Preparation of hafnia ceramic membranes for ultrafiltration

Preparation of hafnia ceramic membranes for ultrafiltration

journalof MEMBRANE SCIENCE Journal of Membrane Science 134 (1997) 109-115 ELSEVIER Preparation of hafnia ceramic membranes for ultrafiltration E B l...

508KB Sizes 0 Downloads 10 Views

journalof MEMBRANE SCIENCE Journal of Membrane Science 134 (1997) 109-115


Preparation of hafnia ceramic membranes for ultrafiltration E B l a n c , A. Larbot*, M. Persin, L. C o t Laboratoire des Mat~riaux et Proc~d~s Membranaires, CNRS - UMR 5635, Ecole Nationale Sup~rieure de Chimie de Montpellier. 8 rue de l'Ecole Normale, 34053 Montpellier Cedex 1, France

Received 23 October 1996; received in revised form 7 April 1997;accepted 16 April 1997

Abstract Stable sols of monoclinic hafnia have been prepared hydrothermally. The reaction conditions can be easily defined by controlling different parameters such as the temperature, the autoclaving time, the concentration of precursor in solution, the filling rate of the autoclave, and the pressure. Hafnia particles are characterized using several methods such as turbidity, granulometry, and microscopy. Particle charge, depending on the electrolyte concentration, is determined by electrophoresis measurements. The hafnia membranes, prepared from the sols, are characterized in static and dynamic conditions. The membrane properties are defined by pore diameters, cut offs, and water permeation. Keywords: Ultrafiltration; Ceramic membranes; Hafnium oxide; Hydrothermal synthesis

1. Introduction Among the materials used as technical ceramics, hafnia belongs to a class of oxides of refractory materials. Hafnia presents a melting point near 2750°C and its first crystalline phase, the monoclinic form, is stable up to 1850°C. This characteristic could be very interesting for high-temperature filtration experiments, such as molten-metal filtration. Hafnia is isomorphous due to the lanthanide contraction such as in the case of zirconia [1]. Hafnia has a higher monoclinic-to-tetragonal transformation temperature (1850°C against ~ 1000°C), a lower volume variation (3.4% for hafnia against 7.5% for zirconia) besides a high chemical resistance in the entire range of pH [2]. Hafnia powders have been prepared by the sol-gel process [3], the precipitation method [4] and hydro*Corresponding author. 0376-7388/97/$17.00 © 1997 Elsevier Science B.V. All rights reserved. P I I S0376-7388(97)00097-5

thermal synthesis [5,6]. Here, we prepare hafnia using the hydrothermal method, allowing us to obtain a crystallized powder at low temperatures. Then, the membrane preparation, from the hafnia colloidal suspension, with the addition of water and organic binder, followed by the steps of coating, drying and sintering will be carried out. The membranes are then characterized by several methods. Lack of defects is confirmed, by SEM, in the texture of the layer. The values of pore diameters justify the results of water permeation and retention of model molecules.

2. Experimental 2.1. H y d r o t h e r m a l s y n t h e s i s o f h a f n i a s u s p e n s i o n

The precursor aqueous hafnium oxychloride, HfOC12.8H20, was heated under pressure in an


P. Blanc et al./Journal of Membrane Science 134 (1997) 109-115

autoclave. Under critical conditions of pressure and temperature, the hydrolysis of the aqueous solution takes place and forms hafnia particles and the acid [7]. Hafnium oxychloride reacts quickly with water to form salts called hafnyl salts. The cationic HfO 2+ species form a tetramer whose stoichiometry is well defined [Hf(OH)2.4H20]48+. The four hafnium atoms are located at the corners of a distorted square [8-10]. These atoms are linked together by -OH bridges above and below the plane of this square. The coordination of the hafnium atoms is completed by the presence of four water molecules bounded to each of them. The solution is then heated and the hafnia suspension is produced [11]. A turbid sol or a white precipitate is produced during the hydrothermal processing. This reaction may be influenced by the pressure, the temperature, the synthesis time, and the salt concentration in solution. The temperature and the pressure are fixed at 180°C and at 10 bars for a filling rate of 70% of the autoclave. The concentration is fixed at 150 g/1 by equivalent oxide (HfO2) in solution. When the inner temperature of the autoclave reaches 180°C, the autoclaving time was varied from 5 min to 3 h 15 min. The autoclave used is simply a cylinder of stainless steel protected by Teflon, a tube closed at one end with a thermocouple and a pressure gauge.

2.2. Characterization of the hafnia powder The hydrothermal synthesis depends only on the autoclaving time. For 5 min, the crystallization does not occur and a transparent solution is obtained. From 10min to 3 h 15 min, the resulting suspensions become more and more turbid and the settling rate increases with the reaction time. This means that the particle size increases with the autoclaving time. The X-ray diffraction patterns, from powders dried at 50°C, show that monoclinic hafnia is obtained, whatever the autoclaving time. The turbidity of the solution is measured after dilution by a factor of two. The values are given in Table 1. Fig. 1 shows the plot of the particle size against reaction time. Measurements, based on the principle of liquid-phase sedimentation, using the optical transmission method, are performed with a granulometer (Horiba Capa 700). Solutions diluted by a

Table 1 Turbidity of twice diluted solution as a function of reaction time Reaction time (min)

Turbidity (NTU)

15 20 25 75

94 570 1130 1940










Particle diameter (nm)



Fig. 1. Panicle size and particle-size distribution vs. autoclaving time.

factor of four are used, just as in case of membrane preparation. Because of the presence of a transparent solution for an autoclaving time > 10 min, we concluded that no particles are formed. From 10 rain to 3 h 15 min, particle size as well as particle-size distribution increase with the autoclaving time. The specific surface area, determined by the BET technique (Micromeritics ASAP 2000), is equal to 80 m2/g. This high value is confirmed by the scanning electron micrograph shown in Fig. 2. For this powder, obtained after 25 min of autoclaving time, each grain is formed by the agglomeration of very small crystallites, although grains are not sintered. The electric charge at the surface of the material is an important parameter for controlling the agglomeration of particles and for determining the filtration membrane properties. The charge depends on the aqueous medium, the pH value, and the nature and concentration of the electrolytes. In the case of NaC1, when the concentration increases from 10 -3 to 10 -2 M, a shift of the pH value of the iso-electric

P. Blanc et al./Journal of Membrane Science 134 (1997) 109-115


loidal suspension. An organic binder, the poly (vinyl alcohol) (12% in aqueous solution), is added in the sol to adjust the viscosity. The sol is then deposited on a macroporous support by slip-casting. After drying at room temperature, the gel layer is sintered at different temperatures, between 500 ° and 1200°C. This procedure allows us to obtain two types of crack-free membranes. We used several different hydrothermal hafnia powders. We will call P25 the powder autoclaved for 25 min, and M25 the membrane realized from this powder. In this way, P45 is the powder autoclaved for 45 min and M45, the membrane realized therefrom. Fig. 2. Powder of 25 min of autoclaving time.

3. Results and discussion

3.1. Characterization of the membranes E



E la p, o



-0.2 -0.3 0.3

Fig. 3. Electrophoretic mobility vs. pH (10 3M and 10-2M NaC1).

point (IEP) is observed from 6.2 to 6.9. This phenomenon corresponds to a specific adsorption of sodium ions at the surface of the particles, and the range of pH where the membrane material is positively charged is larger. Fig. 3 gives the result of this measurement, which is realized with a Rank Brother zetameter.

The pore diameters of the membranes are determined by mercury porosimetry (Micromeritics Autopore II) and results are confirmed by nitrogen adsorption-desorption measurements using the Barrett-Joyner-Halenda (BJH) method. However, only the desorption isotherm is taken into account. Fig. 4 shows the influence of sintering temperature on membrane pore diameters. Pore diameters of M25, smaller than those of M45, undergo a larger increase when the temperature rises due to a higher reactivity of the powder. Pore diameters of both the membranes converge to a common value, near 65 nm, at 1100°C. For M45 and M25 membranes, sintered 1 h at 600°C, the pore diameter values are equal to 35 and 25 nm, respectively. 80 70 6O c 5O

2.3. Preparation of hafnia ultrafiltration membranes

i,o ~ 3O

Hafnia ultrafiltration membranes are prepared with the help of classical process of membrane preparation using a hafnia colloidal suspension. A sol is obtained with the addition of distilled water to a hydrothermal suspension, which enables the reaction of peptization. The sol is prepared from the four times diluted col-

20 t0








11 O0

t 200


Sintedng temperature (°C)

Fig. 4. Pore diameters of M45 and M25 vs. sintering temperature.

P. Blanc et al./Journal of Membrane Science 134 (1997) 109-115

112 00

70 60

~ SO 4O

~ 3o 20



10 0 4O0








(~ -

lh sintering


fOh slntering




Sintering temperature (*C)

Fig. 5. Pore diameter of a M45 vs. sintering time and temperature. Fig. 7. Surface of a M25 (600°C, 1 h).

For high-temperature applications, it is interesting to study the behaviour of the membrane during an extended sintering. Fig. 5 shows the behaviour of the M45 membrane for two sintering times, namely 1 h and 10 h. Above 600°C, a slight increase of 5 nm in the pore diameters can be observed for a 10 h sintering time. The pore size does not appear too much affected by a longer sintering time at higher temperatures and could be used for high-temperature filtration. The characterization of the membrane texture has been performed using a Hitachi S-4500 scanning electron microscope. The morphological aspect of the M45 and M25 layers can be observed, respectively, in Figs. 6 and 7. These show different textures as a function of the granulometry of the powder. The

Fig. 8. Surface of a M45 (1200°C, 1 h).

thicknesses of the membranes are measured at an average value of 0.7 ~tm. Figs. 8 and 9 allows us to observe the texture of these membranes after sintering at a high temperature. As shown in micrographs, the texture layer is not affected by this treatment.

3.2. Dynamic characterization of the two ultrafiltration membranes

Fig. 6. Surface of a M45 (600°C, 1 h).

The water-permeation measurements are carried out on a filtration pilot in the laboratory, using tap water. The membranes, sintered at 600°C for 1 h, present a tubular configuration and a length of 180 mm. A cooling system allows us to maintain the temperature


P. Blanc et al./Journal of Membrane Science 134 (1997) 109-115


350 300





J . . . .

m - | ....

I , - ~_~ ,11. . . .



I ,90 80

~ 250





- -<~. - ~ . . . .









~ 15o







¢p ~


- - nnool











Time (minute)

Fig. l 1. Variation of permeation flux (In; A) for the M25 membrane. Fig. 9. Surface of a M25 (1050°C, 1 ).






n M 25

~= ~ 2so

M 45

~...,_ ^ ~ - m




~ []










600 ~ ~



150 100





~ ~ ~"

and retention rate



900 800






.T: [

~ soo i

0 0






Pressure (bar)

Fig. 10. Variation of permeation flux as a function of working pressure.

at 20°C, the working pressure is between 3 and 9 bar, and the fluid circulation speed is 3.2 m / s . Fig. 10 shows water permeation for two membranes as a function of working pressure. Membranes are first conditioned in water for 24 h before filtration in order to establish interactions between the solute and the surface of the membrane and avoid a decrease of flux. These filtration tests confirm that the pore diameters of the M45 membrane are larger than those of the M25 membrane. Permeabilities of the two membranes are not given because no measurements were performed at a pressure lower than 3 bar, and it is difficult to predict if the value of the stabilized flux is equal to zero when the transmembrane pressure vanishes. The retention rate measurements are performed at a pressure of 4 bar and temperature of 20°C. The initial

~ ~"


o 2e0 L

100 0




100 Time (minute)



Fig. 12. Variation of permeation flux ([7; A) and retention rate (ll; A) for the M45 membrane. concentration of molecules is fixed at 1 g / l . The model molecules used are: Dextran 500 (Mw = 500 × 103 dalton, Aldrich catalog ref. D 5251); and Dextran 260 (Mw = ( 2 0 0 - 3 0 0 ) x 103 dalton, Aldrich catalog ref. D 7265). The permeate concentrations are determined by liquid-phase chromatography with a Waters R 401 refractometer. Figs. 11 and 12 show the evolution of retention rate and flux as a function of time. Whatever the Dextran used, flux is stable or slightly decreased and retention rate is stable or slightly increased during the filtration tests. 3.3. Discussion

The main characteristics of M45 and M25 hafnia ultrafiltration membranes have been summarized in the Table 2.


P Blanc et al./Journal of Membrane Science 134 (1997) 109-115

Table 2 Characteristics of hafnia ultrafiltration membranes Type of membrane

Pore diameter (nm) (sintering temperature = 600°C)



Change in dextran permeation flux (1/h m 2) during filtration at 4 bar

Water permeation flux (1/h m 2) at 4 bar

Dextran 500 : 210-170

D 500 : 89-92% 520

Dextran 260 : 230-210

D 260 : 68-70% 500 D 500 : 87-90%

Dextran 500 : 190-190 M25

Change in retention rate (%) during filtration and molecular weight cut-off (× 103 dalton)


420 Dextran 260 : 260-170

We can note that the water permeation flux (at 4 bar) is higher for the M45 than for the M25 membrane, due to the lower pore diameter of the latter. The cut-offs are the same for the M45 and M25 membranes, but retention rates are different. For M25, the D 500 permeation flux is stable due to the concentration polarization at the surface of the membrane and the D 260 permeation flux decreases whereas the retention rate increases slightly, probably due to a fouling phenomenon. For M45, the decrease of the D 500 permeation flux is most likely due to a fouling of the pores and the presence of a slight fouling phenomenon for the D 260 permeation flux is also possible. The smaller size of the D 260 molecule compared to the D 500 one, most likely explains the lower retention rate and the slight decrease of the permeation flux. We can arrive at the conclusion that there is most likely a competition between two phenomena, fouling and concentration polarization. As explained in the literature from the osmotic pressure model [ 12,13], the latter is an important mechanism for limiting flux in ultrafiltration. In fact, these two phenomena may exist at the same time and it is difficult to establish which one predominates [14]. In the context of our work, we are in the process of studying more closely the influence of these two phenomena.

D 260 : 70-75% 500

ized by an easy formation procedure, provides fine and high purity powders. These are formed by unagglomerated grains showing a narrow grain-size distribution that depends on the autoclaving time. Moreover, the behaviour of the material during high-temperature treatment, without damaging the texture of the membranes, allows us to prepare ceramic membranes with a large range of pore diameters, spanning the ultrafiltration domain. Two membranes have been prepared and characterized by pore diameters, water permeation tests, and cut-off. Another membrane can be synthesized, using the P20 powder and then deposited on a M45 or M25 layer. This membrane presents very small pore diameters, near 4 nm. Water permeation flux, at 4 bar, is equal to 150 1 / h m 2 and the cut-off is near 70 x 103 dalton. The cut-off of this M20 membrane shows a good improvement compared with those of M45 and M25 membranes. Work is in progress to improve the texture of this layer, which could constitute an interesting support for a nanofiltration membrane.

Acknowledgements We are grateful to the EEC for financial support. Grant N ° AVI-CT92-0014.

4. Conclusion Monoclinic hafnia powders prepared from hydrothermal synthesis appear as a very interesting material for membrane preparation. This reaction, character-

References [1] J. Wang, H.P. Li and R. Stevens, Review - Hafnia and hafnia toughened ceramics, J. Mater. Sci., 27 (1992) 5397.

P. Blanc et al./Journal of Membrane Science 134 (1997) 109-115

[2] Pascal, Nouveau trait6 de Chimie Minrrale, Mason, Paris, 1963, pp. 945-996. [3] T. Hours, Elaboration et caracterisation de poudres d'oxyde d'yttrium et d'oxyde d'hafnium par proc~d~s sol-gel, Th~se, Montpellier, 1988. [4] S.L. Dole, R.W. Scheidecker, L.E. Shiers, M.E Berard and O. Hunter Jr., Technique for preparing highly sinterable oxide powders, Mater. Sci. Eng., 32 (1978) 277. [5] S. Somiya, Hydrothermal preparation and sintering of fine ceramic powders, Mat. Res. Soc. Symp. Proc., 24 (1984) 255. [6] S. Somiya, Hydrothermal preparation of fine powders, Adv. Ceram. 3rd Meet, 1988, Elsevier, London, 1990, Chap. 11, p. 207. [7] G. B. Alexander and J. Bugosh, Concentrated ZrO2 and HfO2 aquasols and their preparation, US Patent 2984628, 1961. [8] C.J. Norman, S.L. Jones and B.M. Leigh, The preparation of zirconia powders, Trans. J. Brit. Ceram. Soc., 83(6) (1984) 173.


[9] A. Clearfield, Structural aspects of zirconium chemistry, Rev. Pure Appl. Chemistry, 14 (1964) 96. [10] A. Clearfield, The mechanism of hydrolytic polymerization of zirconyl solutions, J. Mater. Res., 5(1) (1990) 161. [11] V.A. Kuznetzov, Crystallisation of titanium, zirconium and hafnium oxides and some titanate and zirconate compounds under hydrothermal conditions, J. Crystal Growth, 3(4) (1968) 405. [12] V.L. Vilker, C.K. Colton, K.A. Smith and D.L. Green, The osmotic pressure of concentrated protein and lipoprotein solutions and its significance to ultrafiltration, J. Memb. Sci., 20 (1984) 63. [13] J.G. Wijmans, S. Nakao and C.A. Smolders, Hux limitation in ultrafiltration: Osmotic pressure model and gel layer model, J. Memb. Sci., 20 (1984) 115. [14] V. Gekas, P. Aimar, J.P. Lafaille and V. Sanchez, A simulation study of the adsorption-concentration polarization interplay in protein ultrafiltration, Chem. Eng. Sci, 48(15) (1993) 2753.