- Email: [email protected]
New and modified anodic alumina membranes
Journal of Membrane Science 206 (2002) 375–387
New and modified anodic alumina membranes Part III. Preparation and characterisation by gas diffusion of 5 nm pore size anodic alumina membranes夽 Hélio de L. Lira a , Russell Paterson b,∗,1 a
b
Departamento de Engineering de Materiais, Universidade Federal da Para´ıba, Campina Grande 58.109-970, Brazil Colloid and Membrane Science Research Group, Department of Chemistry, University of Glasgow, Glasgow G12 8QQ, Scotland, UK Received 11 April 2001; received in revised form 14 September 2001; accepted 19 November 2001
Abstract This paper describes methods for preparation of anodic alumina membranes with pore diameters 65, 50, 40, 11, and 5 nm, Membranes were separated from the aluminium anode by applied electrical pulses rather than by chemical treatments normally used. In this way pore enlargement caused by etching was effectively avoided; particularly important for the smallest pore sizes. For the two smallest pore sizes direct observation by SEM was inconclusive and a gas diffusion technique was used. For a range of eight low molecular weight gases the permeation mechanism of these anodic alumina membranes was Knudsen diffusion and this was used to estimate pore diameters. For the larger pore sizes, good agreements were obtained for estimates from, SEM and Knudsen methods. For the smallest pores Knudsen estimates alone were used. © 2002 Elsevier Science B.V. All rights reserved. Keywords: Anodic alumina; Membrane preparation; Ceramic membrane; Gas permeation; Knudsen diffusion
1. Introduction Anodic alumina membranes have been studied extensively [1–8]. During anodic oxidation of aluminium the porous film layer adheres strongly to the metal substrate and it is difficult to remove. Normally, anodic alumina membranes are released from the aluminium substrate by chemical attack, either by dissolution of the aluminium foil or by dissolution of the barrier film of alumina which forms a continuous layer at the surface of the aluminium during growth of the alumina 夽 Part I and II of this series is cited in the list of reference as [1,2], respectively. ∗ Corresponding author. E-mail address: [email protected] (R. Paterson). 1 Now retired.
film [9–22]. Such methods normally lead to enlargement of pores due to partial dissolution of pore walls. In a new method developed in Part I [1] a large voltage pulse is applied for a very short time to the aluminium anode after anodisation and the membrane is separated instantly. The barrier layer film of anodic alumina is completely removed and all pores are open. The great advantage of this method is that, being electrical; it is fully controllable and does not involve dissolving chemicals. In this paper, membranes with a range of precisely controlled pore diameters were obtained, in particular, membranes with pores sizes as small as (5 nm) were produced using low anodisation voltages, just above the polarisation potential. Although several direct methods for estimating mean pore sizes are available [23–26], they generally involve some ambiguity and/or require preliminary
0376-7388/02/$ – see front matter © 2002 Elsevier Science B.V. All rights reserved. PII: S 0 3 7 6 - 7 3 8 8 ( 0 1 ) 0 0 7 8 2 - 7
376
H.de L. Lira, R. Paterson / Journal of Membrane Science 206 (2002) 375–387
Nomenclature d n n p r t A D J P M R T U V
pore size diameter (m) number of pores (per m2 ) moles of the gas pressure (Pa) mean pore radius (m) time (s) surface area of membrane (m2 ) cell diameter of anodic alumina film (m) gas flux density (mol s−1 m−2 ) permeability (mol s−1 m−2 Pa−1 ) molecular mass of the gas (kg mol−1 ) gas constant (J mol−1 K−1 ) absolute temperature (K) voltage of anodisation (V) volume of the cell (m3 )
Greek symbols δ thickness of the membrane (m) ε porosity λ mean free path (nm) average velocity (m s−1 ) µ shape factors
sample modification, which can mask small pores. For example, for high-resolution SEM the sample must be electrically conducting and so, for alumina, require pre-coating with a conducting layer. For the smallest pore sizes this can lead to partial blocking or pore filling. Gas permeability measurements have been applied by several authors to characterise microporous membranes [27–31], but unknown factors such pore shape, layer thickness, homogeneity and local defects make such methods unreliable. Anodic alumina membranes have none of these disadvantages. The membrane has a planar structure with uniform parallel pores in a hexagonal array, without intersection and are highly reproducible. The first objective of this paper was to prepare anodic alumina membranes with regular structure and well defined small pore diameter, especially in the 5–11 nm range, using the new method to separate the anodic alumina film from the aluminium anode. The second was to characterise these membranes by gas
diffusion and to compare the results with direct observations using SEM and to investigate empirical correlations between pore size and anodisation conditions. For pore sizes between 5 and 11 nm direct observation by SEM is not reliable. 1.1. Gas diffusion There is a large literature on transport of gases and vapours through porous membranes [32–40]. Diffusion may occur by Poiseuille, surface and Knudsen mechanisms which may apply separately or in combination. Experimental data obtained in this study (using low molecular weight gases) and analysed in Section 3, indicate that only Knudsen diffusion [41] occurs and so the theoretical permeability will be 2εµk νr P = Pk = (1) 3RTδ where P is the total permeability (in mol s−1 m−2 Pa−1 ), and Pk the Knudsen permeability coefficient, ε the porosity, which is nπ r2 , where n is the number of pores m−2 , r the mean pore radius of the membrane (m), µk the shape factor, which is unity for uniform straight pores normal to the planar surface of the membrane, δ the thickness of the membrane (m), R the gas constant (8.314 J mol−1 K−1 ), T the absolute temperature (K), the average velocity (m s−1 ) which is for an ideal gas. 8RT 1/2 ν= (2) πM where M is the molecular mass of the gas (kg mol−1 ). From Eqs. (1) and (2) the permeability will be Knr 3 P= √ δ M
(3)
√ where K is the constant 4/3 2π/(RT) Therefore, for Knudsen diffusion in uniform pores, the radius of the pores can be determined provided the pore density and the membrane (or layer) thickness can be estimated. For a multilayer membrane, with m layers of uniform thickness, the total permeability P is related to the permeabilities of the each layer Pm (m = 1, 2, √ 3) by Eq. (4) and from Eq. (3) will also retain the M
H.de L. Lira, R. Paterson / Journal of Membrane Science 206 (2002) 375–387
relationship if each layer obeys Knudsen law independently 1 1 1 1 1 = + + + ··· + P P1 P2 P3 Pm √ M δ1 δ1 δ3 δm = + (4) + + ··· + 3 K nm rm n1 r13 n2 r23 n3 r33 where P is the permeability of the multilayer membrane, Pm the permeability, δ m the thickness and rm is the mean pore radius of each layer of the membrane. 2. Experimental 2.1. Membrane preparation The membrane preparation was described in detail elsewhere [1,5] and can be summarised as follows: anodic alumina membrane were prepared by anodisation of aluminium foil, purity 99.999%, 70 m supplied by Johnson Matthey, Materials Technology, UK. The anodisation process was performed by using oxalic solution 3% (w/w), at 10 ◦ C under constant voltage conditions (the larger the voltage the larger the pore diameters). Remaining aluminium was removed by a voltage pulse In this method, anodic alumina coated aluminium electrode was immersed in a solution of HClO4 (72% w/w) and (CH3 CO)2 (98%) 1:1 and applied a voltage of 15 V approximately greater then final voltage of anodisation, during 1–3 s. The anodic alumina film separates from the aluminium electrode immediately. At the same time the barrier layer was removed from the film and an open pore membrane is obtained. Conditions of preparations are summarised on Table 1.
377
2.2. Scanning electron microscopy The pore structure of the membrane was observed with a high-resolution scanning electron microscope (SEM, Hitachi S-900). The anodic alumina films were first coated with thin layer of carbon (or gold) by sputtering. 2.3. Gas permeation The gas permeation apparatus is shown in Fig. 1. This system enabled the determination of gas permeability through a flat membrane. Membrane samples 10 mm in diameter were supported in the cell and sealed by a rubber O-ring with 2 mm inner diameter (the exposed membrane area A (m2 ) was π d2 /4). The gas pressure on the high pressure side was raised to 2 atm after the cell had been completely filled with the test gas at 1 atm. The permeability was measured by monitoring pressure in both the high and low pressure sides of the permeation cell as function of time. For each gas three test runs were made. The experiments were performed in a thermostated room at 293 K. The gas flow (J) in the steady state was obtained from Eq. (5). Nitrogen and carbon dioxide were purchased from BOC Ltd., UK. Neon, propane, argon, nitrogen, hydrogen, helium were supplied by Messer Griesheim, Germany. All gases were at least 99.99% of purity. In the experimental measurements, gas permeates from the high to the low pressure side of the test membrane, due to a transmembrane pressure difference, p = p − p > 0. Under experimental conditions the high pressure p is kept constant and as the experiment proceeds the low pressure p increases. The low pressure side has constant volume V and the rate
Table 1 Conditions of preparation of anodic alumina membranesa Sample
T (◦ C)
U1st (V)
t1st (h)
U2nd (V)
t2nd (h)
δ tot (m)
a (m)
b (m)
72/1 60/3 40/5 72/1/10/16 72/1/5/20 72/1/5/40
10 10 10 10 10 10
72 60 40 72 72 72
1 3 5 1 1 1
– – – 10 5 5
– – – 16 20 40
23 37 25 32 29 35
– – – 23 23 23
– – – 9 6 12
a U : initial voltage of anodisation; U 1st 2nd : final voltage of anodisation; T: temperature of electrolyte solution; t1st : time of anodisation for initial voltage; t2nd : time of anodisation for final voltage; δ tot : total thickness of the membrane; a: thickness of the support layer; b: thickness of the active layer.
378
H.de L. Lira, R. Paterson / Journal of Membrane Science 206 (2002) 375–387
Fig. 1. Schematic diagram of gas permeation system.
of permeation of gas dn/dt can be evaluated from the rate of pressure increased dp /dt. For gases obeying the ideal gas law, as here, the flux density, J across a membrane of area A is given by Eq. (5). J =
V dp ART dt
(5)
and the permeability P (mol s−1 m−2 Pa−1 ), defined by Eq. (6). J P= (6) p 3. Results and discussion The code used to identify the method of preparation of various samples discussed throughout this paper was as follows: 72/1–represents anodic alumina film prepared at constant voltage, 72 V for 1 h, 60/3 (60 V for 3 h), 40/5 (40 V for 5 h). The notation for bilayer membranes, e.g. 72/1/10/16, refers to a preparation at constant voltage, 72 V for 1 h, followed by a progressively reduction from 72 to 10 V in decrements of 2 V/min and then maintained at 10 V for 16 h. Other descriptions follow the same code.
3.1. Microstructure of anodic alumina membrane Plates 1–3 are scanning electron micrographs of the anodic alumina film for the sample 72/1. The top surface (Plate 1) shows caps formed during the anodisation using acid to remove aluminium substrate. There are no open pores in this film and all caps have a hemi-spherical shape evenly distributed on the surface in an idealised hexagonal cell structure. The width of the caps corresponds to D, the width of the hexagonal cell. No gas permeation was observed in this case. Plate 2 shows the cross-sectional view for the anodic film after removal of the caps by the electrical pulse technique. The pores appear at the centre of serrated cups, conical depressions in the surface, caused by the removal of the caps. The micrograph shows high regularity in the structure with straight and parallel pores and all caps removed. Plate 3 shows the top surface with all pores open. The pore density is equal to 4.0 × 1013 pores m−2 . The pore diameter (64 ± 14 nm) is less easily determined, due to uneven deposits of gold (or carbon) deposited in the open pores. However, it was possible to estimate the distance between the centres of the pores, 180 ± 14 nm which also corresponds to D, the width of the hexagonal cell.
H.de L. Lira, R. Paterson / Journal of Membrane Science 206 (2002) 375–387
Plate 1. Scanning electron micrograph of anodic alumina film for sample 72/1, showing caps on top surface after dissolving aluminium substrate by acid. (Magnification: 50,000×).
379
Plate 3. Scanning electron micrograph of anodic alumina film for sample 72/1, showing top surface view with open pores after remove caps and aluminium substrate by the electrochemical method. (Magnification: 50,000×).
Plate 2. Scanning electron micrograph of anodic alumina film for sample 72/1, showing cross-sectional view with open pores on top surface after removing caps and aluminium substrate by electrochemical treatment. (Magnification: 50,000×).
380
H.de L. Lira, R. Paterson / Journal of Membrane Science 206 (2002) 375–387
Plates 4 and 5 show micrographs of the anodic alumina film 72/1/10/16. Plate 4 shows the top surface, corresponding to the small pore active layer. There are dark spots ∼30 nm on the micrograph. It seems most likely that these are gold deposit-filled serrated cups (as shown in Plate 2, for the 72/1 sample) rather than the pores themselves and so, 30 nm corresponds to the hexagonal cell dimension, D. There is some evidence of small pores in the middle of these dark areas, but are too indistinct to measure. The pore diameter predicted by the hexagonal cell model and gas permeation, discussed below is nearer 10 nm. The pore density, 7 × 1014 pores m−2 , was estimated from this image assuming each spot corresponds to one obscured small pore. Pore densities for the smallest pores prepared at an anodisation of 5 V (72/1/5/20 and 72/1/5/40) were estimated similarly. The cross-sectional view of the bilayer structure of 72/1/10/16 (Plate 5) shows clearly the interface where the larger (∼64 nm) pore of the support branched into smaller pores when the anodisation voltage was reduced to 10 V. Thickness of support formed at 72 V for 1 h estimated in the Plate 5 was 23 m. This is in good agreement with measurement made by using a micrometer (for sample 72/1) and the thickness for active layer formed at 10 V for 16 h was 9 m.
Plate 4. Scanning electron micrograph of anodic alumina film for sample 72/1/10/16, showing top surface view with open pores on top surface after removing caps and aluminium substrate by electrochemical treatment. (Magnification: 112,500×).
Plate 5. Scanning electron micrograph of anodic alumina film for sample 72/1/10/16, showing cross-sectional view with two layer (23 mm created at 72 V for 1 h and 9 mm at 10 V for 16 h). (Magnification: 1450×).
H.de L. Lira, R. Paterson / Journal of Membrane Science 206 (2002) 375–387
381
Table 2 Comparison of calculated and observed values for pore diameter, pore density and porosity of anodic alumina membranes Membrane sample
72/1 60/3 40/5 72/1/10/16b 72/1/5/20b a b
Hexagonal modela
SEM (observed)
Pore diameter (nm)
Pore density (pores m−2 )
67 55.8 37.2 9.3 4.6
2.8 4.1 9.2 1.5 6.0
× × × × ×
1013 1013 1013 1015 1015
Porosity Eq. (7)
Pore diameter (nm)
Pore density (pores m−2 )
Porosity Eq. (7)
0.10 0.10 0.10 0.10 0.10
64 ± 14 50 ± 6 40 ± 8 – –
4.0 × 1013 5.6 × 1013 7.1 × 1013 7 × 1014 1.5 × 1015
0.13 ± 0.06 0.11 ± 0.02 0.09 ± 0.03 – –
Based on D = 2.8 nm V−1 , D/d = 3.0, Eq. (8), Table 3. Pore diameters refer to the active layer (small pore) surface of the asymmetric membrane.
A compilation of theoretical and observed values for pore density and porosity is given in Table 2. The SEM images show well-defined pore structures for large pore samples. The SEM technique is limited to large pores, >30 nm, due to the effect of gold or carbon deposits (sample preparation) in masking small features. For membranes with pores <20 nm other methods were required. 3.2. Estimation of pore diameters from a hexagonal cell model of the membrane surface The microstructure of anodic alumina membranes (Plates 1–3) shows clearly that the surface consists of hexagonal cells, diameter D. Plates 1 and 2 show the 72/1 membrane with and without caps. Plates 2 and 3 show pores, diameter d, Table 2. For the surfaces of the smallest pore samples (with anodisation voltages of 10 and 5, respectively) the pore diameters are to a greater extent masked by the conducting surface coating required for their preparation, as noted above. For these however, the larger (hexagonal) cell dimension, D, may be estimated with much more confidence. It was observed, Table 3, as in other studies, [3,8,13,15,16,22] that the ratio D/d is constant,
and independent of electrolysis voltage, U, when the same experimental conditions apply. In addition the ratio D/U = 2.8 ± 0.3 nm V−1 , for the three measurements √ available. Since the area of the hexagon cell is ( 3/2)D 2 and the area of the pore is πd2 /4, the porosity ε, and pore diameter d, may be estimated by Eqs. (7) and (8). 2 π d ε= √ (7) 2 3 D d=
D 2.8U = 3 3
(8)
Pore diameters, densities and porosities calculated by this hexagonal cell model show good agreement with SEM estimates available. In addition the model may be used to estimate the smallest pores membranes prepared at 10 (sample 72/1/10/16) and 5 V (sample 72/1/5/20) as 9.3 and 4.6 nm, respectively, Table 2. 3.3. Gas permeability Figs. 2 and 3 show the transmembrane pressure dependence of the flux of various gases at
Table 3 Values of cell diameter D, and pore diameter d, of homogeneous anodic alumina membranes from direct microscopy observation Sample
Anodisation voltage, U (V)
Cell diameter, D (nm)
D/U (nm V−1 )
Pore diameter, d (nm)
D/d
72/1 60/3 40/5 Average value
72 60 40 –
180 ± 14 160 ± 14 120 ± 9 –
2.5 2.7 3 2.8
64 ± 14 50 ± 6 40 ± 8 –
2.80 3.20 3.00 3.00
382
H.de L. Lira, R. Paterson / Journal of Membrane Science 206 (2002) 375–387
Fig. 2. Gas flux vs. transmembrane pressure with different gasses for the sample 72/1.
a constant temperature (293 K) for samples 72/1 and 72/1/10/16, respectively. The observed fluxes were perfectly proportional to the transmembrane pressure difference, indicating that the permeabilities were constant. This was also true of all other membranes tested in this study. Permeabilities were calculated directly from the slopes of the curves, Eq. (6), Table 4. (If the true effect of changing pore size upon permeabilities of homogeneous membranes is required these data need to be multiplied by the corresponding membrane thickness, cf.
Eq. (3)). For asymmetric membranes (72/1/10/16, 72/1/15/20 and 72/1/5/40) the permeabilities of the active layer were calculated from the multilayer permeability relationship, Eq. (4). For example, 1 P72/1/10/16
=
1 P72/1
+
1 P10/16
(9)
where P72/1 and P72/1/10/16 are experimental permeabilities for the samples 72/1 and 72/1/10/16, respectively, and P10/16 , the permeability for the active layer of the samples 72/1/10/16. A compilation of these
H.de L. Lira, R. Paterson / Journal of Membrane Science 206 (2002) 375–387
383
Fig. 3. Gas flux vs. transmembrane pressure with different gases for the sample 72/1/10/16.
calculations are shown in Table 5. It can be observed, by comparison with Table 4, that the addition of the thin active layer to 72/1 giving 72/1/10/16 has decreased the permeability by approximately 50%. For samples 72/1/5/20 and 72/1/5/40, the reductions are approximately 90 and 96%, respectively. (For practical, commercial membranes the active layers would be much thinner.) Here they are made thick enough to allow more accurate estimates of their thicknesses from SEM micrographs, Plate 5.
For membranes 72/1/5/20 and 72/1/5/40, the active layers should have the same pore size and pore density, but, since the anodisation at 5 V is doubled for the 5/40 should be twice as thick as that of the 5/20, Eq. (3), and so have a permeability half as large. This is not so (although it is smaller) and cannot be explained easily, except possibly by an unexpectedly important contribution of the intermediate layer created when the voltage is reduced progressively from 72 to 5 V during the anodisation. Overall the effect should be
384
H.de L. Lira, R. Paterson / Journal of Membrane Science 206 (2002) 375–387
Table 4 Experimental gas permeabilities (mol s−1 m−2 Pa−1 ) × 107 , for anodic alumina membranes, including support plus active layer (multilayer) Gases
H2 He Ne N2 O2 Ar CO2 C3 H8
Membranes 72/1
60/3
40/5
72/1/10/16
72/1/5/20
72/1/5/40
592.30 463.50 215.60 200.80 176.10 158.10 138.70 159.40
314.30 223.20 108.50 86.30 79.30 74.20 72.50 72.70
394.10 287.90 133.70 110.00 94.60 84.80 76.90 82.20
207.00 154.90 54.80 64.80 66.60 51.70 45.00 49.50
70.50 54.60 25.30 23.50 22.60 21.40 16.40 19.80
23.75 17.47 9.54 7.23 7.49 6.86 6.45 8.99
smaller for the membrane with the thicker active layer, 72/1/5/40. For all membranes studied, the permeabilities are inversely proportional to the square root of the molecular weight of the permeant gas, Figs. 4 and 5, proving that the diffusion mechanism is Knudsen flow, Eqs. (3) and (4). Pore diameters for the homogeneous membranes and the active layers of the bilayer membranes were calculated from Eq. (3), and using experimental values for pore densities and membrane/layer thicknesses, of Table 2. Pore densities were obtained from the micrographs even when the pores themselves were small and unresolved, using the assumption that they lay at the centres of the much larger and resolvable cell structure, as discussed in Section 3.1. (The agreement between experimental and hexagonal model pore
densities is not good. Observed SEM pore densities are much lower than the hexagonal model predicts. There is no explanation for this at this time, except it should be stressed that the model may be expected to be less reliable in the limit when extrapolated to the lowest anodisation voltages, close to the polarisation potential.) The results are given in Table 6.
Table 5 Experimental permeability (mol s−1 m−2 Pa−1 ) × 107 , for the active layer of the multilayer membranes Gases
H2 He Ne N2 O2 Ar CO2 C 3 H8 a
Membranes 72/1/10/16a
72/1/5/20a
72/1/5/40a
318.21 232.65 73.48 95.68 107.11 76.82 66.61 71.80
80.02 61.89 28.66 26.61 25.93 24.75 18.60 22.61
24.74 18.15 9.98 7.50 7.82 7.17 6.76 9.53
From Eq. (9) and experimental value of the support 72/1, Table 4
Fig. 4. Permeability of eight test gases plotted against the inverse square root of their molecular weights for the homogeneous membrane sample 72/1.
H.de L. Lira, R. Paterson / Journal of Membrane Science 206 (2002) 375–387
385
Table 7 Comparison of pore size estimated by anodisation voltage, gas permeation and SEM Pore diameter (nm) Membrane sample
Knudsen estimate
72/1 60/3 40/5 72/1/10/16a 72/1/5/20a 72/1/5/40a
59.3 48.7 41.8 13.08 5.86 5.07
a b
Fig. 5. Permeability of eight test gases plotted against the inverse square root of their molecular weights for the bilayer membrane sample 72/1/10/16.
For all eight gases studied there is excellent agreement on estimates of pore diameters, as shown by the low standard deviations on the averaged values. The Knudsen mechanism for diffusion in these membrane
± ± ± ± ± ±
1.4 0.6 0.7 0.68 0.21 0.30
Hexagonal cell model Eq. (8)b
SEM
67.00 55.80 37.20 9.30 4.60 4.60
64 ± 14 50 ± 6 40 ± 8 – – –
Pore size of the active layer of the membrane. D = 2.8 nm V−1 , D/d = 3.0.
is therefore amply validated, as can be seen from comparative data directly from SEM for the larger pore membranes where direct comparisons can be made, Table 7. For the smaller pored membranes the Knudsen technique provides unique experimental estimates of pore size, which is supported by the idealised model calculations based on the hexagonal cell model, Eq. (8). Knudsen techniques can only be applied with confidence for membranes with parallel, uniform pores. It is therefore of limited quantitative application for complex pore structures, due to inhomogeneities and uncertainty in evaluations of pore structure factors, Eq. (1). It has, however, been applied most successfully to track etched membranes by Yamazaki et al. [42].
Table 6 Pore diameter (nm) of active layer of anodic alumina membrane estimated by using Knudsen law, Eq. (3) and pore densities estimated by SEM Gases
Membranes 72/1
60/3
40/5
72/1/10/16a
72/1/5/20a
72/1/5/40a
H2 He Ne N2 O2 Ar CO2 C3 H 8
56.45 58.39 59.16 61.11 59.81 59.88 58.24 61.01
47.87 47.94 49.29 48.30 48.02 48.74 49.14 49.19
41.85 42.31 42.84 42.46 41.29 41.32 40.63 41.55
12.93 13.07 11.64 13.45 14.28 13.26 12.85 13.18
5.53 5.70 5.76 5.95 6.03 6.16 5.69 6.07
4.71 4.77 5.11 4.91 5.09 5.14 5.12 5.74
Average S.D.
59.26 1.45
48.56 0.56
41.78 0.68
13.08 0.68
5.86 0.21
5.07 0.30
a
The pore diameter refer to active surface layer in these asymmetric membranes, underlined.
386
H.de L. Lira, R. Paterson / Journal of Membrane Science 206 (2002) 375–387
4. Conclusions Using an electrical pulse method allowed anodic alumina membranes to be prepared unaffected by the side effects of chemical erosion of pores, normally associated with conventional methods for membrane release from the aluminium anode. As a result membranes with pore diameters as small as 5 nm were successfully produced with clean uneroded pores. The method is precise and simple, using anodic voltages marginally above the polarisation potential of the aluminium anode. Such membranes have a range of potential uses and their regular geometry makes them ideally suited to both practical and research applications. A series of membranes with widely differing pore sizes were prepared by this method. Membranes with smaller pore sizes were prepared supported on thicker larger pored substrates in a bilayer format. Standard SEM measurements were made to determine membrane (or layer) thicknesses and pore densities, but could not resolve pore dimensions for pores with diameters, less than ca. 15 nm. Other methods for measurement of pore diameters were required. Of the two methods used, one was based on a model or idealised concept of the anodic alumina cell structure, the other experimental using gas diffusion permeability measurements. Pore sizes estimated, using an idealised model of the hexagonal cell structure of anodic alumina, were in good agreement with SEM data for large pores where direct comparison was possible. Gas permeabilities for eight test gases were shown to obey Knudsen’s law. From these data estimates of pore diameters of membranes and the active layers of bilayer membranes were made. They too gave excellent agreement with direct SEM measurements for large pore membranes and consequently were used with confidence to estimate pore sizes as small as 5 nm. The gas diffusion technique can only be used with confidence for membranes with regular parallel pore structure, but has been used successfully for track etched membranes (TEMs), [42].
Acknowledgements We are grateful for a scholarship from CNPq (Brazil) to Helio De L. Lira (ref. Proc. 20.1780/92.5)
during the period of this research in Glasgow University, 1994–1997. References [1] P. Mardilovich, A.N. Govyadinov, N.I. Mukhurov, A.M. Rzhevskii, R. Paterson, New and modified anodic alumina membranes. Part I. Thermotreatment of anodic alumina membranes, J. Membr. Sci. 98 (1995) 131. [2] P.P. Mardilovich, A.N. Govyadinov, N.I. Mazurenko, R. Paterson, New and modified anodic alumina membranes. Part II. Comparison of solubility of amorphous (normal) and polycrystalline anodic alumina membranes, J. Membr. Sci. 98 (1995) 143. [3] J. Randon, P.P. Mardilovich, A.N. Govyadinov, R. Paterson, Computer simulation of inorganic membrane morphology. Part III. Anodic alumina films and membranes, J. Colloid Interface Sci. 169 (1995) 335. [4] P.P. Mardilovich, A.N. Govyadinov, V. Kozharinov, R. Paterson, Polycrystalline anodic alumina—new ceramic membranes and matrices, in: P. Vincenzini (Ed.), Ceramics: Charting the Future, Techna. Srl., 1995, p. 2763. [5] H. de L. Lira, Ph.D. Thesis, Preparation and Properties of Ceramic and Surface Modified Ceramic Membranes, Glasgow University, 1996. [6] P.P. Mardilovich, A.N. Govyadinov, J. Randon, S. McFadzean, R. Paterson, New high temperature chemically resistant anodic alumina membranes: preparation, modification, properties, in: Proceedings of the 3rd International Conference on Inorganic Membranes, Worchester, MA, USA, 1994, 457. [7] A.N. Govyadinov, P.P. Mardilovich, V.V. Kozharinov, V.M. Gryaznov, R. Paterson, Composite systems based on anodic alumina membranes, in: Proceedings of the 3rd International Conference on Inorganic Membranes, Worchester, MA, USA, 1994, 541. [8] J. Randon, P.P. Mardilovich, A.N. Govyadinov, R. Paterson, Modelling the pore structure of anodic alumina membranes, in: Proceedings of the 3rd International Conference on Inorganic Membranes, Worchester, MA, USA, 1994, 613. [9] T.P. Hoar, N.F. Mott, Mechanism for the formation of porous anodic oxide films on aluminium, J. Phys. Chem. Solids 9 (1959) 97–99. [10] M. Mårtensson, K.E. Holmberg, C. Lofman, E.I. Eriksson, Some type of membranes for isotope separation by gaseous diffusion, in: Proceedings of the 2nd UN International Conference on Peaceful Uses of Atomic Energy, Vol. 4, 1958, pp. 395–404. [11] A.W. Smith, Process for producing an anodic aluminum oxide membrane, US Patent 3,850,762 (1974). [12] A.W. Smith, Porous anodic aluminum oxide membrane, J. Electrochem. Soc. 120 (8) (1973) 1068–1069. [13] G.E. Thompson, R.C. Furneaux, G.C. Wood, H.A. Richardson, J.S. Goode, Nucleation and growth of porous anodic films on aluminum, Nature 272 (1978) 433–435. [14] R.C. Furneaux, W.R. Rigby, A.P. Davidson, Porous films and method of forming them, EP 0,178,831 Al, 1986, US Patent 4,687,551 (1987).
H.de L. Lira, R. Paterson / Journal of Membrane Science 206 (2002) 375–387 [15] R.C. Furneaux, W.R. Rigby, A.P. Davidson, The formation of controlled porosity membranes from anodically oxidised aluminum, Nature 337 (1989) 147. [16] G.E. Thompson, G.C. Wood, Porous anodic film formation on aluminium, Nature 290 (1981) 230. [17] Anotec Separations, Anopore: a new inorganic membrane unique in the world of microfiltration, Brochure, Whatman International Ltd., Maidstone, UK. [18] R.C. Furneaux, R.W. Philpott, D.M. Jenkins, Controlled pore size membranes produced by anodizing aluminium, in: Proceedings of the First International Conference on Inorganic Membranes, Montpellier, 1989, pp. 47–53. [19] R.C. Furneaux, M.C. Thornton, Porous “ceramic” membranes from anodising aluminium, Adv. Ceram. Chem. Process Eng. 43 (1988) 93. [20] K. Itaya, S. Sugawara, K. Arai, S. Saito, Properties of porous anodic aluminum oxide films as membranes, J. Chem. Eng. Jpn. 17 (1984) 514–520. [21] N. Itoh, K. Kato, T. Tsuji, M. Hongo, Preparation of a tubular anodic aluminum oxide membrane, J. Membr. Sci. 117 (1996) 189–196. [22] T. Parlovic, A. Ignatiev, Optical and microstructural properties of anodically oxidised aluminum, Thin Solid Films 138 (1986) 97–109. [23] J. Rocek, P. Uchytil, Evaluation of selected methods for characterisation of ceramic membranes, J. Membr. Sci. 89 (1994) 119. [24] Y. Yuxun, W. Diyong, Pore size of composite inorganic membrane, in: Proceedings of the third International Conference on Inorganic Membranes, Worchester, July 1994, p. 525. [25] P. Dietz, P.K. Hansma, O. Inacker, H.D. Lehmann, K.H. Hermann, Surface pore structures of micro- and ultrafiltration membranes imaged with the atomic force microscope, J. Membr. Sci. 65 (1992) 101. [26] P. Neufeld, N.K. Nagpaul, R. Ashdown, M. Akbar, Crystallisation of anodic Al2 O3 , Electrochim. Acta 17 (1972) 1543–1546. [27] P. Uchytil, Gas permeation in ceramic membranes. Part I. Theory and testing of ceramic membranes, J. Membr. Sci. 97 (1994) 139. [28] P. Uchytil, Z. Broz, Gas permeation in ceramic membranes. Part II. Modelling of gas permeation through ceramic membrane with one supported layer, J. Membr. Sci. 97 (1994) 145. [29] F.W. Altena, H.A.M. Knoef, H. Heskamp, D. Bargeman, C.A. Smolders, Some comments on the applicability of gas
[30]
[31]
[32]
[33]
[34]
[35] [36]
[37]
[38]
[39]
[40]
[41] [42]
387
permeation methods to characterise porous membranes based on improved experimental accuracy and data handling, J. Membr. Sci. 12 (1983) 313. P. Uchytil, Z. Wagner, J. Rocek, Z. Broz, Possibility of pore size determination in separation layer of ceramic membrane using permeation method, J. Membr. Sci. 103 (1995) 151. P. Uchytil, X.Q. Nguyen, Z. Broz, Characterisation of membrane skin defects by gas permeation method, J. Membr. Sci. 73 (1992) 47. K. Keizer, R.J.R. Uhlhorn, R.J. Van Vuren, A.J. Burggraaf, Gas separation mechanisms in microporous modified ␥-Al2 O3 membranes, J. Membr. Sci. 39 (1988) 285. R.J.R. Uhlhorn, K. Keizer, A.J. Burggraaf, Gas and surface diffusion in modified ␥-alumina systems, J. Membr. Sci. 46 (1989) 225. R.J.R. Uhlhorn, K. Keizer, A.J. Burggraaf, Gas transport and separation with ceramic membranes. Part I. Multilayer diffusion and capillary condensation, J. Membr. Sci. 66 (1992) 259. P. Levitz, Knudsen diffusion and excitation transfer in random porous media, J. Phys. Chem. 97 (1993) 3818. R.L. Rowell, S.A. Carrano, A.J. Bethune, A.P. Malinauskas, Gas and vapour permeability: surface flow through porous media, J. Colloid Interface Sci. 37 (1971) 242. R.W. Schofield, A.G. Fane, C.J.D. Fell, Gas and vapour transport through microporous membranes. Part I. Knudsen–Poiseuille transition, J. Membr. Sci. 53 (1990) 159. R. Datta, S. Dechapanichkul, J.S. Kim, L.Y. Fang, Uehara, A generalised model for the transport of gases in porous, non-porous, and leaky membranes. Part I. Application to single gases, J. Membr. Sci., 75 (1992) 245. M. Bhandarkar, A.B. Shelekhin, A.G. Dixon, Y.H. Ma, Adsorption, permeation, and diffusion of gases in microporous membranes. Part I. Adsorption of gases on microporous glass membranes, J. Membr. Sci. 75 (1992) 221. K. Keizer, R.J.R. Uhlhorn, V.T. Zaspalis, A.J. Burggraaf, Transport and related (gas and vapour) separation in ceramic membranes, Key Eng. Mater. 61/62 (1991) 143. M. Knudsen, Kinetic Theory of Gases: Some Modern Aspects, 3rd Edition, Methuen & Co. Ltd., 1950. M. Yamazaki, L.P. Geraldo, R. Paterson, Characterisation of Polycarbonate Nuclear Track Etched Membranes by Means of a Gas Permeation Method, Nuclear Instruments and Methods in Physics Research, Vol. 418A, 1998, pp. 491–496.
New and modified anodic alumina membranes Part III. Preparation and characterisation by gas diffusion of 5 nm pore size anodic alumina membranes夽 Hélio de L. Lira a , Russell Paterson b,∗,1 a
b
Departamento de Engineering de Materiais, Universidade Federal da Para´ıba, Campina Grande 58.109-970, Brazil Colloid and Membrane Science Research Group, Department of Chemistry, University of Glasgow, Glasgow G12 8QQ, Scotland, UK Received 11 April 2001; received in revised form 14 September 2001; accepted 19 November 2001
Abstract This paper describes methods for preparation of anodic alumina membranes with pore diameters 65, 50, 40, 11, and 5 nm, Membranes were separated from the aluminium anode by applied electrical pulses rather than by chemical treatments normally used. In this way pore enlargement caused by etching was effectively avoided; particularly important for the smallest pore sizes. For the two smallest pore sizes direct observation by SEM was inconclusive and a gas diffusion technique was used. For a range of eight low molecular weight gases the permeation mechanism of these anodic alumina membranes was Knudsen diffusion and this was used to estimate pore diameters. For the larger pore sizes, good agreements were obtained for estimates from, SEM and Knudsen methods. For the smallest pores Knudsen estimates alone were used. © 2002 Elsevier Science B.V. All rights reserved. Keywords: Anodic alumina; Membrane preparation; Ceramic membrane; Gas permeation; Knudsen diffusion
1. Introduction Anodic alumina membranes have been studied extensively [1–8]. During anodic oxidation of aluminium the porous film layer adheres strongly to the metal substrate and it is difficult to remove. Normally, anodic alumina membranes are released from the aluminium substrate by chemical attack, either by dissolution of the aluminium foil or by dissolution of the barrier film of alumina which forms a continuous layer at the surface of the aluminium during growth of the alumina 夽 Part I and II of this series is cited in the list of reference as [1,2], respectively. ∗ Corresponding author. E-mail address: [email protected] (R. Paterson). 1 Now retired.
film [9–22]. Such methods normally lead to enlargement of pores due to partial dissolution of pore walls. In a new method developed in Part I [1] a large voltage pulse is applied for a very short time to the aluminium anode after anodisation and the membrane is separated instantly. The barrier layer film of anodic alumina is completely removed and all pores are open. The great advantage of this method is that, being electrical; it is fully controllable and does not involve dissolving chemicals. In this paper, membranes with a range of precisely controlled pore diameters were obtained, in particular, membranes with pores sizes as small as (5 nm) were produced using low anodisation voltages, just above the polarisation potential. Although several direct methods for estimating mean pore sizes are available [23–26], they generally involve some ambiguity and/or require preliminary
0376-7388/02/$ – see front matter © 2002 Elsevier Science B.V. All rights reserved. PII: S 0 3 7 6 - 7 3 8 8 ( 0 1 ) 0 0 7 8 2 - 7
376
H.de L. Lira, R. Paterson / Journal of Membrane Science 206 (2002) 375–387
Nomenclature d n n p r t A D J P M R T U V
pore size diameter (m) number of pores (per m2 ) moles of the gas pressure (Pa) mean pore radius (m) time (s) surface area of membrane (m2 ) cell diameter of anodic alumina film (m) gas flux density (mol s−1 m−2 ) permeability (mol s−1 m−2 Pa−1 ) molecular mass of the gas (kg mol−1 ) gas constant (J mol−1 K−1 ) absolute temperature (K) voltage of anodisation (V) volume of the cell (m3 )
Greek symbols δ thickness of the membrane (m) ε porosity λ mean free path (nm) average velocity (m s−1 ) µ shape factors
sample modification, which can mask small pores. For example, for high-resolution SEM the sample must be electrically conducting and so, for alumina, require pre-coating with a conducting layer. For the smallest pore sizes this can lead to partial blocking or pore filling. Gas permeability measurements have been applied by several authors to characterise microporous membranes [27–31], but unknown factors such pore shape, layer thickness, homogeneity and local defects make such methods unreliable. Anodic alumina membranes have none of these disadvantages. The membrane has a planar structure with uniform parallel pores in a hexagonal array, without intersection and are highly reproducible. The first objective of this paper was to prepare anodic alumina membranes with regular structure and well defined small pore diameter, especially in the 5–11 nm range, using the new method to separate the anodic alumina film from the aluminium anode. The second was to characterise these membranes by gas
diffusion and to compare the results with direct observations using SEM and to investigate empirical correlations between pore size and anodisation conditions. For pore sizes between 5 and 11 nm direct observation by SEM is not reliable. 1.1. Gas diffusion There is a large literature on transport of gases and vapours through porous membranes [32–40]. Diffusion may occur by Poiseuille, surface and Knudsen mechanisms which may apply separately or in combination. Experimental data obtained in this study (using low molecular weight gases) and analysed in Section 3, indicate that only Knudsen diffusion [41] occurs and so the theoretical permeability will be 2εµk νr P = Pk = (1) 3RTδ where P is the total permeability (in mol s−1 m−2 Pa−1 ), and Pk the Knudsen permeability coefficient, ε the porosity, which is nπ r2 , where n is the number of pores m−2 , r the mean pore radius of the membrane (m), µk the shape factor, which is unity for uniform straight pores normal to the planar surface of the membrane, δ the thickness of the membrane (m), R the gas constant (8.314 J mol−1 K−1 ), T the absolute temperature (K), the average velocity (m s−1 ) which is for an ideal gas. 8RT 1/2 ν= (2) πM where M is the molecular mass of the gas (kg mol−1 ). From Eqs. (1) and (2) the permeability will be Knr 3 P= √ δ M
(3)
√ where K is the constant 4/3 2π/(RT) Therefore, for Knudsen diffusion in uniform pores, the radius of the pores can be determined provided the pore density and the membrane (or layer) thickness can be estimated. For a multilayer membrane, with m layers of uniform thickness, the total permeability P is related to the permeabilities of the each layer Pm (m = 1, 2, √ 3) by Eq. (4) and from Eq. (3) will also retain the M
H.de L. Lira, R. Paterson / Journal of Membrane Science 206 (2002) 375–387
relationship if each layer obeys Knudsen law independently 1 1 1 1 1 = + + + ··· + P P1 P2 P3 Pm √ M δ1 δ1 δ3 δm = + (4) + + ··· + 3 K nm rm n1 r13 n2 r23 n3 r33 where P is the permeability of the multilayer membrane, Pm the permeability, δ m the thickness and rm is the mean pore radius of each layer of the membrane. 2. Experimental 2.1. Membrane preparation The membrane preparation was described in detail elsewhere [1,5] and can be summarised as follows: anodic alumina membrane were prepared by anodisation of aluminium foil, purity 99.999%, 70 m supplied by Johnson Matthey, Materials Technology, UK. The anodisation process was performed by using oxalic solution 3% (w/w), at 10 ◦ C under constant voltage conditions (the larger the voltage the larger the pore diameters). Remaining aluminium was removed by a voltage pulse In this method, anodic alumina coated aluminium electrode was immersed in a solution of HClO4 (72% w/w) and (CH3 CO)2 (98%) 1:1 and applied a voltage of 15 V approximately greater then final voltage of anodisation, during 1–3 s. The anodic alumina film separates from the aluminium electrode immediately. At the same time the barrier layer was removed from the film and an open pore membrane is obtained. Conditions of preparations are summarised on Table 1.
377
2.2. Scanning electron microscopy The pore structure of the membrane was observed with a high-resolution scanning electron microscope (SEM, Hitachi S-900). The anodic alumina films were first coated with thin layer of carbon (or gold) by sputtering. 2.3. Gas permeation The gas permeation apparatus is shown in Fig. 1. This system enabled the determination of gas permeability through a flat membrane. Membrane samples 10 mm in diameter were supported in the cell and sealed by a rubber O-ring with 2 mm inner diameter (the exposed membrane area A (m2 ) was π d2 /4). The gas pressure on the high pressure side was raised to 2 atm after the cell had been completely filled with the test gas at 1 atm. The permeability was measured by monitoring pressure in both the high and low pressure sides of the permeation cell as function of time. For each gas three test runs were made. The experiments were performed in a thermostated room at 293 K. The gas flow (J) in the steady state was obtained from Eq. (5). Nitrogen and carbon dioxide were purchased from BOC Ltd., UK. Neon, propane, argon, nitrogen, hydrogen, helium were supplied by Messer Griesheim, Germany. All gases were at least 99.99% of purity. In the experimental measurements, gas permeates from the high to the low pressure side of the test membrane, due to a transmembrane pressure difference, p = p − p > 0. Under experimental conditions the high pressure p is kept constant and as the experiment proceeds the low pressure p increases. The low pressure side has constant volume V and the rate
Table 1 Conditions of preparation of anodic alumina membranesa Sample
T (◦ C)
U1st (V)
t1st (h)
U2nd (V)
t2nd (h)
δ tot (m)
a (m)
b (m)
72/1 60/3 40/5 72/1/10/16 72/1/5/20 72/1/5/40
10 10 10 10 10 10
72 60 40 72 72 72
1 3 5 1 1 1
– – – 10 5 5
– – – 16 20 40
23 37 25 32 29 35
– – – 23 23 23
– – – 9 6 12
a U : initial voltage of anodisation; U 1st 2nd : final voltage of anodisation; T: temperature of electrolyte solution; t1st : time of anodisation for initial voltage; t2nd : time of anodisation for final voltage; δ tot : total thickness of the membrane; a: thickness of the support layer; b: thickness of the active layer.
378
H.de L. Lira, R. Paterson / Journal of Membrane Science 206 (2002) 375–387
Fig. 1. Schematic diagram of gas permeation system.
of permeation of gas dn/dt can be evaluated from the rate of pressure increased dp /dt. For gases obeying the ideal gas law, as here, the flux density, J across a membrane of area A is given by Eq. (5). J =
V dp ART dt
(5)
and the permeability P (mol s−1 m−2 Pa−1 ), defined by Eq. (6). J P= (6) p 3. Results and discussion The code used to identify the method of preparation of various samples discussed throughout this paper was as follows: 72/1–represents anodic alumina film prepared at constant voltage, 72 V for 1 h, 60/3 (60 V for 3 h), 40/5 (40 V for 5 h). The notation for bilayer membranes, e.g. 72/1/10/16, refers to a preparation at constant voltage, 72 V for 1 h, followed by a progressively reduction from 72 to 10 V in decrements of 2 V/min and then maintained at 10 V for 16 h. Other descriptions follow the same code.
3.1. Microstructure of anodic alumina membrane Plates 1–3 are scanning electron micrographs of the anodic alumina film for the sample 72/1. The top surface (Plate 1) shows caps formed during the anodisation using acid to remove aluminium substrate. There are no open pores in this film and all caps have a hemi-spherical shape evenly distributed on the surface in an idealised hexagonal cell structure. The width of the caps corresponds to D, the width of the hexagonal cell. No gas permeation was observed in this case. Plate 2 shows the cross-sectional view for the anodic film after removal of the caps by the electrical pulse technique. The pores appear at the centre of serrated cups, conical depressions in the surface, caused by the removal of the caps. The micrograph shows high regularity in the structure with straight and parallel pores and all caps removed. Plate 3 shows the top surface with all pores open. The pore density is equal to 4.0 × 1013 pores m−2 . The pore diameter (64 ± 14 nm) is less easily determined, due to uneven deposits of gold (or carbon) deposited in the open pores. However, it was possible to estimate the distance between the centres of the pores, 180 ± 14 nm which also corresponds to D, the width of the hexagonal cell.
H.de L. Lira, R. Paterson / Journal of Membrane Science 206 (2002) 375–387
Plate 1. Scanning electron micrograph of anodic alumina film for sample 72/1, showing caps on top surface after dissolving aluminium substrate by acid. (Magnification: 50,000×).
379
Plate 3. Scanning electron micrograph of anodic alumina film for sample 72/1, showing top surface view with open pores after remove caps and aluminium substrate by the electrochemical method. (Magnification: 50,000×).
Plate 2. Scanning electron micrograph of anodic alumina film for sample 72/1, showing cross-sectional view with open pores on top surface after removing caps and aluminium substrate by electrochemical treatment. (Magnification: 50,000×).
380
H.de L. Lira, R. Paterson / Journal of Membrane Science 206 (2002) 375–387
Plates 4 and 5 show micrographs of the anodic alumina film 72/1/10/16. Plate 4 shows the top surface, corresponding to the small pore active layer. There are dark spots ∼30 nm on the micrograph. It seems most likely that these are gold deposit-filled serrated cups (as shown in Plate 2, for the 72/1 sample) rather than the pores themselves and so, 30 nm corresponds to the hexagonal cell dimension, D. There is some evidence of small pores in the middle of these dark areas, but are too indistinct to measure. The pore diameter predicted by the hexagonal cell model and gas permeation, discussed below is nearer 10 nm. The pore density, 7 × 1014 pores m−2 , was estimated from this image assuming each spot corresponds to one obscured small pore. Pore densities for the smallest pores prepared at an anodisation of 5 V (72/1/5/20 and 72/1/5/40) were estimated similarly. The cross-sectional view of the bilayer structure of 72/1/10/16 (Plate 5) shows clearly the interface where the larger (∼64 nm) pore of the support branched into smaller pores when the anodisation voltage was reduced to 10 V. Thickness of support formed at 72 V for 1 h estimated in the Plate 5 was 23 m. This is in good agreement with measurement made by using a micrometer (for sample 72/1) and the thickness for active layer formed at 10 V for 16 h was 9 m.
Plate 4. Scanning electron micrograph of anodic alumina film for sample 72/1/10/16, showing top surface view with open pores on top surface after removing caps and aluminium substrate by electrochemical treatment. (Magnification: 112,500×).
Plate 5. Scanning electron micrograph of anodic alumina film for sample 72/1/10/16, showing cross-sectional view with two layer (23 mm created at 72 V for 1 h and 9 mm at 10 V for 16 h). (Magnification: 1450×).
H.de L. Lira, R. Paterson / Journal of Membrane Science 206 (2002) 375–387
381
Table 2 Comparison of calculated and observed values for pore diameter, pore density and porosity of anodic alumina membranes Membrane sample
72/1 60/3 40/5 72/1/10/16b 72/1/5/20b a b
Hexagonal modela
SEM (observed)
Pore diameter (nm)
Pore density (pores m−2 )
67 55.8 37.2 9.3 4.6
2.8 4.1 9.2 1.5 6.0
× × × × ×
1013 1013 1013 1015 1015
Porosity Eq. (7)
Pore diameter (nm)
Pore density (pores m−2 )
Porosity Eq. (7)
0.10 0.10 0.10 0.10 0.10
64 ± 14 50 ± 6 40 ± 8 – –
4.0 × 1013 5.6 × 1013 7.1 × 1013 7 × 1014 1.5 × 1015
0.13 ± 0.06 0.11 ± 0.02 0.09 ± 0.03 – –
Based on D = 2.8 nm V−1 , D/d = 3.0, Eq. (8), Table 3. Pore diameters refer to the active layer (small pore) surface of the asymmetric membrane.
A compilation of theoretical and observed values for pore density and porosity is given in Table 2. The SEM images show well-defined pore structures for large pore samples. The SEM technique is limited to large pores, >30 nm, due to the effect of gold or carbon deposits (sample preparation) in masking small features. For membranes with pores <20 nm other methods were required. 3.2. Estimation of pore diameters from a hexagonal cell model of the membrane surface The microstructure of anodic alumina membranes (Plates 1–3) shows clearly that the surface consists of hexagonal cells, diameter D. Plates 1 and 2 show the 72/1 membrane with and without caps. Plates 2 and 3 show pores, diameter d, Table 2. For the surfaces of the smallest pore samples (with anodisation voltages of 10 and 5, respectively) the pore diameters are to a greater extent masked by the conducting surface coating required for their preparation, as noted above. For these however, the larger (hexagonal) cell dimension, D, may be estimated with much more confidence. It was observed, Table 3, as in other studies, [3,8,13,15,16,22] that the ratio D/d is constant,
and independent of electrolysis voltage, U, when the same experimental conditions apply. In addition the ratio D/U = 2.8 ± 0.3 nm V−1 , for the three measurements √ available. Since the area of the hexagon cell is ( 3/2)D 2 and the area of the pore is πd2 /4, the porosity ε, and pore diameter d, may be estimated by Eqs. (7) and (8). 2 π d ε= √ (7) 2 3 D d=
D 2.8U = 3 3
(8)
Pore diameters, densities and porosities calculated by this hexagonal cell model show good agreement with SEM estimates available. In addition the model may be used to estimate the smallest pores membranes prepared at 10 (sample 72/1/10/16) and 5 V (sample 72/1/5/20) as 9.3 and 4.6 nm, respectively, Table 2. 3.3. Gas permeability Figs. 2 and 3 show the transmembrane pressure dependence of the flux of various gases at
Table 3 Values of cell diameter D, and pore diameter d, of homogeneous anodic alumina membranes from direct microscopy observation Sample
Anodisation voltage, U (V)
Cell diameter, D (nm)
D/U (nm V−1 )
Pore diameter, d (nm)
D/d
72/1 60/3 40/5 Average value
72 60 40 –
180 ± 14 160 ± 14 120 ± 9 –
2.5 2.7 3 2.8
64 ± 14 50 ± 6 40 ± 8 –
2.80 3.20 3.00 3.00
382
H.de L. Lira, R. Paterson / Journal of Membrane Science 206 (2002) 375–387
Fig. 2. Gas flux vs. transmembrane pressure with different gasses for the sample 72/1.
a constant temperature (293 K) for samples 72/1 and 72/1/10/16, respectively. The observed fluxes were perfectly proportional to the transmembrane pressure difference, indicating that the permeabilities were constant. This was also true of all other membranes tested in this study. Permeabilities were calculated directly from the slopes of the curves, Eq. (6), Table 4. (If the true effect of changing pore size upon permeabilities of homogeneous membranes is required these data need to be multiplied by the corresponding membrane thickness, cf.
Eq. (3)). For asymmetric membranes (72/1/10/16, 72/1/15/20 and 72/1/5/40) the permeabilities of the active layer were calculated from the multilayer permeability relationship, Eq. (4). For example, 1 P72/1/10/16
=
1 P72/1
+
1 P10/16
(9)
where P72/1 and P72/1/10/16 are experimental permeabilities for the samples 72/1 and 72/1/10/16, respectively, and P10/16 , the permeability for the active layer of the samples 72/1/10/16. A compilation of these
H.de L. Lira, R. Paterson / Journal of Membrane Science 206 (2002) 375–387
383
Fig. 3. Gas flux vs. transmembrane pressure with different gases for the sample 72/1/10/16.
calculations are shown in Table 5. It can be observed, by comparison with Table 4, that the addition of the thin active layer to 72/1 giving 72/1/10/16 has decreased the permeability by approximately 50%. For samples 72/1/5/20 and 72/1/5/40, the reductions are approximately 90 and 96%, respectively. (For practical, commercial membranes the active layers would be much thinner.) Here they are made thick enough to allow more accurate estimates of their thicknesses from SEM micrographs, Plate 5.
For membranes 72/1/5/20 and 72/1/5/40, the active layers should have the same pore size and pore density, but, since the anodisation at 5 V is doubled for the 5/40 should be twice as thick as that of the 5/20, Eq. (3), and so have a permeability half as large. This is not so (although it is smaller) and cannot be explained easily, except possibly by an unexpectedly important contribution of the intermediate layer created when the voltage is reduced progressively from 72 to 5 V during the anodisation. Overall the effect should be
384
H.de L. Lira, R. Paterson / Journal of Membrane Science 206 (2002) 375–387
Table 4 Experimental gas permeabilities (mol s−1 m−2 Pa−1 ) × 107 , for anodic alumina membranes, including support plus active layer (multilayer) Gases
H2 He Ne N2 O2 Ar CO2 C3 H8
Membranes 72/1
60/3
40/5
72/1/10/16
72/1/5/20
72/1/5/40
592.30 463.50 215.60 200.80 176.10 158.10 138.70 159.40
314.30 223.20 108.50 86.30 79.30 74.20 72.50 72.70
394.10 287.90 133.70 110.00 94.60 84.80 76.90 82.20
207.00 154.90 54.80 64.80 66.60 51.70 45.00 49.50
70.50 54.60 25.30 23.50 22.60 21.40 16.40 19.80
23.75 17.47 9.54 7.23 7.49 6.86 6.45 8.99
smaller for the membrane with the thicker active layer, 72/1/5/40. For all membranes studied, the permeabilities are inversely proportional to the square root of the molecular weight of the permeant gas, Figs. 4 and 5, proving that the diffusion mechanism is Knudsen flow, Eqs. (3) and (4). Pore diameters for the homogeneous membranes and the active layers of the bilayer membranes were calculated from Eq. (3), and using experimental values for pore densities and membrane/layer thicknesses, of Table 2. Pore densities were obtained from the micrographs even when the pores themselves were small and unresolved, using the assumption that they lay at the centres of the much larger and resolvable cell structure, as discussed in Section 3.1. (The agreement between experimental and hexagonal model pore
densities is not good. Observed SEM pore densities are much lower than the hexagonal model predicts. There is no explanation for this at this time, except it should be stressed that the model may be expected to be less reliable in the limit when extrapolated to the lowest anodisation voltages, close to the polarisation potential.) The results are given in Table 6.
Table 5 Experimental permeability (mol s−1 m−2 Pa−1 ) × 107 , for the active layer of the multilayer membranes Gases
H2 He Ne N2 O2 Ar CO2 C 3 H8 a
Membranes 72/1/10/16a
72/1/5/20a
72/1/5/40a
318.21 232.65 73.48 95.68 107.11 76.82 66.61 71.80
80.02 61.89 28.66 26.61 25.93 24.75 18.60 22.61
24.74 18.15 9.98 7.50 7.82 7.17 6.76 9.53
From Eq. (9) and experimental value of the support 72/1, Table 4
Fig. 4. Permeability of eight test gases plotted against the inverse square root of their molecular weights for the homogeneous membrane sample 72/1.
H.de L. Lira, R. Paterson / Journal of Membrane Science 206 (2002) 375–387
385
Table 7 Comparison of pore size estimated by anodisation voltage, gas permeation and SEM Pore diameter (nm) Membrane sample
Knudsen estimate
72/1 60/3 40/5 72/1/10/16a 72/1/5/20a 72/1/5/40a
59.3 48.7 41.8 13.08 5.86 5.07
a b
Fig. 5. Permeability of eight test gases plotted against the inverse square root of their molecular weights for the bilayer membrane sample 72/1/10/16.
For all eight gases studied there is excellent agreement on estimates of pore diameters, as shown by the low standard deviations on the averaged values. The Knudsen mechanism for diffusion in these membrane
± ± ± ± ± ±
1.4 0.6 0.7 0.68 0.21 0.30
Hexagonal cell model Eq. (8)b
SEM
67.00 55.80 37.20 9.30 4.60 4.60
64 ± 14 50 ± 6 40 ± 8 – – –
Pore size of the active layer of the membrane. D = 2.8 nm V−1 , D/d = 3.0.
is therefore amply validated, as can be seen from comparative data directly from SEM for the larger pore membranes where direct comparisons can be made, Table 7. For the smaller pored membranes the Knudsen technique provides unique experimental estimates of pore size, which is supported by the idealised model calculations based on the hexagonal cell model, Eq. (8). Knudsen techniques can only be applied with confidence for membranes with parallel, uniform pores. It is therefore of limited quantitative application for complex pore structures, due to inhomogeneities and uncertainty in evaluations of pore structure factors, Eq. (1). It has, however, been applied most successfully to track etched membranes by Yamazaki et al. [42].
Table 6 Pore diameter (nm) of active layer of anodic alumina membrane estimated by using Knudsen law, Eq. (3) and pore densities estimated by SEM Gases
Membranes 72/1
60/3
40/5
72/1/10/16a
72/1/5/20a
72/1/5/40a
H2 He Ne N2 O2 Ar CO2 C3 H 8
56.45 58.39 59.16 61.11 59.81 59.88 58.24 61.01
47.87 47.94 49.29 48.30 48.02 48.74 49.14 49.19
41.85 42.31 42.84 42.46 41.29 41.32 40.63 41.55
12.93 13.07 11.64 13.45 14.28 13.26 12.85 13.18
5.53 5.70 5.76 5.95 6.03 6.16 5.69 6.07
4.71 4.77 5.11 4.91 5.09 5.14 5.12 5.74
Average S.D.
59.26 1.45
48.56 0.56
41.78 0.68
13.08 0.68
5.86 0.21
5.07 0.30
a
The pore diameter refer to active surface layer in these asymmetric membranes, underlined.
386
H.de L. Lira, R. Paterson / Journal of Membrane Science 206 (2002) 375–387
4. Conclusions Using an electrical pulse method allowed anodic alumina membranes to be prepared unaffected by the side effects of chemical erosion of pores, normally associated with conventional methods for membrane release from the aluminium anode. As a result membranes with pore diameters as small as 5 nm were successfully produced with clean uneroded pores. The method is precise and simple, using anodic voltages marginally above the polarisation potential of the aluminium anode. Such membranes have a range of potential uses and their regular geometry makes them ideally suited to both practical and research applications. A series of membranes with widely differing pore sizes were prepared by this method. Membranes with smaller pore sizes were prepared supported on thicker larger pored substrates in a bilayer format. Standard SEM measurements were made to determine membrane (or layer) thicknesses and pore densities, but could not resolve pore dimensions for pores with diameters, less than ca. 15 nm. Other methods for measurement of pore diameters were required. Of the two methods used, one was based on a model or idealised concept of the anodic alumina cell structure, the other experimental using gas diffusion permeability measurements. Pore sizes estimated, using an idealised model of the hexagonal cell structure of anodic alumina, were in good agreement with SEM data for large pores where direct comparison was possible. Gas permeabilities for eight test gases were shown to obey Knudsen’s law. From these data estimates of pore diameters of membranes and the active layers of bilayer membranes were made. They too gave excellent agreement with direct SEM measurements for large pore membranes and consequently were used with confidence to estimate pore sizes as small as 5 nm. The gas diffusion technique can only be used with confidence for membranes with regular parallel pore structure, but has been used successfully for track etched membranes (TEMs), [42].
Acknowledgements We are grateful for a scholarship from CNPq (Brazil) to Helio De L. Lira (ref. Proc. 20.1780/92.5)
during the period of this research in Glasgow University, 1994–1997. References [1] P. Mardilovich, A.N. Govyadinov, N.I. Mukhurov, A.M. Rzhevskii, R. Paterson, New and modified anodic alumina membranes. Part I. Thermotreatment of anodic alumina membranes, J. Membr. Sci. 98 (1995) 131. [2] P.P. Mardilovich, A.N. Govyadinov, N.I. Mazurenko, R. Paterson, New and modified anodic alumina membranes. Part II. Comparison of solubility of amorphous (normal) and polycrystalline anodic alumina membranes, J. Membr. Sci. 98 (1995) 143. [3] J. Randon, P.P. Mardilovich, A.N. Govyadinov, R. Paterson, Computer simulation of inorganic membrane morphology. Part III. Anodic alumina films and membranes, J. Colloid Interface Sci. 169 (1995) 335. [4] P.P. Mardilovich, A.N. Govyadinov, V. Kozharinov, R. Paterson, Polycrystalline anodic alumina—new ceramic membranes and matrices, in: P. Vincenzini (Ed.), Ceramics: Charting the Future, Techna. Srl., 1995, p. 2763. [5] H. de L. Lira, Ph.D. Thesis, Preparation and Properties of Ceramic and Surface Modified Ceramic Membranes, Glasgow University, 1996. [6] P.P. Mardilovich, A.N. Govyadinov, J. Randon, S. McFadzean, R. Paterson, New high temperature chemically resistant anodic alumina membranes: preparation, modification, properties, in: Proceedings of the 3rd International Conference on Inorganic Membranes, Worchester, MA, USA, 1994, 457. [7] A.N. Govyadinov, P.P. Mardilovich, V.V. Kozharinov, V.M. Gryaznov, R. Paterson, Composite systems based on anodic alumina membranes, in: Proceedings of the 3rd International Conference on Inorganic Membranes, Worchester, MA, USA, 1994, 541. [8] J. Randon, P.P. Mardilovich, A.N. Govyadinov, R. Paterson, Modelling the pore structure of anodic alumina membranes, in: Proceedings of the 3rd International Conference on Inorganic Membranes, Worchester, MA, USA, 1994, 613. [9] T.P. Hoar, N.F. Mott, Mechanism for the formation of porous anodic oxide films on aluminium, J. Phys. Chem. Solids 9 (1959) 97–99. [10] M. Mårtensson, K.E. Holmberg, C. Lofman, E.I. Eriksson, Some type of membranes for isotope separation by gaseous diffusion, in: Proceedings of the 2nd UN International Conference on Peaceful Uses of Atomic Energy, Vol. 4, 1958, pp. 395–404. [11] A.W. Smith, Process for producing an anodic aluminum oxide membrane, US Patent 3,850,762 (1974). [12] A.W. Smith, Porous anodic aluminum oxide membrane, J. Electrochem. Soc. 120 (8) (1973) 1068–1069. [13] G.E. Thompson, R.C. Furneaux, G.C. Wood, H.A. Richardson, J.S. Goode, Nucleation and growth of porous anodic films on aluminum, Nature 272 (1978) 433–435. [14] R.C. Furneaux, W.R. Rigby, A.P. Davidson, Porous films and method of forming them, EP 0,178,831 Al, 1986, US Patent 4,687,551 (1987).
H.de L. Lira, R. Paterson / Journal of Membrane Science 206 (2002) 375–387 [15] R.C. Furneaux, W.R. Rigby, A.P. Davidson, The formation of controlled porosity membranes from anodically oxidised aluminum, Nature 337 (1989) 147. [16] G.E. Thompson, G.C. Wood, Porous anodic film formation on aluminium, Nature 290 (1981) 230. [17] Anotec Separations, Anopore: a new inorganic membrane unique in the world of microfiltration, Brochure, Whatman International Ltd., Maidstone, UK. [18] R.C. Furneaux, R.W. Philpott, D.M. Jenkins, Controlled pore size membranes produced by anodizing aluminium, in: Proceedings of the First International Conference on Inorganic Membranes, Montpellier, 1989, pp. 47–53. [19] R.C. Furneaux, M.C. Thornton, Porous “ceramic” membranes from anodising aluminium, Adv. Ceram. Chem. Process Eng. 43 (1988) 93. [20] K. Itaya, S. Sugawara, K. Arai, S. Saito, Properties of porous anodic aluminum oxide films as membranes, J. Chem. Eng. Jpn. 17 (1984) 514–520. [21] N. Itoh, K. Kato, T. Tsuji, M. Hongo, Preparation of a tubular anodic aluminum oxide membrane, J. Membr. Sci. 117 (1996) 189–196. [22] T. Parlovic, A. Ignatiev, Optical and microstructural properties of anodically oxidised aluminum, Thin Solid Films 138 (1986) 97–109. [23] J. Rocek, P. Uchytil, Evaluation of selected methods for characterisation of ceramic membranes, J. Membr. Sci. 89 (1994) 119. [24] Y. Yuxun, W. Diyong, Pore size of composite inorganic membrane, in: Proceedings of the third International Conference on Inorganic Membranes, Worchester, July 1994, p. 525. [25] P. Dietz, P.K. Hansma, O. Inacker, H.D. Lehmann, K.H. Hermann, Surface pore structures of micro- and ultrafiltration membranes imaged with the atomic force microscope, J. Membr. Sci. 65 (1992) 101. [26] P. Neufeld, N.K. Nagpaul, R. Ashdown, M. Akbar, Crystallisation of anodic Al2 O3 , Electrochim. Acta 17 (1972) 1543–1546. [27] P. Uchytil, Gas permeation in ceramic membranes. Part I. Theory and testing of ceramic membranes, J. Membr. Sci. 97 (1994) 139. [28] P. Uchytil, Z. Broz, Gas permeation in ceramic membranes. Part II. Modelling of gas permeation through ceramic membrane with one supported layer, J. Membr. Sci. 97 (1994) 145. [29] F.W. Altena, H.A.M. Knoef, H. Heskamp, D. Bargeman, C.A. Smolders, Some comments on the applicability of gas
[30]
[31]
[32]
[33]
[34]
[35] [36]
[37]
[38]
[39]
[40]
[41] [42]
387
permeation methods to characterise porous membranes based on improved experimental accuracy and data handling, J. Membr. Sci. 12 (1983) 313. P. Uchytil, Z. Wagner, J. Rocek, Z. Broz, Possibility of pore size determination in separation layer of ceramic membrane using permeation method, J. Membr. Sci. 103 (1995) 151. P. Uchytil, X.Q. Nguyen, Z. Broz, Characterisation of membrane skin defects by gas permeation method, J. Membr. Sci. 73 (1992) 47. K. Keizer, R.J.R. Uhlhorn, R.J. Van Vuren, A.J. Burggraaf, Gas separation mechanisms in microporous modified ␥-Al2 O3 membranes, J. Membr. Sci. 39 (1988) 285. R.J.R. Uhlhorn, K. Keizer, A.J. Burggraaf, Gas and surface diffusion in modified ␥-alumina systems, J. Membr. Sci. 46 (1989) 225. R.J.R. Uhlhorn, K. Keizer, A.J. Burggraaf, Gas transport and separation with ceramic membranes. Part I. Multilayer diffusion and capillary condensation, J. Membr. Sci. 66 (1992) 259. P. Levitz, Knudsen diffusion and excitation transfer in random porous media, J. Phys. Chem. 97 (1993) 3818. R.L. Rowell, S.A. Carrano, A.J. Bethune, A.P. Malinauskas, Gas and vapour permeability: surface flow through porous media, J. Colloid Interface Sci. 37 (1971) 242. R.W. Schofield, A.G. Fane, C.J.D. Fell, Gas and vapour transport through microporous membranes. Part I. Knudsen–Poiseuille transition, J. Membr. Sci. 53 (1990) 159. R. Datta, S. Dechapanichkul, J.S. Kim, L.Y. Fang, Uehara, A generalised model for the transport of gases in porous, non-porous, and leaky membranes. Part I. Application to single gases, J. Membr. Sci., 75 (1992) 245. M. Bhandarkar, A.B. Shelekhin, A.G. Dixon, Y.H. Ma, Adsorption, permeation, and diffusion of gases in microporous membranes. Part I. Adsorption of gases on microporous glass membranes, J. Membr. Sci. 75 (1992) 221. K. Keizer, R.J.R. Uhlhorn, V.T. Zaspalis, A.J. Burggraaf, Transport and related (gas and vapour) separation in ceramic membranes, Key Eng. Mater. 61/62 (1991) 143. M. Knudsen, Kinetic Theory of Gases: Some Modern Aspects, 3rd Edition, Methuen & Co. Ltd., 1950. M. Yamazaki, L.P. Geraldo, R. Paterson, Characterisation of Polycarbonate Nuclear Track Etched Membranes by Means of a Gas Permeation Method, Nuclear Instruments and Methods in Physics Research, Vol. 418A, 1998, pp. 491–496.