results
To put in practice the theory, a very simple experimental set-up was used (Figure 42). Two platinum electrodes were connected to a generator and the bottoms of the electrodes were immersed in a few millimetres of the seeding solution.
The electrodes were cleaned with acetone and cotton between each EPD experiment. In the beginning, the electrodes were rectangles cut from a platinum sheet. Their dimensions were 4 cm * 8.5 cm.
For the first trial, we used a wafer already used for seeding. It was glued onto the anode with silver paint. The fact that the wafer was already used did not really matter because it was only for testing the set-up. For this first trial, the EPD was made for 20 minutes, using 2 volts and with a seed dispersion of 0.1 wt.% in dry content. The solution pH was 10.4. After the EPD, the wafer was rinsed with distilled water.
The SEM observations did not show real differences between the pictures before and after the EPD trial. The deposition did not seem to be successful. Thus, in a second attempt, four new wafers were used: two on each electrode, side by side, with, in each pair, one coated with gold. Gold was to improve the electric conductivity, and both electrodes were carefully inspected after deposition to ascertain whether there was any deposited material. Besides, the time of the EPD was extended to 40 minutes and wafers were not rinsed at the end.
As it can be seen on Figure 43, there was indeed almost nothing on the cathode: as expected the deposition did not take place there.
Concerning the wafers on the anode, the result is very different from what we obtained with the classic way of seeding studied in part 2.4.2. In this case, the seeds repartition was far from being uniform. As shown in Figure 44, there are areas with almost nothing, others with a dense monolayer or thick packs of seeds. Similar results were obtained for both the gold coated and the uncoated wafer. Thus, the gold coating did not
contribute to any improvement. The problem did not seem to to be related to the electric conductivity of the wafers.
Figure 42. Experimental setup for EPD.
Figure 43. Wafer on the cathode.
40 Subsequently, we tried to increase voltage. Because of water electrolysis, it cannot be increased too much, but a trial was made at 2.5 V. For this one, two wafers, new and uncoated, were used on the anode. One of them was rinsed at the end, not the other.
For the rinsed wafer (Figure 45), there were only few seeds. Some areas had a bit more seeds, especially on the edges, but still, the seeds seemed to be removed by the rinsing step. This means they did not stuck to the wafer as they had with the EPD technique.
For the unrinsed wafer, the results were not much better. There were more seeds and less empty areas, but it was still not a uniform deposition (Figure 47). Moreover, there were many big particles in cubic or star shape as on Figure 46. An elemental analysis of these particles showed a high silver content. They certainly came from the silver glue used to glue the wafers to the electrode. The former probably dissolved into the seeding solution and contaminated it.
Figure 44. Seeding on the gold coated wafer from the anode.
41
Figure 45. EPD 2.5 V, rinsed wafer.
Figure 46. Star and cube shaped particles
42 For the rest of the study, smaller electrodes were used. Indeed, another problem could be due to a too much weak current density. This could be increased by using smaller electrodes. Besides, to simplify the process and avoid silver pollution, the EPD was made directly on the platinum electrodes, without using wafers anymore. In order to do that, both electrodes were carefully polished.
The results were this time more promising and seemed globally better than with large electrodes.
The deposition was not perfectly uniform, but a major part of the seeded area was covered with a thick layer of seeds (Figure 48). It was only at the top of the area that the seeds became scarcer. But the electrode had not been rinsed after the EPD, so we had to do it again and rinse to see if the seeds remained in place.
But with rinsing at the end of the EPD, it was as usual: almost no seeds, except at the bottom edge of the electrode (certainly because a drop of liquid remained there at the end). It is noteworthy that, when the electrode was rinsed, the boundary zone between the seeded (i.e. the area of the electrode immersed in the seeding solution during the EPD) and non-seeded areas was visible.
Indeed, the seeded area was hydrophilic whereas the area above was hydrophobic, and the water on the electrode allowed distinguishing the limit. So even though there was almost no seeds in the seeded area, there was something different between the two areas.
Figure 47. EPD 2.5 V, unrinsed wafer
43
Figure 48. EPD on the small platinum anode (from bottom to top of the seeded area).
44 Then we tried to change another parameter: the dry content of the seeding solution. From this moment, a 1wt.% colloidal dispersion was used.
First an EPD followed by rinsing was carried out. The results were not particularly interesting: as usual with rinsing, large areas with almost nothing, a few more seeds at the bottom, and, this time, a very thick pack of seeds at the top of the seeded area were observed (Figure 49). But it could simply be because of a drop which might have dried there.
Since seeds are removed by rinsing, we tried to use the cationic polymer used in part 2.4.2 to keep them on the electrode. To do so, the electrode was immersed for one hour in that polymer solution.
Then it was used as the anode for the EPD and rinsed with water.
This time seeding was more uniform and all the seeded area was covered with seeds. It looks more like a classic seeded wafer. In comparison, the seeds seemed in this case a bit better distributed on the surface, but on the other hand, there were dense areas of seeds which were not present on samples produced by classic seeding. In any case, the polymer worked well: despite rinsing with water, the seeds were still there. But it could not be really considered as an electrophoretic deposition.
Figure 49. Top of the seeded area.
45 As previously explained, it was at this stage we realized the pH-meter was giving wrong values.
Thus, another EPD was made with this time a solution which really was at a pH value around 10. It was again carried out for 40 minutes, at 2 volts and with a 1 wt.% solution. The electrode was rinsed with water at the end of the process. Once again, there were larges areas almost empty, with sometimes some pieces of very thick layers of seeds (Figure 51). It seems likely that there was a thick layer of seeds on the electrode and that most of it was chopped off during the rinsing step.
Figure 50. Electrode with cationic polymer.
46 In parallel to those trials with aqueous seeding solution, we wanted to try to make an EPD in an organic solution to increase the voltage and avoid water electrolysis issues. For this purpose, we chose to make a seeding solution in isopropyl alcohol (IPA). Starting from the aqueous solution, we used centrifugation and redispersion in IPA, four times, to get the seeding solution in IPA, with the same dry content than the aqueous seeding solution. This EPD was performed at 20 volts for 10 minutes. Then the electrode was rinsed with water.
The results were actually interesting for a rinsed electrode: there were no empty areas. The deposition was not uniform but almost all the seeded area was covered (Figure 52). For some reasons, the seeds were not, at least, completely removed during the washing. There was a thick layer of seeds at the top of the seeded area.
Figure 51. EPD with right pH value, rinsed anode.
47
Figure 52. EPD in IPA (pictures recorded from
bottom to top of the sample).
48 For aqueous EPD, we tried another technique: when the 40 minutes were elapsed, the seeding solution was slowly removed from the container with a syringe. Then the drop at the bottom of the electrode was absorbed by holding it vertically on filter paper and the electrode was let to dry. By doing this this way, very good results were obtained: a large part of the seeded area was covered with a dense monolayer of seeds (Figure 53). There were some bare areas and some thick layers (Figure 55), especially on the bottom part, but globally it was rather successful. The holes in the monolayer seemed to be often caused by the presence of some unknown black dots (Figure 54).
Figure 53. Monolayer of seeds. The solution was slowly removed with a syringe, no rinsing.
49 So the results were interesting with this technique, but it must be noted that this was not really an EPD rather dip-coating.
Another idea to improve the set-up was to put the electrodes closer so that the current can pass through the solution more easily. They were maybe too far from each other previously. The new set-up can be seen in Figure 56. The space between the two electrodes was around 2 mm.
But the results were not significantly improved by this modification.
Figure 54. Holes in the monolayer of seeds. Still the same sample: solution removed with a syringe and no rinsing.
Figure 55. Thick layers, especially at the bottom of the sample (the same as in Fig.53).
Figure 56. New set-up with closer electrodes.
50 Afterwards, we thought the problem was maybe the generator, so another EPD was made with a new generator but there still were no improvement.
Finally, we did an “EPD without electricity” i.e. exactly the same set-up was used but just without turning on the generator. The aim of this trial was to see if the current had an influence on our process or if it was really just some kind of dip-coating as we suspected. As shown in Figure 57, the pictures are not very different from those shown previously.
So, our assumption was good: something is wrong in the set-up and we surely never really made EPD on our different samples. That is why the seeds are simply removed by rinsing. Maybe the polished platinum surface was actually too smooth as there were always a lot of seeds in scratches.
Figure 57. "EPD without electricity".
51
Conclusion
In this thesis, we spent a lot of time on electrophoretic deposition, always trying to make it work.
And for some still unknown reasons it never did. According to Berenguer and his team, who also worked on EPD of silicalite-1 (15), the seeds would actually be deposed on the cathode because of the TPA+ cations preferentially adsorbed around the zeolites seeds. This should be studied. It would actually explain why the seeds sometimes seem not to like the cationic polymer. EPD at acidic pH should also be tried, maybe it would work better than with basic pH.
Besides, it was very interesting to follow the growth of the film after different hydrothermal treatment times. It really seems that the nanoparticles that could be seen in the sample after 8 hour synthesis were taking part to the growth process.
The multi-seeding method is very promising since it gives a very dense deposition of seeds with good reproducibility. The drawback is that the deposition is slightly too thick. But repeating seeding ten times is not really necessary to obtain a surface totally covered. Maybe four or five times would be sufficient.
Seeding on polished alumina discs also produced interesting results. It was not perfect because the fine layer was removed and thus there were large holes between alumina grains. But if it can be polished without removing it, then it could be very promising.
And finally, if the EPD did not work as it was expected, at least we obtained a very good seeding when we did some kind of dip-coating. So, maybe it would be a good idea to work further on it.
Carefully controlled, it could really give what we aimed for. But the drawback is that the seeds do not seem to be firmly attached to the support.
52
Bibliography
1. S. M. Auerbach, M. H. Ford and P. A. Monson. New insights into zeolite formation from molecular modeling. Curr. Opin. Colloid Interface Sci. 2005, 10, pp. 220-225.
2. Valtchev, Lubomira Tosheva and Valentin P. Nanozeolites: Synthesis, Crystallization Mechanism, and. Chem. Mater. 2005, 17, pp. 2494-2513.
3. C. E. A. Kirschhock, R. Ravishankar, P. A. Jacobs, and J. A. Martens. Aggregation mechanism of nanoslabs. J. Phys. Chem. B. 1999, 103, pp. 11021-11027.
4. Duoduo Liang, Lana R. A. Follens, Alexander Aerts, Johan A. Martens, Gustaaf Van Tendeloo, and Christine E. A. Kirschhock. TEM Observation of Aggregation Steps in Room-Temperature Silicalite-1 Zeolite. J. of Phys. Chem. C. 2007, 111, pp. 14583-14285.
5. David Lesthaeghe, Peter Vansteenkiste, Toon Verstraelen, An Ghysels, Christine E. A. Kirschhock, Johan A. Martens, Veronique Van Speybroeck, and Michel Waroquier. MFI Fingerprint: How
Pentasil-Induced IR Bands Shift during Zeolite Nanogrowth. J. Phys. Chem. C. 2008, 112, pp. 9186- 9191.
6. An Interdisciplinary Approach to Understanding the Growth of Nanoporous. P. A. Monson, S. M.
Auerbach, M. Tsapatsis, R. Lobo, D. G. Vlachos. Dec 11-13, 2002. NSF Nanoscale Science and Engineering Grantees Conference. Grant # : 0103010.
7. Begüm Tokay, Mehmet Somer, Ayse Erdem-Senatalar, Ferdi Schüth, Robert W. Thompson.
Nanoparticle silicalite-1 crystallization from clear solutions: Nucleation. Microporous and Mesoporous Materials. 2009, 118, pp. 143-151.
8. Sandeep Kumar, Zhuopeng Wang, R. Lee Penn and Michael Tsapatsis. A Structural Resolution Cryo-TEM Study of the Early Stages of MFI Growth. J.AM.CHEM.SOC. 2008, Vol. 130, 51, pp. 17284- 17286.
9. Regrev, Brian J. Schoeman and Oren. A study of the initial stage in the crystallisation of TPA- silicalite-1. Zeolites. 1996, 17, pp. 447-456.
10. Schoeman, Brian J. Analysis of the nucleation and growth of TPA-silicalite-1 at elevated
temperatures with the emphasis on colloidal stability. Microporous and mesoporous materials. 1998, 22, pp. 9-22.
11. Sandeep Kumar, Tracy M. Davis, Harikrishnan Ramanan, R. Lee Penn and Michael Tsapatsis.
Aggregative Growth of Silicalite-1. J. Phys. Chem. B. 2007, Vol. 111, pp. 3398-3403.
12. S. Yang, A. Navrotsky, D.J. Wesolowski, J.A. Pople. Chem. Mater. 2004, 16, p. 210.
13. C. Ke, W. L. Yang, Z. Ni, Y. J. Wang, Y. Tang, Y. Gu and Z. Gao. Electrophoretic assembly of nanozeolites: zeolite coated fibers and hollow zeolite fibers. Chem. Commun. 2001, pp. 783-784.
53 14. Bovornlak Oonkhanond, Michael E. Mullins. The preparation and analysis of zeolite ZSM-5.
Journal of Membrane Science. 2001, 194.
15. M. Abdollahi, S.N. Ashrafizadeh, A. Malekpour. Preparation of zeolite ZSM-5 membrane by electrophoretic. Microporous and Mesoporous Materials. 2007, 106.
16. Wei Shan, Yahong Zhang, Wuli Yang, Chen Ke, Zi Gao, Yunfen Ye, Yi Tang. Electrophoretic deposition of nanosized zeolites in non-aqueous medium and its application in fabricating thin zeolite membranes. Microporous and Mesoporous Materials. 2004, 69, pp. 35-42.
17. A. Berenguer-Murcia, E. Morallon, D. Cazorla-Amoros, A. Linares-Solano. Preparation of thin silicalite-1 layers on carbon materials by electrochemical methods. Microporous and Mesoporous Materials. 2003, 66, pp. 331-340.
18. Torsten Mäurer, Steffen P. Müller, and Bettina Kraushaar-Czarnetzki. Aggregation and Peptization Behavior of Zeolite Crystals in Sols and Suspensions. Ind. Eng. Chem. Res. 2001, 40, pp.
2573-2579.
19. Vladimiros Nikolakis, Efrosini Kokkoli, Matthew Tirrell, Michael Tsapatsis, and Dionisios G.
Vlachos. Zeolite Growth by Addition of Subcolloidal Particles:. Chem. Mater. 2000, 12, pp. 845-853.
