ultrafiltration membrane synthesis by nanoscale templating

7/27/2019 Ultrafiltration Membrane Synthesis by Nanoscale Templating

1/14

Journal of Membrane Science 198 (2002) 173186

Ultrafiltration membrane synthesis by nanoscale templating ofporous carbon

Michael S. Strano a,1, Andrew L. Zydney a,1, Howard Barth b,Gilber Wooler b, Hans Agarwal a,1, Henry C. Foley c,

a Colburn Laboratory, Department of Chemical Engineering, University of Delaware, Academy Street, Newark, DE 19716, USAb DuPont Company, Central Research and Development, Experimental Station, P.O. Box 80228, Wilmington, DE 19880-0228, USA

c Department of Chemical Engineering, Center for Catalytic Science and Technology,

Pennsylvania State University, University Park, PA 16802, USA

Received 22 March 2001; received in revised form 28 June 2001; accepted 9 July 2001

Abstract

A novel method for producing carbon membranes for ultrafiltration applications is presented using a spray deposition and

pyrolysis of poly(furfuryl alcohol)/poly(ethylene glycol) mixtures on macroporous stainless steel supports. The poly(ethylene

glycol) or PEG employed as a carbonization template creates a mesoporosity that leads to pores in the ultrafiltration range.

Scanning electron microscopy (SEM) shows that the membranes consisted of 12- to 15-m thick carbon films. Gas permeation

and water permeability data were used for the calculation of mean pore sizes, which were found to decrease with decreasingaverage molecular weight of the PEG template. Ultrafiltration of a polydisperse dextran solution was used to quantify the

retention properties of the membranes. Molecular weight cutoffs determined from dextran retention data were shown to vary

with template molecular weight: values of 2 104, 3.5 104, and 6 104 gmol1 dextran were measured for respective

templates of 2000, 3400, and 8000g mol1 PEG. For PEG molecular weights of 2000 or below, the templating effect was

ill defined, membrane film cracking became more prominent, and membrane selectivity and reproducibility were adversely

affected. 2002 Published by Elsevier Science B.V.

Keywords: Ultrafiltration; Carbon membranes; Membrane formation; Inorganic membranes

1. Introduction

Membrane filtration technologies are critical to a

variety of industrial process applications including

cell harvesting, sterile filtration, protein enrichment,

and removal of particulate matter [1]. Ultrafiltration

Corresponding author. Tel.: +1-814-865-2574;

fax: +1-814-865-7846.

E-mail addresses: [email protected] (M.S. Strano),

[email protected] (H.C. Foley).1 Tel.: +1-302-831-2345; fax: +1-302-831-2085.

in particular, where the membrane retains compo-

nents with kinetic diameters from 1 to 100 nm, haswidespread utility for industrial separation processes.

Traditionally, membranes used for ultrafiltration have

been polymeric in nature. Asymmetric ultrafiltration

membranes are commonly synthesized using phase

inversion, where a polymer solution consisting of

a base and pore-former in a solvent is induced to

form two interdispersed liquid phases. Membranes

synthesized in this manner include the bilayer type

containing slit shaped fissures or cracks [2] and those

that contain plasticizers and are stable while dry [3,4].

0376-7388/02/$ see front matter 2002 Published by Elsevier Science B.V.

P I I : S 0 3 7 6 - 7 3 8 8 ( 0 1 ) 0 0 5 7 4 - 9


2/14

174 M.S. Strano et al. / Journal of Membrane Science 198 (2002) 173186

Nomenclature

B0 porous structure factor

(mol ideal gas m

1

)Cb concentration of dextran in

bulk solution (g l1)Cp concentration of dextran in

permeate (g l1)

Cw concentration of dextran at

membrane surface (g l1)dm average membrane pore size from

gas permeation (m)

J flux of gas (mol m2 s1)

Jv filtrate flux during ultrafiltration (m s1)

k solute mass transfer coefficient (m s1)K0 Knudsen permeability

(mol m1 s1 Pa1)

l carbon layer thickness (m)

M molecular weight (g mol1)p average trans-membrane pressure

difference (Pa)

p pressure driving force under gas

permeation (Pa)

rpore cylindrical pore radius (m)

R ideal gas constant (Pa m3 mol1)

Rc rejection coefficient for dextran

ultrafiltrationT temperature of gas permeation (K)

Greek symbols

geometric constant for porous materials

geometric constant for porous materials

proportionality constant in mass

transfer correlation

stagnant film thickness above

membrane in stirred-cell (m) membrane porosity

gas viscosity (Pa s) viscosity of solvent (Pa s)

pore tortuosity

Despite the widespread use of these types of

membrane materials, they suffer from several disad-

vantages. The membranes have limited mechanical

integrity which can lead to deformation during op-

eration and adversely affect membrane performance.

These materials often have limited temperature and

chemical stability, prohibiting their use in applica-

tions involving harsh organic solvents or at elevated

temperature [4]. The use of caustic and hypochlorite

solutions as cleaning agents can also lead to chemicaldegradation of the polymeric materials, significantly

reducing the life of these membranes. In addition,

membranes made from hydrophobic polymers like

polysulfone and polyethersulfone must contain a

humectant such as glycerol to avoid problems asso-

ciated with re-wetting, while membranes made from

hydrophilic polymers like cellulose must remain wet

at all times to avoid collapse of the pore structure.

While inorganic ultrafiltration membranes have

attracted increasing attention in recent years [57],

pore size modified carbon membranes have been

relatively unexplored. Yet compared to ceramic and

metallic ultrafiltration membrane materials, carbon

can be considerably less expensive. Nanoporous car-

bon is also a promising material because it is chem-

ically inert under typical processing conditions and

thermally stable at temperatures well above 200 C

[8,9]. Formed from the pyrolysis of carbonizing nat-

ural or synthetic polymeric precursors, this material

is a disordered solid having a pore size approach-

ing molecular dimensions and has been shown to

possess highly shape-selective molecular transport

properties [10]. Poly(furfuryl alcohol) (PFA)-derivednanoporous carbons have a mean pore size about

0.5 nm as measured from N2 and methyl chloride ad-

sorption isotherms [11]. Attempts to use this material

in the synthesis of defect free, micron scale films on

structurally stable macroporous supports have been

successful. Nanoporous carbon membranes have been

fabricated with high selectivity for some gas phase

separations of small molecules [1214].

While such membranes have a mean pore size on

the molecular scale (


3/14

M.S. Strano et al. / Journal of Membrane Science 198 (2002) 173186 175

and 0.28 cm3 g1 for 2000, 3400, and 8000 g mol1

templates, respectively. These results have been used

in the synthesis of various nanoporous carbon cat-

alysts with greatly enhanced effectiveness factorswhen compared to catalysts prepared from native

nanoporous carbon alone [16,17].

Studies involving the use of porous carbon for ul-

trafiltration membrane formation in the literature are

scarce. Schindler and Maier [4] claimed to have syn-

thesized such a membrane through the carbonization

of a pre-existing polymer membrane with identical

pore structure. Although pyrolysis of a polymeric

microfiltration membrane can reasonably produce a

carbon membrane with similar macroporosity, there is

little theoretical or empirical evidence to suggest that

pores in the entire ultrafiltration range (1100 nm)

would be preserved during the carbonization of a

pre-existing film. Additionally, ultrafiltration range

pores were undetectable using the bubble-point test-

ing methodology employed by the authors [4].

In this paper, a novel method for producing carbon

ultrafiltration membranes is presented. Spray deposi-

tion and pyrolysis on macroporous stainless steel of

poly(furfuryl alcohol)/poly(ethylene glycol) mixtures

is the straightforward fabrication technique employed.

The methodology is demonstrated in Fig. 1. As in our

earlier work, the PEG is used as a non-carbonizingtemplate molecule to introduce controllable meso-

porosity into the carbon film. Permeation of a poly-

disperse dextran solution was used to quantify the

Fig. 1. Mesoporous carbon membranes via nanoscale templating.

retention properties of the membranes as a function

of template size. A correlation of partial rejection

coefficients for 90% of the dextran mixture with tem-

plate molecular weight was observed. Membraneswere also characterized using scanning electron

microscopy (SEM) and gas permeation.

2. Experimental

2.1. Membrane synthesis

Carbon based ultrafiltration membranes were syn-

thesized using an adaptation of a spray coating proce-

dure used for creating supported nanoporous carbon

membranes [13]. Furfuryl alcohol resin in the form of

Durez Resin #16,470 (PFA) from Occidental Chem-

ical Corporation diluted in reagent grade acetone

(Aldrich Chemical) was combined with poly(ethylene

glycol) (PEG) and spray deposited upon a macrop-

orous stainless steel support. The PFA resin has a

specific gravity of 1.21 and a viscosity of 200 cP

at 25 C that decreases to about 5 cP at 80 C. PEG

samples of average molecular weights of 1000, 1500,

2000, 3400, and 8000 g mol1 were obtained from

Aldrich Chemical Company.

This PFA/PEG/acetone precursor was prepared bycombining PFA resin with a given average molecular

weight of PEG in a 50/50 wt.% ratio at 70 C and sub-

sequent dilution in acetone forming a stable mixture


4/14


Table 1

Summary of synthesis and performance characteristics for membranes used in this worka

Initial weight

support (g)

Wet weight

support (g)

Net carbon

weight (g)

Pressure

applied (psig)

Water flux

(m s

1

)

Water permeance

(m s

1

Pa

1

)M8000-1-0.021 10.5107 10.6393 0.021 50.5 1.8 106 5.1 1012

M8000-2-0.018 10.8591 10.9768 0.018 54 1.1 105 2.9 1011

M8000-3-0.022 10.4478 10.58 0.022 54 1.8 106 4.7 1012

M1500-1-0.019 10.4659 10.5728 0.019 55 5.9 107 1.5 1012

M3400-1-0.019 10.7327 10.8409 0.0188 75 8.3 107 1.6 1012

M3400-2-0.016 10.286 10.3696 0.0161 75 8.3 107 1.6 1012

M3400-3-0.018 10.5444 10.6481 0.018 75 8.3 107 1.6 1012

M2000-1-0.019 10.6382 10.7455 0.0189 50 2.0 108 5.6 1014

M2000-2-0.019 10.4353 10.534 0.019 50 1.3 107 3.8 1013

M2000-3-0.019 10.3555 10.4623 0.019 50 8.8 108 2.5 1013

M2000-4-0.019 10.6816 10.7785 0.0189 50 1.0 107 3.0 1013

M1000-1-0.029 10.5297 10.6695 0.0288 50 1.7 107 5.0 1013

M1000-2-0.023 10.7057 10.8189 0.0227 50 5.7 108 1.6 1013

a Labels are as follows: M(PEG MW)-(sample #)-(carbon mass (in g)).

at room temperature with a viscosity near 5 cP. The

resulting solution was subsequently spray deposited

onto circular porous stainless steel supports (0.2m

pore size, 11.4 cm2 area, and 1 mm thickness) sup-

plied by Mott Metallurgical Co. Before spray coating

the stainless steel supports were cleaned by sonica-

tion in acetone. After spray coating approximately

200 mg of the polymer blend solution (wet weight)was deposited on the support. The resulting system

was subsequently pyrolyzed in a stream of flowing

He (50sccm) at 5 Cmin1 to 600 C, held for 2h

at this temperature, and then allowed to cool to room

temperature. The final carbon mass on each support

was approximately 20 mg. Table 1 presents data on

the membranes used in this work. Membrane labels

contain the PEG template molecular weight and total

coat mass of carbon for each sample.

2.2. Scanning electron microscopy

SEM was used to characterize the internal mor-

phology of the membranes. Membrane cross-sections

were cut orthogonal to the support surface using a

diamond-wafering saw. These sections were mounted

in an epoxy resin, polished, and given a coating of

Au for imaging with a Hitachi S-4000 field emis-

sion scanning electron microscope. Imaging was

performed on areas of these cross-sections external

to and within the macroporosity of the stainless steel

support and at various radii from the center of the

disk-shaped membranes.

2.3. Characterization using gas permeation

Gas phase transport of He, Ar, N2, O2, SF6 and CO2was used to characterize the selective porosity and

integrity of the carbon membrane. The disk-shapedmembranes were sealed using VitonTM gaskets into a

stainless steel module setup to measure the transport

of a single gas through the membrane. The pure com-

ponent flux of each gas was measured as a function

of pressure. Details concerning the experimental setup

and technique can be found in [18].

2.4. Ultrafiltration of polydisperse dextrans

Ultrafiltration of polydisperse dextrans was used to

characterize the retention properties of the templatedcarbon films. Dextrans having average molecular

weights of 2 106, 1.7 105, 7.0 104, 3.9 104,

9.9 103 gmol1 (Pharmacia) were obtained from

Aldrich Chemical Company. The dextrans were added

in equal mass ratios to create a 10 g l1 solution in a

phosphate buffer prepared from monobasic (NaHPO4)

and dibasic (NaH2PO4) sodium phosphate to achieve

a solution pH of 7.40.3. When applicable, 1 wt.% of

methanol was added to prevent bacterial contamina-

tion.


5/14


Retentates and permeates were analyzed using size

exclusion chromatography (SEC) with a model 1050

HP chromatograph (Hewlett Packard) and a Waters

410 differential refractometer. SEC analysis wasperformed using a 7.5-mm i.d. 30-cm TSK G4000

SW column (Tosoh Corp.), thermostated at 30 C,

with 0.1 M sodium sulfate used as the mobile phase

at a flow rate of 0.6 ml min1. Samples were injected

with an autosampler without additional dilution using

a 100-l sample loop. The above mentioned dex-

trans were injected individually at a concentration of

1mgml1, and retention volume data were used to

construct a molecular weight calibration curve. Data

was processed using Waters Millennium 32 HPLC

software.

Membrane separation was carried out using a

dead end filtration setup. A 50 ml volume of dextran

solution was loaded above the carbon membrane in

a 30 mm diameter Amicon stirred-cell. An attached

Ar line supplied gas pressure to the solution above

the membrane. The permeate was collected in 30 ml

vials, with the total permeate mass weighed contin-

uously by placing the vial directly on a laboratory

balance. From this time series data mass flow through

the membrane was computed. A series of permeate

and retentate samples were collected over time for

off-line dextran analysis.

2.5. Batch depletion dextran adsorption

Granular carbon membrane samples were prepared

using the identical procedure described above for

making thin carbon films, except that the solution

was poured into a quartz boat and pyrolyzed as a

bulk solution. These samples were analyzed for total

dextran adsorption capacity using a batch depletion

method. Each carbon was ground to 60/140 mesh

and suspended in a water solution. After settling, theremaining solution was decanted to remove fine parti-

cles. The process was repeated until all fine particles

were removed. A mass of 0.5 g of carbon was sealed

with 30 ml of the polydisperse dextran solution in

a Pyrex vial as described above. Sealed vials were

vigorously agitated for 32 h, after which the solutions

were passed through a 1m filter and analyzed using

SEC as described above. The average concentration

of dextran remaining in the solution above the carbon

was integrated over the molecular weight distribu-

tion, with the results compared to the corresponding

integral for the original solution.

3. Results and discussion

Table 1 lists the initial support weight, the weight

after spray coating (wet), and the weight after car-

bonization (dry) for each membrane studied in this

work. All samples except M8000-3-0.022 were given

a single coat and thermal treatment. This particular

sample was fabricated in two coating/carbonization

cycles. During the carbonization process, the evolu-

tion of gases reduced the polymer mass on the support

leaving behind the solid carbon film. For poly(furfuryl

alcohol) carbonization at 600 C, the carbon yield pro-duced in this way is approximately 2530% for sim-

ilar thin films [14]. The PEG template is expected to

produce very little residual carbon as seen in the syn-

thesis of mesoporous catalyst supports (


6/14


Fig. 3. Water permeance through carbon membranes as a functionof template molecular weight.

the entire template being emitted as gaseous prod-

ucts. This suggests that volatilization of the template

does not cause any significant loss of poly(furfuryl

alcohol) prior to carbonization.

It is noteworthy that a similar trend is observed for

the water permeability (Fig. 3) plotted as a function of

Fig. 4. Scanning electron micrograph of a porous carbon ultrafiltration membrane (M8000-1-0.021) cross-section showing external carbon

layer and stainless steel support.

template size although there is considerable variabil-

ity between samples formed from the same template

size. Generally, water permeability decreases with

decreasing template size and this may correspond toa decrease in membrane pore size and porosity as dis-

cussed below. In Table 1, the values for the hydraulic

permeabilities range from 1014 to 1011 m s1 Pa1,

which are considerably lower than commercial ul-

trafiltration membranes (typical values are around

1010 m s1 Pa1. These reduced permeabilities are

attributable to both the much greater thickness of the

membranes and the decrease in the film porosity that

occurs with decreasing template size. These factors

are discussed in detail below.

3.1. Scanning electron microscopy

Fig. 4 is a scanning electron micrograph of a

cross-section of the external porous carbon layer and

the stainless steel support for a membrane synthesized

using a 50% mixture of PFA: 8000 g mol1 PEG at

600 C. The demarcation between the stainless steel

support and external carbon layer is clearly visible,


7/14


and there are also localized along with localized sep-

arations or cracks between the two layers. These film

separations apparently have a negligible impact on

coating integrity and mechanical stability, as therewas no observed carbon mass loss during testing or

property variations after repeated testing. The com-

bination of the diamond saw cutting and surface

polishing of the SEM sample has a smoothing affect

on the morphology of the membrane cross-sections.

The actual porosity of the steel support is roughly

0.6 as reported by the manufacturer. The carbon layer

is 12- to 15-m thick and relatively uniform over

the membrane surface, consistent with observations

of similarly prepared carbon films for gas separation

[14]. The calculated thickness range based on the

mass of carbon deposited (Table 1) and assuming

a porous carbon density of 1.6 g cm3 is 915m,

which agrees well with this measurement.

Fig. 5 is a 500 nm scale micrograph of the mi-

crostructure of the PFA/PEG carbon taken from

the material external to the support surface. This

cross-section was prepared by fracturing the carbon

surface while attached to the support using a scalpel.

Fig. 5. Scanning electron micrograph of the pore structure of the carbon ultrafiltration membrane showing grain boundaries between

nanoporous carbon domains (M8000-1-0.021).

The microstructure suggests pores with diameters in

the nanometer range. As expected from the inferred

hypothesis for pore formation, these pores are in the

form of fissures and voids between purely nanoporouscarbon domains. Porous carbon was also observed to

be present on the opposite side of the support (the

side opposite from the initial deposition) and may

fill some portion of the macroporosity of the sup-

port. During pyrolysis the polymer blend drops in

viscosity prior to carbonization, making deeper pore

penetration more facile.

3.2. Porosity characterization by gas permeation

Fig. 6(a) is a plot of the flux versus pressure for

several gases through an uncoated stainless steel

0.2m support. The data show both a high rate of

permeation and low gas selectivity. A nanoporous

carbon film pyrolized in the absence of polymer tem-

plate shows a linear relationship between the flux and

applied pressure for He, O2, N2 and Ar (Fig. 6(b)).

This permeation behavior is typical of PFA derived

carbon membranes [18]. Fig. 7 is characteristic of gas


8/14


Fig. 6. Flux as a function of driving force pressure for: (a) uncoated stainless steel support; (b) 0% PEG/PFA carbon membrane carbonized

at 600 C; (c) 50% 8000 amu PEG/PFA templated membrane carbonized at 600 C.

permeation through a templated carbon membrane

(50% PFA/8000 g mol

1 PEG) showing a quadraticdependence of the flux versus pressure and Knudsen

gas selectivity at low pressure. This behavior is

Fig. 7. Ultrafiltration membrane (M3400-1-0.019) gas permeability

as a function of trans-membrane pressure for mean pore size

characterization.

observed on all PEG/PFA carbon membranes regard-

less of coating deposition mass or template size emplo-yed. Hence, the permeation results cannot necessarily

be attributed to partially coated or partially formed

nanoporous films such as that shown in Fig. 6(b).

For a membrane with a mean pore size between

100 and 1 nm, the pore dimension can be estimated

by examining data for the membrane gas perme-

ability versus trans-membrane pressure. The linear

relationship yields an intercept equal to the Knud-

sen permeability with slope inversely proportional to

the gas viscosity (Fig. 7). The gas flux through the

membrane is written as

J=

K0 +B0

p

p

l(1)

The structure factor B0 is proportional to the square

of the mean pore size

B0 =

2d2m (2)

with = 2.5 for consolidated media [20]. The

Knudsen permeability K0 can be written as


9/14


Table 2

Mean pore sizes of membranes measured by gas permeation

dm (m)

PEG1500 PEG3400 PEG8000

He 9.92 109 1.19 108

N2 1.03 108 1.28 108 2.83 108

Ar 1.37 108 6.81 109 2.59 108

O2 9.15 109 2.39 108 2.21 108

CO2 1.43 108 1.22 108 2.87 108

SF6 1.13 108 1.10 108

Mean 1.15 108 1.31 108 2.63 108

S.D. 2.10 109 5.69 109 3.05 109

K0 =4

32 dm8RTM (3)

where is 0.8 for consolidated media [20]. The gas

permeation data can be used to evaluate both B0 and

K0, with the ratio of these two parameters used to

evaluate the mean pore size of the membrane, dmwithout any assumptions regarding the membrane

porosity, or tortuosity factor, . Here, we do assume

that the pore size of the carbon film is significantly

less than that of the support (i.e.


10/14


Fig. 8. Observed (dotted line) and actual (solid line) partial rejection coefficients vs. dextran molecular weight for carbon membranes

synthesized from poly(ethylene glycol) template: (a) 1000 MW PEG; (b) 1500 MW PEG; (c) 2000 MW PEG; (d) 3400 MW PEG; (e)

8000 MW PEG.


11/14


as a function of dextran molecular weight for mem-

branes synthesized using the five different template

PEG sizes. The results for 1000 MW PEG shown

in Fig. 8(a) demonstrate a lack of dextran retentionwith respect to molecular weight indicating that large,

non-selective pores exist across the film. The polar-

ization correction reveals that the broad, non-selective

behavior of the membrane cannot be attributed to po-

larization, but rather that the membrane is defective.

As the template size tends toward zero, the carbon

membrane behaves as an untemplated film with a sig-

nificant tendency towards cracking. It has been well

established that furfuryl alcohol resin-based mem-

branes require several coatings before a crack-free

layer can be constructed [12,14]. The curves for

1500 (Fig. 8(b)) and 2000 PEG templates (Fig. 8(c))

show an increase in membrane selectivity. However,

a tendency towards film cracking is retained and

membrane reproducibility is affected. Fig. 8(c) in

particular shows a lack of reproducibility for four

membranes synthesized at the same carbon deposition

mass. Membrane M2000-4-0.019 (Fig. 8(c)) shows a

significantly enhanced retention compared to those in

Fig. 8(a), which may represent the properties of the

film in the absence of any significant cracking. Such

retention could not be attained using a 1000 MW pre-

cursor. The templating of nanoporous carbon usingPEG has been observed to yield accessible meso-

porosity only at molecular weights of 2000 g mol1

and higher for granular materials [15], although

this study represents the most comprehensive work

undertaken to determine the exact threshold.

Problems of film cracking and reproducibility

diminish for higher molecular weight templates as

transport through these films becomes dominated by

that which takes place through the porosity created via

the templates rather than through adventitious cracks

and fissures. Fig. 8(d) and (e) show reproduciblebehavior for 3400 and 8000 MW PEG templated

membranes, respectively. Retention properties tend to

increase from those seen in Fig. 8(b) and (c) and are

apparently independent of carbon thickness and water

permeance as expected. Membrane M8000-3-0.022 is

the result of two successive coating and carbonization

cycles and shows an increase in retention behavior

likely from a reduction in pore size after the initial

coat. Here, the pore structure of the first coat may be-

come impregnated with a second coating of precursor

Fig. 9. Dextran R90 value and calculated pore size as a function

of template molecular weight of carbon membrane.

material that modifies the retention characteristics

as shown. This procedure, which is analogous to a

slip casting design, may in fact provide a route to

the synthesis of ultra-thin nanofiltration membranes.

With successive layer deposition, the retentive prop-

erties decrease with only a marginal increase in film

thickness and deposition mass.

In Fig. 9, the molecular weight of dextran corre-sponding to 90% rejection is plotted as a function of

the molecular weight of the PEG template. This R90value is obtained by interpolation or extrapolation of

the data in Fig. 8(a)(e) as needed. The extrapolated

value for the 1000 g mol1 template determined from

Fig. 8(a) is actually an underestimate of this molecular

weight. As observed above, the analysis suggests that

these films possess large cracks and defects that are

non-retentive. The trend shows an increase in dextranR90 with increasing PEG molecular weight starting

from the 2000 MW PEG template size. Below thiscutoff, the selectivity decreases sharply with decreas-

ing template size suggesting the formation of cracks

and defects. Fig. 9 also plots the calculated pore size

based upon a simple partitioning model assuming a

cylindrical pore structure and using the Stokes radius

of the dextran molecule for the corresponding R90molecular weight [1]

Rc = 1

1rdextran

rpore

2(6)


12/14


Pore sizes estimated in this way for the membranes

formed from the larger PEG templates are generally

smaller than those calculated from gas permeability

data. For example, the values of 15 and 12.4 nm cor-responding to 8000 and 3400 molecular weight PEG

compare to 26.3 6.1 and 13.1 11.4 nm calculated

from gas permeation, respectively. In contrast, the

pore size calculated from the dextran data for the

1500 PEG membrane was 28.2 nm, which is signifi-

cantly greater than the 11.54.2 nm determined from

gas permeation. This large discrepancy is due to the

greater effect of the pore defects on dextran transport

than on the gas permeation results.

The water permeability can also be used to provide

an estimate of the average pore size or porosity using

the HagenPoiseuille equation. Here, the membrane

porosity/tortuosity ratio, /, can be expressed as a

function of the pore size and water permeability Lpassuming a cylindrical pore structure [1]

=

8lLp

(rpore)2(7)

Using pore sizes from dextran rejection data (Eq. (6)),

application of Eq. (7) predicts that the ratio of porosity

to tortuosity decreases with decreasing template size.

Average values are 0.024, 0.005, 0.001 and 0.001 for

8000, 3400, 2000, and 1000 g mol1 template. Thisdecrease in / is consistent with the greater carbon

yield, and thus, lower porosity, of the membranes

formed using the smaller PEG templates.

Results of the batch depletion of dextran over gran-

ular templated carbons yield relative values of the

accessible pore volumes of the various carbons. This

allows for a comparison between properties resulting

from the templating process and those resulting from

pores developed as a result of film cracking. Fig. 10

is a plot of relative dextran amount adsorbed versus

PEG molecular weight used during carbon synthesis.The results show negligible uptake at values of 2000

MW PEG and below with substantial increases in

accessible area at larger template molecular weights

(sizes). The data suggest a maximum in adsorption

uptake near 8000 MW PEG. At 18,500 MW PEG, the

extent of adsorption falls to values below that of the

3400 MW PEG. The initial portion of the curve agrees

well with previous gas adsorption characterization,

where it has also been noted that accessibility of the

mesopore distribution becomes hampered for lower

Fig. 10. Dextran uptake from batch depletion on granular carbons

synthesized using templates of 30018,500 g mol1 of PEG.

template molecular weights [15]. It is this lack of ac-

cessibility that may be responsible for the trapping

of template material within the carbonizing system

and hence the increase in carbon yield at low template

molecular weights (Fig. 2). This retention of material

would explain the observed decrease in membrane

porosity. In Fig. 10, the subsequent decrease in capac-

ity with 18,500 PEG may be the effect of a decrease inthe rate of pyrolysis of this larger polymeric template.

The subsequent decrease in the rate of evolution of the

pyrolysis gases could also lead to a retention of tem-

plate carbon within the forming solid, as in the case

of the diffusion limitations associated with smaller

template sizes.

The internal pore volume of membranes made from

the smaller templates is largely inaccessible, leading

to a small amount of dextran uptake. As the template

size increases, the accessibility of the pore volume

increases, but at the expense of an increasing aver-age pore radius, which ultimately reduces the internal

surface area.

The emerging picture of the actual templating

process from the above observations is one that is

kinetically driven by the diffusion of template decom-

position products from the forming solid. At small

template pore sizes (


13/14


resistance is exceedingly high. These diffusion lim-

itations increase carbon yield and decrease carbon

porosity as observed. The lack of mesopore intercon-

nection also means that the membrane behaves as apurely nanoporous thin film, and hence is subject to

cracking and the formation of defects after a single

coating. As the template size increases, this diffusion

length scale increases and membrane properties be-

come a function of the templated carbon itself with a

subsequent improvement in retention properties and

reproducibility.

The results presented here underscore several

potential improvements that can be made through

modification of the synthesis procedure. The most

notable of these is improvement in water perme-

ability by decreasing carbon film thickness. The

spray deposition on macroporous support was readily

adapted from an analogous method of gas-separation

membrane fabrication [15], although this may not be

the optimal system for liquid separation applications.

The results of microscopy and water permeation mea-

surements suggest that decreasing the carbon deposi-

tion mass while maintaining coating uniformity can

accomplish this goal. Additionally, this work presents

indirect evidence that lower molecular weight tem-

plates have a discernable affect on pore structure and

may not be entirely flashed off during carbonizationas originally proposed [17]. This warrants a more

complete investigation of the templating process with

a particular emphasis on the composition and molec-

ular weight dependence of the template as well as the

pyrolysis ramp rate during carbonization.

4. Conclusions

A novel method for producing ultra and nanofil-

tration carbon membranes is presented using spraydeposition on macroporous stainless steel and pyrol-

ysis of poly(furfuryl alcohol)/poly(ethylene glycol)

mixtures. Membranes characterized by SEM show

12- to 15-m thick carbon films. Pores in the ultrafil-

tration range clearly vary with the average molecular

weight of the template for PEG with molecular weight

of 2000 g mol1 and above. Below this threshold, the

film behaves similar to an untemplated carbon mem-

brane with crack formation a significant factor for the

single coated membranes. Transport of a polydisperse

dextran solution was used to quantify the retention

properties of the membranes. Partial rejection coef-

ficients for 90% of the dextran mixture were shown

to vary with template molecular weight: 90% cutoffsof 2 104, 3.5 104, 6 104 gmol1 of dextran

were measured for 2000, 3400, and 8000 g mol1 of

template employed, respectively. Although generally

of less accuracy, the mean pore sizes as measured

from gas permeation data also demonstrate a similar

decrease in value with decreasing template size.

Acknowledgements

Support for this research was provided by the

Department of Energy Office of Basic Energy Science

and the DuPont Co. Michael Strano is grateful for fi-

nancial support in the form of a Presidential Graduate

Fellowship from the University of Delaware.

References

[1] L.J. Zeman, A.L. Zydney, Microfiltration and Ultrafiltration:

Principles and Applications, Marcel Dekker, New York, 1996.

[2] A.S. Michaels, High flow membrane, US Patent #3615024

(1971).

[3] T.A. Tweddle, W.L. Thayer, O. Kutowy, S. Sourirajan, Methodof gelling cast, polysulfone memrbrane, US Patent #4451424

(1984).

[4] E. Schindler, F. Maier, Manufacture of porous carbon

membranes, US Patent #4919860 (1990).

[5] W.M. Clark, A. Bansal, M. Sontakke, Y.H. Ma, Protein

adsorption and fouling in ceramic ultrafiltration membranes,

J. Membr. Sci. 55 (1991) 2138.

[6] H.P. Hsieh, R.R. Bhave, H.L. Fleming, Microporous alumina

membranes, J. Membr. Sci. 39 (1988) 221241.

[7] J.C.S. Wu, L.C. Cheng, An improved synthesis of ultrafil-

tration zirconia membranes via the solgel route using

alkoxide precursor, J. Membr. Sci. 167 (2) (2000) 253261.

[8] H.C. Foley, Carbogenic molecular-sieves synthesis, proper-ties and applications, Micropor. Mater. 4 (6) (1995) 407433.

[9] H.C. Foley, Nanoporous carbons and related materials for

small molecule separations, Abstr. Papers Am. Chem. Soc.

211 (1996) 2.

[10] M. Acharya, M.S. Strano, J. Mathews, S.J.L. Billinge, et al.,

Simulation of nanoporous carbons: a chemically constrained

structure, Philos. Mag. B 79 (10) (1999) 14991518.

[11] R.K. Mariwala, H.C. Foley, Calculation of micropore sizes

in carbogenic materials from the methyl-chloride adsorption-

isotherm, Ind. Eng. Chem. Res. 33 (10) (1994) 23142321.

[12] M. Acharya, B.A. Raich, H.C. Foley, M.P. Harold, J.J. Lerou,

Metal-supported carbogenic molecular sieve membranes:


14/14


synthesis and applications, Ind. Eng. Chem. Res. 36 (8) (1997)

29242930.

[13] M. Acharya, H.C. Foley, Spray-coating of nanoporous carbon

membranes for air separation, J. Membr. Sci. 161 (1999) 15.

[14] M.B. Shiflett, H.C. Foley, Ultrasonic deposition of highselectivity nanoporous carbon membranes, Science 285 (17)

(1999) 19021905.

[15] D.S. Lafyatis, J. Tung, H.C. Foley, Poly(furfuryl alcohol)-

derived carbon molecular-sieves dependence of

adsorptive properties on carbonization temperature, time and

poly(ethylene glycol) additives, Ind. Eng. Chem. Res. 30 (5)

(1991) 865873.

[16] M.S. Kane, L.C. Kao, R.K. Mariwala, D.F. Hilscher, H.C.

Foley, Effect of porosity of carbogenic molecular sieve

catalysts on ethylbenzene oxidative dehydrogenation, Ind.

Eng. Chem. Res. 35 (10) (1996) 33193331.

[17] M.G. Stevens, H.C. Foley, Alkali metals on nanoporous

carbon: new solid-base catalysts, Chem. Commun. (6) (1997)

519520.

[18] M.S. Strano, H.C. Foley, Deconvolution of permeance through

supported nanoporous membranes, AIChE J. 46 (3) (2000)

651658.

[19] M. Stevens, Cesium/nanoporous carbon composite materials:

synthesis, characterization, and base catalysis, Ph.D. Thesis,University of Delaware, Newark, 1999 (Chapter 2).

[20] P.C. Carman, Flow of Gases through Porous Media, Academic

Press, London, 1956.

[21] R. Nobrega, H. Balmann, P. Aimar, V. Sanchez, Transfer of

dextran through ultrafiltration membranes: a study of rejection

data analysed by gel permeation chromatography, J. Membr.

Sci. 45 (1989) 1736.

[22] K.A. Granath, Solution properties of branched dextrans, J.

Colloid Interf. Sci. 13 (1958) 308.

[23] K.A. Smith, C.K. Colton, E.W. Merrill, L.B. Evans,

Convective transport in a batch dialyzer: determination of

the true membrane permeability from a single measurement,

AIChE Symp. Ser. 64 (1968) 45.

ultrafiltration membrane synthesis by nanoscale templating

Documents