-
American Fisheries Society Symposium 38:387–401, 2003© 2003 by
the American Fisheries Society
Sepiolite Membrane for Ultrafiltration
Q. K. WANG1, TAKESHI MATSUURA2, C. Y. FENGIndustrial Membrane
Research Institute, Department of Chemical Engineering
University of Ottawa, Ottawa, Ontario K1N 6N5, Canada
M. R. WEIR, CHRISTIAN DETELLIEROttawa-Carleton Chemistry
Institute, Department of Chemistry
University of Ottawa, Ottawa, Ontario K1N 6N5, Canada
E. RUTIDINKA, R. LE VAN MAODepartment of Chemistry and
Biochemistry, Laboratories for Inorganic Materials
Concordia University, Montreal, Quebec H3G 1M8, Canada
387
tion of sepiolite membranes by pore size and poresize
distribution is another objective.
Many works have been published on the poresize and pore size
distribution of synthetic mem-branes (Michaels 1980; Zeman and
Wales 1981;Singh et al. 1998). One of the methods of determin-ing
pore size and pore size distribution is based onthe separation of
solutes of known sizes. Michaelsconcluded that the lognormal
probability functionwas generally accounted as means for
describingsieving curves for ultrafiltration membranes, and
acomplete sieving curve could be constructed for agiven membrane
using only two experimental siev-ing coefficient values for two
different solutes ofknown Einstein–Stokes radius. On the basis of
thisobservation, Michaels proposed a method to deter-mine pore size
and pore size distribution. Singh et al.studied various sulfonated
poly(2,6-dimethyl-1,4-phenylene oxide) membranes and found that
thesieving coefficients and Einstein–Stokes radii werecorrelated
well with the lognormal distributionfunction. Aimar et al. (1994)
described the prob-lems that may be encountered when developing
amethod for membrane characterization based onsolute sieving (e.g.,
concerning macromoleculartransport through capillaries). They also
showed theapplicability of the lognormal curve to fit the
soluteseparation–solute size correlation.
Scanning electron microscopy (SEM) is apowerful tool for
investigating membrane struc-ture. However, because of the low
conductivity ofthe membrane surface, the sample must be coatedwith
a heavy metal, and the coating process maycause some damage to the
membrane. Therefore,SEM is not a reliable method for measuring
poresize (Hsieh et al. 1979; Aimar et al. 1994). Hence,in this
work, membrane pore size and pore size dis-tribution are determined
by a method proposed byMichaels (1980).
In recent years, a great deal of research has beendevoted to the
development of a new kind of inor-ganic membrane that exhibits
improved resistanceto heat, chemicals, and corrosion. Rapid
develop-ment and innovation have already been realized inthis area
(Cot 1998). Clay minerals are a well-known class of naturally
occurring inorganic mate-rials with well-known structural
adsorption, rheo-logical, and thermal properties (Nagata et al.
1974;Serna et al. 1975; Jones and Galán 1988; Pérez-Rodríguez and
Galán 1994). Research on clay as amembrane material has
concentrated mainly onpillared clays (Cool et al. 1997; Mishra and
Parida1997). Studies of membranes prepared entirelyfrom clay have
begun (Ishiguro et al. 1995; Le VanMao et al. 1999).
Sepiolite, one of the most important gel-form-ing clays, can
give rise to stable suspensions of highviscosity at relatively low
concentrations. It is char-acteristically fibrous as observed under
an electronmicroscope (Jones and Galán 1988). Its structuraland
morphological changes that occur on heatingcan be divided into
three phases: low-temperature(600°C) regions. In the
high-temperatureregion, dehydroxylation of the structure takes
placeat about 800°C, together with a change in entropydue to
structural collapse (Jones and Galán 1988).This thermal behavior
suggests that there is a limi-tation in temperature for sintering
the membrane.
The objective of the present study was toestablish the
preparation procedure for pure sepio-lite membranes and to
investigate their propertiesand potential for ultrafiltration. The
characteriza-
1Present address: Department of Food Science and
Engineering,Dalian Fisheries University, Heishijiao, Dalian 116023,
PRChina. E-mail: [email protected] or
[email protected] author. E-mail:
[email protected]
-
388
TheoreticalMean pore size and pore size distribution
According to the equation for solute separation (f; in %),
ƒ = (1–Cp/Cf)×100 (1)
where Cp and Cf are the solute concentration in thepermeate and
in the feed solution, respectively. InEquation (1), the effects of
dispersive or electro-static interactions and concentration
polarizationare not considered. Solute separation of an
ultrafil-tration membrane, plotted versus the solute diame-ter on a
lognormal probability paper, can yield alinear relationship
(Michaels 1980). If solute diam-eter correlates with the solute
separation accordingto the lognormal probability function, then
therelationship is given as
ƒ = erf (z) = —— ∫z e 2 du (2)
where
lnds –ln µsz = ————— (3)lnσg
and ds is the solute diameter, µs is the geometricmean diameter
of a solute molecule, and σg is thegeometric standard deviation
about the meandiameter. On lognormal probability
coordinates,Equations (2) and (3) linearize in the form of
F(ƒ) = A0 + A1(1nds) (4)
where A0 and A1 are the intercept and the slope,respectively
(Michaels 1980); µs can be determinedfrom ds corresponding to f =
50%, and σg can bedetermined from the ratio of ds at f = 84.13%
andat 50%. By neglecting the effects of steric andhydrodynamic
interaction between solute andpore, the mean pore size (µp) and the
geometricstandard deviation (σp) of the membrane can beconsidered
to be the same as µs and σg, respective-ly (Michaels 1980; Ishiguro
et al. 1995). Depend-ing on µp and σp, the pore size distribution
of anultrafiltration membrane can be expressed by thefollowing
probability density function (Youm andKim 1991):
df(dp) 1 (lndp –lnµp)2—— = —————— exp [– —————] (5)ddp dp (lnσp)
���� 2(lnσp)2
where dp is the pore diameter.
Pore density and surface porosity
Pore density can be calculated from the permeabil-ity data of
the membrane through the Hagen–Poiseuille equation. According to
this equation,the solvent flux (Ji) through the pores of diameterdi
can be expressed as
Niπdi4∆pJi = —————128η δ
where Ni is the number of pores per unit area hav-ing the
diameter di, δ is the length of the pores, ∆pis the pressure drop
from one end to the other endof the pore, and η is the solvent
viscosity. The totalflux through membrane is the sum of all the
fluxesthrough the pores of different sizes; therefore,
J = Σ Jiπ∆pJ = ——— (N1d14 + N2d24 + N3d34 + Nndn4)
128η δ
π∆pJ = ———— (ƒ1Nd14 + ƒ2Nd24 + ƒ3Nd34 + ƒnNdn4)128η δ
π∆pN d maxJ = ————— Σ fidi4 (6)128η δ d min
where N is the number of pores per unit area andcalled pore
density and fi is the fraction of the num-ber of pores with
diameter di. From Equation (6),the total number of pores can be
expressed as
128ηδJN = ————— (7)d max
π∆p Σ fidi4d min
The pore length δ is assumed to be equal to thethickness of the
membrane based on the consider-ation that the solvent can pass
through only thepores with the shortest length; otherwise, the
poreswill be dead ended.
The surface porosity (Ps), defined as the ratiobetween the area
of pores to the total membranesurface area, can be calculated
as
Nπd max
Ps = ( —— Σ fidi2) × 100 (8)4d min
Einstein–Stokes radius
Einstein–Stokes radius is used as a parameter tocharacterize the
size of a macrosolute in the ultra-
WANG ET AL.
1����2π
�u2__– �
2π
-
389SEPIOLITE MEMBRANE FOR ULTRAFILTRATION
filtration study. It is defined as the “apparent equiv-alent
spherical radius” of the macromolecule byusing Einstein–Stokes
equation
DAB = kT/6πηα (9)
where DAB is diffusivity, k is Boltzmann’s constant,and α is the
Einstein–Stokes radius. Therefore, ds asmacrosolute diameter can be
calculated from α.
Singh et al. (1998) have given the followingequations to obtain
Einstein–Stokes radius of poly-ethylene glycol (PEG) and
polyethylene oxide (PEO)solutes from their molecular weights. For
PEG,
� = 16.73 × 10–12 M0.557 (10)
And for PEO,
� = 10.44 × 10–12 M0.587 (11)
Experimental MethodsPurification of sepiolite
Brute sepiolite was purchased from the Source ClayMinerals
Repository, University of Missouri–Columbia. Crude sepiolite (30 g)
was ground witha small amount of water to form a thick paste,
thenadded to 400 mL of distilled water and stirredovernight. The
suspension was centrifuged. Thesupernatant was discarded, and the
clay particleson the top layer of the sediment were scraped offwith
a spatula.
The clay particles were suspended in distilledwater and
acidified with 1 M HCl to pH 3.5 (todestroy the carbonates), then
centrifuged. The sed-iment was washed with dilute HCl several
times,centrifuged, and finally dispersed in distilled water.The pH
of the suspension was increased to 8 byadding 0.1 M NaOH and
leaving it overnight forsedimentation. The upper layer of the
suspensionwas removed. Water (200 mL) was added to theremaining
sediment and stirred to homogenize.The saturated NaCl solution was
added, andexchange occurred overnight. The suspension
wastransferred to dialysis bags to expel the chlorideions. The
contents in the dialysis bags were finallycentrifuged and dried at
60°C.
Measurement and observation of sepiolite particle size
Two drops of the suspension sample were placed onthe glass
slides. The sepiolite particles wereobserved by using an
Olympus-BX40 optical micro-
scope connected to a Polaroid model DMCI digitalmicroscope
camera.
Membrane preparation
The aqueous suspension of clay fibers was preparedby dispersing
0.25 or 0.5 g of sepiolite in 30 or 60mL of distilled water. The
mixture was stirred witha magnetic stirrer for about 24 h or
subjected toultrasonic agitation for about 30 s. Then, the
sus-pension was poured into a 9-cm-diameter petri dishand left at
room temperature to evaporate to dry-ness. The air-dried membrane
was calcined at120°C for 2 h, then the temperature was increasedto
550°C for 5 h.
The membranes were labeled as 0.5A, 0.25A,0.5B, and 0.25B, where
the numbers indicate theamount of sepiolite used and the letters
representthe method of dispersion (A, magnetic stirring;
B,ultrasonic wave). Although the mechanicalstrength of sepiolite
membrane was not measured,the membranes were self-sustaining during
theultrafiltration experiments.
Ultrafiltration
Ultrafiltration was carried out in a dead-end stirredcell
configuration. The cell was pressurized withnitrogen, and the
transmembrane pressure was readon a calibrated gauge. The feed
concentration wasassumed to remain constant during
ultrafiltration,because small amounts of permeate were
collected.Five solutes were used: PEG with 35,000 molecularweight
(MW) and PEO with 100,000, 200,000,300,000, and 400,000 MW. Feed
concentrationwas 200 ppm in each experiment. The permeatewas
collected for a predetermined period to meas-ure the permeation
rate. The total organic carbonwas measured with a Folio DC-190
Total OrganicCarbon Analyser, and the solute separation
wascalculated as in Equation (1).
Scanning electron microscopy
Scanning electron microscopic images wereobtained by using a
JEOL 6400 scanning electronmicroscope, and the images were captured
digitallyon a personal computer.
Results and DiscussionPurity of sepiolite and structure of
sepiolitesuspension
The purity and structure of sepiolite were examinedby X-ray
diffraction (Philips pw3710 diffractome-ter). The X-ray powder
patterns showed that themajor impurities in the crude material are
calcite,
-
dolomite, and quartz, which were removed by acidtreatment and
sedimentation during purification.The main fraction on the powder
X-ray diffractionpatterns of purified material is at 12.1 Å, which
is acharacteristic reflection associated with sepiolitestructure
(Serna et al. 1975), whereas the calcite,dolomite, and quartz are
no longer detectable.
The sepiolite fibers observed under the elec-tron microscope are
shown in Figure 1. The lengthof the fiber was 1.0–7.0 µm, and the
diameter was0.2–0.4 µm. The ultrasonified sample was morehighly
dispersed than the magnetically stirred sam-ple. In the latter,
aggregated sepiolite fiber bundleswere apparent.
The membrane thickness before and after cal-cination was
measured by using a micrometer;results are listed in Table 1. The
membranes made
by applying magnetic stirring were thinner thanthose made by
applying ultrasonic wave. This dis-crepancy probably results
because less fiber aggre-gation occurred when the suspension was
sonified.As a result, the porosity of membranes preparedfrom
ultrasonified dispension was greater.
Ultrafiltration
The permeate flux changed almost linearly withapplied pressure
(30–120 pounds per square inchgauge [psig]; Figure 2). The flux
depends mainly onthe thickness of the membrane. For the
membranesmade from magnetically stirred suspensions, thethinner the
membrane, the higher the permeateflux. Similarly, for the membranes
made from soni-fied suspensions, the thickness of membrane is
the
390 WANG ET AL.
Figure 1. Sepiolite fibers: magnetic stirred suspension (a) and
sonified suspension (b).
Table 1. Thickness of sepiolite membranes before and after
calcination.
Membrane Thickness (µm)
Magnetically stirred
0.5A 97 before85 after
0.25A 52 before43 after
Sonified
0.5B 123 before114 after
0.25B 64 before57 after
-
main factor that controls flux rate. The membranesmade by
sonified suspension tend to get higher fluxthan the membranes made
by magnetic stirring,probably because of the higher porosity in
themembranes made by sonified suspension.
Even though the flux was lowered significant-ly in the presence
of PEG and PEO solutes, this
391SEPIOLITE MEMBRANE FOR ULTRAFILTRATION
effect was not permanent, because the flux of waterobtained
after filtration experiments with PEG andPEO became closer to the
initial water flux.
Mean pore size and pore size distributions
Figure 3 illustrates the separations of PEG and PEOof different
molecular weights. Although the flux is
Figure 2. Flux rates of four different sepiolite membranes
(0.5A, 0.25A, 0.5B, and 0.25B) calcined at 550ºC.
-
paper. High correlation coefficients were obtained(r2 =
0.92-0.96). The slope of the line indicates thepore size
distribution, and all slopes are quite similar.
Table 2 lists the geometric mean pore size andthe geometric
standard deviation for the sepiolitemembranes. Again, the values of
the geometricmean pore size (23–26 nm) and the geometric
392 WANG ET AL.
quite different, the separation of all these mem-branes is quite
similar.
The pore size and the pore size distribution ofmembranes were
calculated from the data in Figure3. As illustrated in Figure 4,
straight lines weredrawn between the separation and
Einstein–Stokesdiameters of solutes using a lognormal
probability
Figure 2. continued
-
standard deviation (1.91–2.04) are quite similar.The different
dispersion means for membranepreparation did not obviously affect
the pore sizes.This finding demonstrates that sepiolite mem-branes
can be quite easily prepared, and the poresizes can be kept stable
even when the preparationmethod is altered.
Figure 5 illustrates the cumulative pore sizedistributions for
different membranes. More than80% of the pores have diameters less
than 50 nm.The largest pore size is more than 130 nm.
Probability density function curves are illus-trated in Figure 6
by using the values of mean poresize and geometrical standard
deviation. All the
393SEPIOLITE MEMBRANE FOR ULTRAFILTRATION
Figure 3. Separations of four different sepiolite membranes
(0.5A, 0.25A, 0.5B, and 0.25B) calicined at 550ºC.
-
membranes have quite similar trends, and pore sizescover a broad
range. That is why the separation isquite low, even for PEO, with
400,000 MW. Asshown in Figure 7, the pores of sepiolite mem-branes
are neither spherical nor cylindrical asviewed with SEM; the
sepiolite membrane also was
formed by layers of fibers. This may be the reasonfor the broad
distribution of pore sizes.
Pore density and surface porosity
Pore density and surface porosity were calculatedaccording to
Equations (7) and (8); results are listed
394 WANG ET AL.
Figure 3. continued
-
395SEPIOLITE MEMBRANE FOR ULTRAFILTRATION
Figure 4. Solute separation curves (solute diameter vs. their
separation) plotted on a lognormal probabilitypaper for 0.5A,
0.25A, 0.5B, and 0.25B membranes.
Figure 5. Cumulative pore size distribution of four different
sepiolite membranes (0.5A, 0.25A, 0.5B, and 0.25B).
Table 2. Geometric mean pore size (µp) and geometric standard
deviation (σp) of sepiolite membranes.
Membrane µp (nm) σp
0.5A 25.7 1.950.25A 25.1 1.910.5B 23.4 2.040.25B 23.0 1.99
90
70
50
30
10
Solu
te s
epar
atio
n (
%)
10 100Solute diameter (nm)
0.5A (r2=0.92)0.25A (r2=0.93)0.5B (r2=0.96)0.25B (r2=0.96)
-
in Table 3. Pore density and surface porosity did notdepend very
much on the amount of sepiolite used.Instead, they depended highly
on the method of claydispersion. The aggregation of the fiber
bundles wasprevented by sonifying the clay suspension. As a
396 WANG ET AL.
Figure 5. continued
consequence, the pore density and surface porosityof the
membrane increased, resulting in higher flux,despite the fact that
the membrane thicknessincreased as a result of sonification. The
thickness ofthe membrane (43–114 µm) is 107.5–570 times as
-
397SEPIOLITE MEMBRANE FOR ULTRAFILTRATION
Figure 5. continued
Figure 6. Probability density function curves for four different
sepiolite membranes (0.5A, 0.25A, 0.5B, and 0.25B).
large as the diameter of the sepiolite fiber (0.2–0.4µm); this
difference indicates that the membraneswere formed by assembling
layers of completely dis-ordered fibers. Therefore, a large portion
of the pores
formed are dead ended or blocked between layers offibers. The
pore density and surface porosity calcu-lated above are only for
the pores that could passthrough the entire cross section of
membrane.
-
Characterization by SEM images
Figure 7 shows SEM images of membranes con-structed of layers of
disordered sepiolite. It is easyto prepare pinhole- and crack-free
membranesfrom sepiolite because of its fibrous nature. The
SEM image also shows that fibers are better dis-persed in the
membrane prepared from ultrasoni-fied suspension, unlike the
membrane made byapplying magnetic stirring, where a network offiber
bundles is observed.
398 WANG ET AL.
Figure 6. continued
-
Arribas, J. I., F. Martinez, A. Hernández, P. Prádanos, andG.
Caruana. 1991. Morphological study of surfaceinorganic membranes by
scanning electron micro-scopy and image analysis. Key Engineering
Materials61/62:371–374.
Cool, R., A. Clearfield, V. Mariagnanam, L. J. McElli-strem, R.
M. Crooks, and E. F. Vansant. 1997. Self-assembly of
aluminum-pillared clay on a gold sup-port. Journal of Materials
Chemistry 7(3):443–448.
Cot, L. 1998. Inorganic membranes: academic exercise
orindustrial reality. In Proceedings of the Fifth Inter-nal
Conference on Inorganic Membranes, June22–26, Nagoya, Japan.
Hsieh, F.-U., T. Matsuura, and S. Sourirajan. 1979.
Reverseosmosis separation of polyethylene glycols in diluteaqueous
solutions using porous cellulose acetate mem-brane. Journal of
Applied Polymer Science 23:561–573.
Ishiguro, M., T. Matsuura, and C. Detellier. 1995.
Reverseosmosis separation for a montmorillonite mem-brane. Journal
of Membrane Science 107:87–92.
Jones, B. F., and E. Galán. 1988. Sepiolite and
palygorskite.Pages 631–674 in S. W. Bailey, editor. Reviews
inmineralogy. Volume 19: Hydrous phyllosilicates.Mineralogical
Society of America, Washington, DC.
Le Van Mao, R., E. Rutinduka, C. Detellier, P. Gougay,
V.Hascoet, S. Tavakoliyan, S. V. Hoa, and T. Matsuu-ra. 1999.
Mechanical and pore characteristics of zeo-lite composite membrane.
Journal of MaterialsChemistry 9:783–788.
Michaels, A. S. 1980. Analysis and prediction of sieving
curvesfor ultrafiltration membranes: a universal
correlation?Separation Science Technology 15 (6):1305–1322.
399SEPIOLITE MEMBRANE FOR ULTRAFILTRATION
Figure 6. continued
Conclusions• Ultrafiltration membranes can be prepared
from sepiolite clay material.• The method of preparing sepiolite
mem-
branes requires only one step (spreadingthe clay suspension on a
smooth surface)before calcination, whereas the sol–gelmethod
requires three steps (precipitation,peptization, and gelling).
Thus, sepiolitemembrane preparation is much simpler.
• The sepiolite membrane has a broad poresize distribution
because of its multilayerstructure of sepiolite fibers.
• Sepiolite fiber is better dispersed by sonifi-cation than by
magnetic stirring.
AcknowledgmentsThe authors thank the China Scholarship
Councilfor providing a scholarship. They also thank theNatural
Sciences and Engineering Research Coun-cil of Canada (NSERC) for a
research grant (toCD) and a postdoctoral fellowship.
ReferencesAimar, P., M. Meireles, and V. Sanchez. 1994. A
contri-
bution to the translation of retention curves intopore size
distributions for sieving membranes. Jour-nal of Membrane Science
54:321–338.
-
400 WANG ET AL.
Figure 7. Scanning micrographs of sepiolite membranes: cross
sections of 0.5A (a), 0.5B (b), 0.25A (e), and0.25B (f); surfaces
of 0.5A (c), 0.5B (d), 0.25A (g), and 0.25B (h).
-
Mishra, T., and K. Parida. 1997. Transition-metal oxidepillared
clays. 2. A comparative study of textural, andacidic properties of
manganese(III) pillared montmo-rillonite, and pillared
acid-activated montmoril-lonite. Journal of Materials Chemistry
7(1):147–152.
Nagata, H., S. Shimoda, and T. Sudo. 1974. On dehydra-tion of
bound water of sepiolite. Clay and Clay Min-erals 22:285–293.
Pérez-Rodríguez, J. L., and E. Galán. 1994. Determinationof
impurity in sepiolite by thermal analysis. Journalof Thermal
Analysis 42:131–141.
Serna, C., J. L. Ahlrichs, and J. M. Serratosa. 1975. Fold-ing
in sepiolite crystals. Clays and Clay Minerals23:452–457.
Singh, S., K. C. Khulbe, T. Matsuura, and P. Ramamurthy.1998.
Membrane characterisation by solute trans-port and atomic force
microscopy. Journal of Mem-brane Science 142:111–127.
Youm, K. H., and W. S. Kim. 1991. Prediction of intrinsicpore
properties of ultrafiltration membrane by soluterejection curves:
effects of operating conditions onpore properties. Journal of
Chemical EngineeringJapan 24:1–7.
Zeman, L., and M. Wales. 1981. Steric rejection of poly-meric
solutes by membranes with uniform pore sizedistribution. Separation
Science and Technology 16(3):275–290.
401SEPIOLITE MEMBRANE FOR ULTRAFILTRATION
Table 3. Pore density and surface porosity of sepiolite
membranes calculated from solute transport data.
Membrane Pore density (pores/µm2) Surface porosity (%)
0.5A 27 3.150.25A 28 3.040.5B 67 7.150.25B 70 7.03