This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
is defined by <M> -~ = Xl' Mi ~ + (1 - Xl)" M2 ~
(9.25)
then the Knudsen permeation of the mixture is obtained by inserting <M> -~ from (9.25) for (M) -~ from (9.7). Equation (9.24b) in combination with (9.6) predicts for a non-isobaric and equimolecular mixture (xl = x2 = 0.5) that the ratio Jk,1/Jk,2 is proportional to sqrtM2/M1. This is the ideal permselectivity of the mixture. 9.2.4.3 Viscous Flow and the Transition Region
Viscousflow The viscous flow of a binary mixture which is fully in the continuum regime does not affect the concentration of both gases and relation (9.2) applies for the mixture as for a single gas with the mixture viscosity rl(x) of the mixture with a constant mole fraction x. The viscous flow Jv~ of each species i of the mixture equals the total flow Jv,t multiplied by the mole fraction x (proportional to the partial pressure p): Jv,i = xi. Jv,t and Jv~ given by Eq. (9.2).
(9.26)
358
9 ~ TRANSPORT A N D SEPARATION PROPERTIES OF MEMBRANES WITH GASES A N D VAPOURS
The transition region of Knudsen and continuum diffusion or viscous]low Two important cases must be considered: (i) non-isobaric, and (ii) isobaric situations. The non-isobaric situation will first be discussed. - Estimate of magnitude of different contributions: According to Eq. (9.9a) the viscous flow increases with r 2 and with P, while the Knudsen diffusion increases with r and is independent of pressure. This means that the contribution of the viscous flow to the total permeation increases with r and p. Using relations (9.3) and (9.4b) or (9.7) it can easily be shown that in a first approximation the total permeation F can be written as: F=Fk.
(
3F'P/ l+l--6.rl.v
(1-~) =Fk" 1+ .A
(9.27a)
Using the gas kinetic relations between r, rl, v and ;~ we find A = n K~ and so:
(
f =f k 91 + ~ 16 9K n
/
(9.27b)
Equation (9.27b) is useful to estimate the contribution of viscous flow to the total permeation. For argon at 1 bar and 293 K it is found that with r = 10 n m (K n = 7), 98% is Knudsen diffusion, with r = 1 ~tm (Kn = 0.07), 67% is viscous flow and 33% is Knudsen diffusion. So with larger pores and higher pressure in non-isobaric systems viscous flow is the dominant contribution and molecular diffusion can be assumed to be negligible. Note that in this treatment m o m e n t u m transfer is ignored. - The extended Fick model: An extended Fick type of equation is used by Veldsink [46] to incorporate this m o m e n t u m transfer. The total flux Ji of component i can be written as a superposition of the total pressure driven viscous flow on the diffusional flow component. 5(xiP) Bo 1 De ~ + ~ x ~ P J i - - RT 5z 11
5P)
(9.28)
where Bo is the permeability coefficient, xiP the partial pressure and D e is the effective diffusion coefficient of i in the mixture. The term 'effective' indicates that geometric effects of the pore structure are incorporated in D e and Bo (with D e = TI/'cD ~ with D ~ the expression for a cylindrical pore). In the transition region the transport resistances are assumed to be in series as expressed by the Bosanquet equation: 1
1
~ = ~ +
Die,j
~
1
Diem Diek
(9.29)
9 m TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
359
where D ei,m and D ei,k a r e the effective diffusion coefficients for continuum and Knudsen diffusion respectively. Di,k is given by the gas kinetic expression. So it follows:
4 ,N/8RT e D i,k -----T, -3 Kn ~M i
(9.30)
If the mixture continuum diffusivity is unknown it can be estimated using Blanc's law: tl
De z,m
= ~1 1
--
y__, Died 9xj X i
(9.31a)
j=l, j~l
For binary diffusion (9.31a) reduces to d In p _ Dll (Fick) Diem = D12 d In c i
(9.31b)
Dij (here D12) is the diffusion coefficient of the pair i-j. It can be experimentally measured by Wicke-Callenbach type (isobaric) measurements (see Sections 9.2.4.3 and 9.4.2,3) or calculated with the help of the first order approximation Chapman-Enskog relation [1,4] which is written as
.N/,r3 D12-
0.00262
L
+__M___2.1
M 2 P (~12 ~'~12
j
(9.32)
where oh2 is the collision diameter (taken as the arithmetical mean of the individual component diameters), ~'~12 is the first order collision integral, which is tabulated by e.g. Hirschfelder [4] and which is a function of the temperature. P is the pressure in atm and D12 is obtained in c m 2 s -1.
The Dusty Gas Model (DGM) In the DGM model as presented by Mason and Malinauskas [11a] all the different contributions to the transport are taken into account. The wall of the porous medium is considered as a very heavy component and so contributes to the momentum transfer. The model is schematically represented in Fig. 9.12 for a binary mixture (in analogy with an electrical network). As can be seen from this figure, the flux contributions by Knudsen diffusion Jk,i and of molecular (continuum) diffusion of the mixture Jm,12are in series and so are coupled. The total flux of component i (i = 1,2) due to these contributions is Ji, km" Note that Jk,i = Jm,12. The contribution of the viscous flow Jv,i and of the surface diffusion Js,i are parallel with Ji, km and so are considered independent of each other (no coupling terms, e.g. no transport interaction between gas phase and surface diffusion).
360
9 -- TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANESWITH GASES AND VAPOURS
L
P Jr
I
L_
J v,i
J,.,
J m,1I2 =''"
,i. Fig. 9.12. Schematic representation of the Dusty Gas Model. Ji is the molar flux of component i; k = knudsen, m = molecular diffusion, v = viscous flow, s = surface diffusion.
The flux expression for a single species i in a multi-component mixture with n components according to the DGM model results in 1"/
~_~ xi ' Jj- x.j. ' Ji Ji 1 ~)xi xi l BoP ) SP j=l, j,i Pi Dfj - p . Di,---~k- RT 8z + PRT ('11 Di,--~k+ 1_ ~
(9.33)
with Diek- g/T,.Di~ given by (9.30) and D,~ = 8/t.D~ given by (9.32)or directly measured. B0 is the permeability coefficient for a porous medium (m 2) and it can be obtained from the slope of the curve obtained by plotting the permeation F (in the transition region) versus the average pressure, as discussed in Section 9.2.3.2. For multi-component mixtures the flux ]i as described by (9.33) can only be obtained in implicit form. For binary mixtures (9.33) can be solved directly in explicit form.
- Comparison of DGM and extended Fick models, some data: A comparison of DGM and the extended Fick model for the transition region has been made by Veldsink et al. [46] and is illustrated by many transport data and applied to describe transport in a macro-porous membrane reactor. Their main conclusion is that for ternary mixtures the use of the DGM model is necessary and predicts the transport of a gas mixture within a few percent (5%). For binary gases usually the extended Fick model is sufficient, but with an overall pressure over the membrane the accuracy is less than that obtained by use of the DGM. A further discussion will be given in Section 9.7.
9 ~ TRANSPORT AND SEPARATION PROPERTIESOF MEMBRANES WITH GASES AND VAPOURS
361
TABLE 9.3 Relative importance of molecular (continuum), Knudsen diffusion and Poiseuille flow for air at 20~ in a straight cylindrical pore (after Karger and Rutven [3]) Dpoiseuille
p (atm)
Dm (cm2/s)
r (cm)
Dk (cm2/s)
D (cm2/s)
DPoisuille (cm2/s)
Dtotal (cm2/s)
1.0
0.2
10-6 10-5 10-4
0.03 0.3 3.0
0.027 0.121 0.19
0.0007 0.07 7.0
0.027 0.19 7.2
10 -6
0.03
0.012
0.007
0.019
0.37
10
0.02
10 -5
0.3
0.019
0.7
0.719
0.97
10-4
3.0
0.020
70
70
1.0
Dtotal
0.026 0.37 0.97
The relative importance of different transport contributions in a porous structure is given in Table 9.3 which shows that the contribution of Poiseulle (viscous) flow becomes important in larger pores (range 0.1-0.3 ~tm). At high pressure (10 bar) the Poiseuille flow is already important in pores with a radius of 10 nm. - The extended P - D model:
Present and De Bethune [48] were the first to develop a model (P-D model) including diffusion, intermolecular m o m e n t u m transfer and viscous flow. Based on the P-D model, Eickmann and Werner [18] incorporated two parameters (n k and [5) i n t h e P-D equations to account for geometric and reflection characteristics of a real membrane. This extended P-D model is very successful to describe the effect of a variety of parameters on permeation and separation [18] and will also be used in Section 9.3. Note that surface diffusion is not incorporated in the model. The flux of component i in a binary mixture is given by: Ji
g'[ = L
~176 d ( x . P ) ~f~ dP , dP] 1 + B'--------P d-----~-+ 1 + B ' P d z + x A P--~z
(9.34)
with the mol fractions for components 1 and 2 (i = 1 or 2) given by x and l-x, respectively. The terms in (9.34) describe the Knudsen diffusion (1st term), m o m e n t u m transfer (2nd term) and viscous flow, respectively. The different coefficients in (9.34) are described below:
8r[ g = -~
~ 2kTM
(9.34a)
362
9 - - TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
g ' = nkr I r 2 ~g
(9.34b)
Mathematically nk[-] accounts for the porosity (r and the tortuosity (z) for gas permeation dominated by Knudsen diffusion (see Eq. (9.16)). ~[-] is used to correct for behaviour deviating from the ideal Knudsen behaviour, e.g., due to reflection conditions deviating from elastic specular collisions with the pore wall. 3r
A=~ 16rlv2
(9.34c)
and
A
A' = --
(9.34d)
B=8r qrckT q Mx 1 3---ff"- - -2--M " M1 + M2 PD12
(9.34e)
B' = B. [3
(9.34f)
= ~ / M 1 + (1 - x ) ~4M 2
MIM2
M+=
(9.34g) (9.34h)
M 1+ M 2
(9.34i) with M2 > M1
o0: - x/f 8kT /
(9.34j)
34k
D12 in (9.34e) can be calculated from (9.32) or directly measured. Equation 9.34 is used by Eichmann and by Wu et al [19] to study separation in porous media and this will be discussed in Section 9.3. Wu et al. [19] used (9.34) for single gas permeation (see Fig. 9.4a,b) to obtain values of n k and ~ in an asymmetric membralox membrane consisting of a top layer of T-A1203 (thickness 3-4 gm, pore radius r = 4(-7.8) nm) supported on an o~-A1203 support.
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
363
Gases studied were He, N 2, H 2 and CO in the temperature range 20-815~ and pressure range 3-38 bar. The pore diameter calculated from the measurements is not the same for all gases. The same holds for the values of n k which vary from 2.43x10 -12 m -2 (H2) to 4.3x10 -16 m -2 (N2). The values of [3 (representing reflection conditions of molecules after colliding with the pore wall) decrease with increasing temperature for all four gases and strongly different values are found for the different gases. Especially the ~ values formed for CO are much lower: 0.27 (20~ and 0.06 (T = 538~ compared with that for N 2 w i t h ~ = 0.40 (20~ and 0.024 (815~ respectively. This unexpected behaviour of CO may be attributed to the interaction of CO with the aluminium oxide surface. The small value of [~ explains the much lower permeation of CO compared with the theoretical Knudsen diffusion in the membrane (for the other gases there is a good agreement) (see Fig. 9.4a,b). It should be noted that surface diffusion of CO is possible, but is probably negligible because the permeation is decreased (with respect to expectations based on Knudsen diffusion) instead of increased (if surface diffusion is important).
- Determination of effective diffusion coefficients: The effective diffusion coefficient, and so the permeation of a component in a mixture, can be determined with the so-called Wicke-Callenbach cell [7]. The cell has a similar design to that given in Fig. 9.11 but in this Wicke-Callenbach type of measurement there is no total pressure difference across the membrane (isobaric). The feed is in this case gas a, the permeate in Fig. 9.11 is replaced by an incoming flow of gas b (countercurrent configuration). Gases a and b diffuse through the membrane (counter diffusion) with fluxes Ja and Jb, and so the retentate (Fig. 9.11) is now a flow of gases a+db, the outgoing stream ('sweep' in Fig. 9.11) is b+da. In the measurement of D a, the volume flow ~v,d of the gas mixture b+da in the bottom compartment (d) and the concentration Ca,e in Qba,d are measured; this gives the mol fraction Ja,e. In the equilibrium state using a mass balance over the cell and using the DGM expression for a binary gas (under isobaric conditions) it can be described that [49] 2r
P d ' Ya,d " Tcell
" a - "- (I)v'd Pcell 9Ya,cell " T d = D e ' a
1 - 2ya, d Ya, cell
AP + K e ' a Pcell
(9.35)
This equation takes into account that usually P, y and T are measured not in the cell but at a different site in the measuring equipment. A plot of the left-hand side of Eq. (9.35) v e r s u s AP/Pcell yields the effective flow factor Ke~ from the slope of the curve. The value of De,a c a n be calculated from the intersection of the curve at AP/Pcell = 0 because the mol fraction Ya,d is known.
364
9 - - TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
9.3 S E P A R A T I O N OF BINARY MIXTURES IN SIMPLE M E S O P O R O U S MEMBRANES
9.3.1 Important Concepts The separation of gas mixtures in practice can be performed in a variety of modes e.g. counter- or concurrent flows and cross flow (dead end mode) with different conditions concerning the variation of pressure and concentration on the feed and permeate chambers and along the membrane surface. Examples are discussed by e.g. Eichmann and Wemer [18]. The most simple experimental set-up, suitable to define some important parameters, is given in Fig. 9.11. Assumptions made here are well mixed flow on both permeate and retentate streams (this means constant concentration), no pressure drop throughout the permeate and retentate sides respectively and ideal gas behaviour. These assumptions hold usually in small modules with not too large membrane permeation. In large modules (long tubes, capillaries or large plates) with high membrane fluxes other conditions prevail. This will be discussed in Section 9.5. The parameter to describe the separation efficiency for a binary mixture is the separation factor (xwhich is a measure of the enrichment of a gas component after it has passed the membrane. (x=
y .1-x 1-y x
(9.36)
with x and y the mol fractions of feed and permeate respectively. For a given mixture, 0~is influenced by the membrane and the process specific parameters. In mesoporous membranes the most effective separation mechanism outside the capillary condensation region is Knudsen diffusion. In this case the ideal separation factor 0~*equals the square root of the ratio of masses: 0~* = ~ M 2 / M 1 with
M 2 >
M1
(9.37)
In general c~* is not equal to ~ due to back diffusion, caused by non-zero pressure at the permeate side, or to contributions of non-separative mechanisms to the total flow and concentration polarisation on feed or the permeate side. Also the presence of surface diffusion influences the ideal separation factor. Back diffusion due to a non-zero value of the pressure at the permeate side is a very general phenomenon to decrease the value of (x. The permeant gases at the permeate side of the membrane are removed by pumping or by a sweep gas. In the last case the total pressure is usually relatively large, but the partial pressure of the permeant is low. Using a sweep gas makes the mixture effectively a ternary system and ignoring the effect of the sweep gas (as is frequently done) is not always allowable as will be discussed in Sections 9.4 and 9.5.
9 - - TRANSPORT AND SEPARATION PROPERTIESOF MEMBRANES WITH GASES AND VAPOURS
365
If the pressure at the downstream (permeate) side is in the transition or continuum regime and is not negligible, there is a back-diffusional flux into the membrane decreasing the value of c~. Equation 9.38 gives the effect of back diffusion on the actual separation factor [23,24]. a = 1+
(1 - Pr) (a* - 1) 1 + Pr(1 - y )
(9.38)
(0~*- 1)
where Pr is the ratio of permeate pressure divided by the feed pressure. It is obvious from (9.38) that the permeate pressure directly after the separation layer should be kept low. This is in principle possible with single wall, symmetric membranes. With asymmetric (supported) membranes the support represents always a certain flow resistance and this means that the actual, or partial, pressure of the interface between separation layer and support is larger than the pressure at the permeate side of the support. This implies that the flow resistance of the support should be as small as possible to minimise back-diffusional effects. The separation factor a as determined from gas mixtures is generally not the same as the permselectivity which is defined as the ratio of the permeation of the single gases at a given membrane thickness. They are similar only when all interactions between the different phases and between gases and the pore wall can be neglected, e.g., in the Knudsen region and at high temperature (surface diffusion negligible).
9.3.2. Separation in the Knudsen and Transition Regions As discussed above, the ideal separation factor (x" in the case of pure Knudsen diffusion is given by Eq. (9.37) and is equal to the permselectivity provided that surface diffusion is not present (high temperature). As can be seen from (9.37) the highest ideal separation factors are obtained for mixtures of light and heavy gases. Back-diffusion effects are taken into account by Eq. (9.38) to give the real separation factor. The support can have a considerable influence on the separation factor of the membrane consisting of separation layer and support when its flow resistance is not negligible and the gases in the support pores are in the transition or viscous flow regime [20]. This point will be discussed in Section 9.5. In the transition region intermolecular momentum transfer decreases the separation factor considerably. The effects of the pressure ratio Pr, with feed pressure as a parameter of temperature of pore size and of concentration, are analysed by Wu et al. [18] and by Eichmann and Werner [19]. Wu et al. used Eq. (9.34) to simulate the permeate composition and separation factor for H2/N2, H2/Co and H e / O 2 gas mixtures and compared them with experimental results obtained on a Membralox asymmetric membrane system,
366
9 w TRANSPORT AND SEPARATION PROPERTIESOF MEMBRANES WITH GASES AND VAPOURS
4.00
II
3.S0.
1
3.00
2
~
_ I_
]
II
..
".'~_-~
I I
II
I
II ILl
Ilill
lab
I
I
In
_
;
I
Ill
.
i
-__] . . . .
Ideal
|
I
I
I
I
sepmidon
2.S0 3.O0
8. I . s o m
1.00
0.S0
I
. . . . .
m
m
i
m
m
m
-
m
m
m
m
nomuaaon -
0.( lO
9" " '
I ""'" "
0.20
9 I 0.40
9 i
,
" I
'
0.60
I I I
I
0.80
I'l
,
m
J m
9 9 I 1.00
Pressure Ratio, Pr Fig. 9.13. Feed pressure effect on separation of H2-N2 mixtures at T = 538~ feed H2X0 = 0.5, stage cut = 0.01, p o r e diameter 5.6 run. Feed pressures (1) I atm; (2) 7 atm; and (3) 34 atm. After Wu et al. [19].
whose characteristics are described in Section 9.2.4.3. Correction for the support resistance was not applied. The simulation is generally in good agreement with the experimental results generated for a wide range of operation conditions (20-815~ P = 1-34 atm, P~ = 0.1-0.8, stage cut 0.01-0.36). Deviations between predicted and experimental mole fractions are within 10%, with a consistent overestimate of the light component in the permeate. The effect of the pressure ratio Pr of permeate and feed and feed pressure on the c~ value of a H2/N2 membrane is given in Fig. 9.13 for T = 538~ and feed pressures ranging from 1-34 arm using a separation layer of 5.6 nm. As is shown in Fig. 9.13 for a given pressure ratio, the higher the feed pressure, the lower the separation factor. At all pressures (1-34 atm) the separation factor decreases continuously with P~ (0.10--0.70). At P~ = 0.70 allseparation factors converge to a value of 1.5. Note that even at the lowest pressure (1 atm) and lowest value of P~ = 0.10 the value of (~ = 3.20 which is considerably smaller than the ideal value (~* = 3.70) as given by Eq. (9.37). So even a small amount of non-Knudsen contribution to the total flow in a pore considerably influences the separation. At higher temperature the separation factor increases because the mean free path increases and consequently less momentum loss is expected for H2. The effect is stronger at lower Pr value, and at Pr = 0.10 and P = 7 arm the values of
9 ~ TRANSPORT A N D SEPARATION PROPERTIES OF MEMBRANES WITH GASES A N D VAPOURS
367
for H 2 / N 2 are 2.50 and 3.00 at 20~ and 815~ respectively for the same membrane as used in Fig. 9.13. As the H 2 concentration in the discussed H 2 / N 2 mixture decreases the separation factor also decreases under selected operating conditions. When the partial pressure (concentration) of H2 decreases the number of H 2 to N2 collisions increase relative to that of the H 2 to H 2 collisions and consequently more H 2 momentum is lost at low H 2 concentration and the separation efficiency decreases. The effect is weak however compared to that of pressure and temperature, because collisions with the pore wall are much more frequent compared with intermolecular collisions. Finally the 'stage cut' Sc = Qp/Qb(Qp,f = feed and permeate flow respectively, see Fig. 9.11) is important. At high stage cut the driving force for gas separation in terms of a partial pressure difference is reduced to maintain the material balance. At low Pr the effect of Sc is largest and the lower the value of Sc the larger the separation. For H 2 / N 2 and the conditions given for Fig. 9.13 with Pr = 0.01 and P = 7 atm the values of ~ are 1.90 and 2.90 at Sc = 0.4 and 0.01 respectively (ideal separation is 3.70). The effect of the pressure, temperature and pore radius on the separation factor is investigated also by Eichmann and Werner [19] using Eq .(9.34) with a constant and experimentally determined value of ~ for all gas membrane combinations, in contrast to Wu et al. Who fitted the value of ~ for each gas membrane combination. Figure 9.14 shows the effect of the pressure ratio Pr for different mean pressure levels P (assuming a linear pressure drop in the membrane) on the separation factor of a N 2 / C O 2 mixture (ideal separation factor equals 1.25) in a membrane with pore radius Rp = 0.03 ~tm. In contrast to the situations given in Fig. 9.13, maxima can be seen which shift to larger Pr values with higher pressures. Similar curves are obtained for different pore radii as shown in Fig. 9.15, where the maxima become smaller and shift to larger Pr values with increasing pore radius. The maximum is caused by a viscous flow contribution in the relatively large pores (0.015-0.12 ~tm) considered here. In the rising part of the curves the (non-separative) viscous flow contribution decreases with increasing Pr (smaller pressure difference). The contribution of the viscous flow decreases with decreasing pore radius and with small enough pores the maximum vanishes and continuously decreasing curves are obtained which exhibit greatly reduced pressure dependency. This is shown in Fig. 9.16 with similar shapes to those of Fig. 9.13. The results of Fig. 9.16 are obtained on membranes of y-A1203 with a pore radius of 2.5 nm as prepared by Leenaars and Burggraaf [17b]. In conclusion it can be said that the key operating parameters to approach the ideal Knudsen separation factor (determined by mass ratio) in mesoporous membranes are: small pore diameters; low pressure ratio, adjusted to produce maximum separation; relatively low pressure level; and high process temperature.
368
9 - - TRANSPORT A N D SEPARATION PROPERTIES OF MEMBRANES WITH GASES A N D VAPOURS
1.15 =2x10s
1.10 C 0 0
1.05
Ix10s
1.00 0
0.2 0.4 0.6 0.8
1.0
PR : I~ I p . Fig. 9.14. Influence of pressure ratio Pr on the separation factor of N2/CO2 mixtures. Pore radius is 0.03 l~n. After Eickmannand Wemer [18]. To increase the separation factor above the ideal Knudsen separation factor requires contribution of surface diffusion a n d / o r capillary condensation or the presence of micropore systems.
9.3.3 Separation with Surface Diffusion and Capillary Condensation The permeation of gases in membranes due to surface diffusion and capillary condensation has been discussed in Section 9.2.3.3. together with some illustrative data. The total flux of a single gas is usually calculated as the sum of the flux by surface diffusion and the flux through the gas phase. As shown the surface flux can contribute considerably to the total flux (increased by factor 2-3 of gas diffusional flux), especially with smaller and uniform pore sizes (compare Eqs. (9.9a) and (9.15). With decreasing pore size the flux through the bulk gas decreases while the surface diffusional flux increases. With very small pore diameter (< 2-3 nm) the effective diameter for bulk gas transport is less than the geometric pore diameter due to the thickness of the absorbed layer which
9 - - TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
I
,
9
I
I
I
.
.
.
.
.
.
.
369
.
1.15 i.
0
o
90 015 pm
0
C o
....
1.10
,.mlw
o
0 Q. o ul
0.03
1.0S
1.00
0
(106
0.20.t.
06 (18 1.0
P,,
Fig. 9.15.Influence of pore radius r onthe separation factor of N2/CO2mixture at a pressure of 2 bar. After Eickmann and Wemer [18]. decreases the space available for the gas phase. With gas mixtures this means that the bulk gas phase diffusion of a non absorbing molecule is decreased by absorption of an adsorbing molecular species in the mixture resulting in an increase of the separation factor. This is especially the case with lower temperatures of a few hundred degrees and intermediate pressures which give rise to partial blocking by capillary condensation. Some illustrative examples and special phenomena will be discussed below.
Separation by surface diffusion With gas mixtures, enhancement of the separation factor can be obtained by preferential sorption of mobile species of one of the components of the gas mixture. Adsorption does not always lead to enhanced separation. In a mixture of light non-adsorbing molecules and heavy molecules, the heavy molecules move slower than the lighter ones but in many cases are preferentially adsorbed. Consequently the flux of the heavier molecules is better enhanced by surface diffusion and the separation factor decreases. This occurs, e.g., in CH4/CO 2
370
9 - - TRANSPORT A N D SEPARATION PROPERTIES OF MEMBRANES WITH GASES A N D VAPOURS
I 1.20
6,2,~os./.z a,@ O
u 1.15
0
2
-O,
D 1.10-
0
1
a,10s
1.05 0
0.2 0.~, 06
0.8
1.0
Pn = p, I P, Fig. 9.16. Influence of pressure level on the separation factor of a N2/CO2 mixture. Pore radius is 2.5 nm. After Eickmann and Wemer [18].
mixtures in Vycor glass membranes. With two adsorbing molecular species, competition for the adsorption sites m a y exist and sorption isotherms for single gas species are no longer valid. Uhlhorn et al [28] reported for a H 2 / N 2 mixture a separation factor of about 9 compared to the Knudsen value of 3.74. As shown in Fig. 9.17 the ratio of the H 2 flux over that of the N 2 flux decreases from 9 at a pressure of 50 kPa to 5 at 200 kPa. This result is obtained on ~'-A1203 membranes (thickness 100 ~tm, pore diameter 2.5-4.0 nm) impregnated with 17 wt% (finely dispersed) Ag. The increase of the H 2 flux is obtained by the Ag impregnation. Probably the decrease of the separation factor is caused by a decreasing contribution of the surface diffusion to the total flux with increasing pressure due to saturation of the adsorption. Keizer et al [20] found a similar p h e n o m e n o n for C O 2 / N 2 separation (with C - O ~as 2 the fastest diffusing species) on non-modified ~'-A1203 m e m b r a n e s (0~ = 1.5-2.0 at 240 K, c( = 0.8 (Knudsen value) at 360 K, pressure I bar). In order to
9 - - TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
371
12
Z 9
e~e--.~....e 9l i o r e ~ r A i
~..
0
.
..._
.
.
A
100
ratm
___1.,.~,,,,~...~,~.-,~.,,,~,,,,~
2O0
30O
Fig. 9.17.Experimental (o) and theoretical flux ratio of H2 and N2at 25~ on a nonsupported 7-Al203 layer modified with 17 wt% silver, measured in counter-diffusion configuration. After Uhlhom et al. [28]. enhance the surface contribution the 7-A103 membrane was modified with 2.2 wt% MgO [28,20]. The result was a decrease of the separation factor to 1.0 due to the formation of strongly bonded, immobile CO2 species, the total concentration of adsorbed CO2 remaining constant. As shown by Eq. (9.15) this results indeed in a lower CO2 surface flux.
Separation by multilayer diffusion and capillary condensation (see also Section 9.2.3.3) Brief overviews are given by Keizer et al. [50] and Sperry et al. [39] and these show that very high separation factors in combination with large permeation can be obtained in cases of mixtures of an easily condensable gas (vapour) and a difficult (non)-condensable gas which has a low solubility in the condensed phase. Pore blocking by capillary condensation takes place at 0.5--0.8 of the saturated vapour pressure (depending on pore size) and is preceded by multilayer diffu-
372
9 -- TRANSPORTAND SEPARATIONPROPERTIESOF MEMBRANESWITHGASESAND VAPOURS
A
X a.
0
~ii 5
3o
A
-
,
.
.
.
.
.
.
.
prol:)ytene
.
.
.
.
.~
.
.
.
30
-'
.....
-
B
cleS~:~Dhon
2o .15
"
~ads~ption
Oesor
E C 0
m
0
E
10
s
r.trogen
0 0.00
(xs0
O
t .00
0.00
.
.
.
Rel. prestmm
.
.
.
.
0.50
1.00
R~a.~ m
ql"
r
20
9O
"E 0
E
Q.
.
.
.
'
.
.
D
.
im
o
~ 60
~.
cle~pt~n
J l
rutrooen (~
.
aclsorptton + c~mcxpt|~
to
m
E
'.
p, o o y J ~ / "
c 0
Q
-
C
~
o~oo
iior 30
, O
~L p ~
-
ooo
''
oa0
,~x~
Fig, 9.18. Permeation and separation factors of supported y-alumina thin film for nitrogen and propylene at 263 K (A,B). Propylene is the preferentially permeating component; dashed line gives the relative pressure at which the maximum in the permeation of Fig. 9.18a. occurs. (C) and (D) as (A) and (B) but for a supported film modified with MgO. After Uhlhom et al. [37].
sion and an increased flux of the condensable gas and an increased separation factor. Uhlhom et al. [37] reported separation factors of c~up to 27 for propene/N2 (60:40) mixtures at 263 K (with propene the fastest permeating species). Note that the Knudsen factor is 0.8 and the permselectivity (ratio of single gas phase fluxes) amounts 7.4. As shown in Fig. 9.18, the region with the highest separation factors coincides with the maxima in the permeation curves which in turn are determined by the blocking of pores by adsorbate (capillary condensation). The permeation of propylene at the maximum amounts 30x10 -6 m o l / m 2 s Pa. A further improvement of the separation factor is obtained by modification of the ~/-A1203 membrane with the reservoir method [51]. The membrane pores are filled up to 85% of the pore volume with MgO. This process enhances the value of 0c to 85 with a corresponding decrease of both propylene and N2 permeation values to 15x10 -6 m o l / m 2 s Pa for propene (equivalent to 300 N m 3 / m 2 day bar.
9 ~ T R A N S P O R T A N D S E P A R A T I O N PROPERTIES OF M E M B R A N E S W I T H GASES A N D V A P O U R S
373
The shape of the curves (steeper, shift of desorption branch to lower relative pressures) indicates a narrower pore size distribution with smaller average pore size (below 3 nm) and less defects for the modified membrane. Sperry et al. [39] reported capillary condensation up to 473 K in methanol/H2 mixtures for certain pressure ranges. They used a similar type of membrane as used by Uhlhorn, but treated with NaOH to poison the surface for chemical dehydration reactions. Using the Wicke-Callenbach method (no absolute pressure drop) the highest value of o~ equals 680 (methanol being the faster permeating species) and is obtained at 373 K and 2.2 bar methanol pressure, with a methanol permeability of 51x10 -6 cm3(STP) c m / c m 2 s cmHg. At higher temperatures the maximum obtainable values of both {x and permeability decrease and {x = 110 (with methanol permeability is 4.2x10 -6 m o l / m 2 s Pa) at 473 K. (Note: 1 cmB(STP) c m / c m 2 s cmHg is equivalent to 3.12x10 -6 mol m / m 2 s Pa). Capillary condensation takes place at Pr = 0.60. This is considerably lower than predicted by the Kelvin equation (9.18) for pores with a diameter of 4 nm. Separations with a pressure drop must be carried out with pressure drops smaller than 0.25-0.28 at T < 448 K or 0.05 bar at 473 K due to blow-out of the condensate under these conditions. The observed flow rates in the capillarycondensation regime are larger than those obtained for Knudsen diffusion at lower pressures. Together with the results reported by Sperry e t al., the conclusion is that separation by capillary condensation yields a combination of large separation factor and high permeation even at increased temperature provided the appropriate temperaturepressure, pore size combination is chosen. A disadvantage is the sensitivity of the process for pressure changes (blow-out phenomena). Finally, Asaeda and co-workers [52,53,64] reported separation results using membranes which are modified in such a way that pore sizes below the mesopore range ( 100 for water-light-alcohol mixtures at 70-90~ in alumina-silica membranes. The water permeability is dependent on its concentration in the mixture. At atmospheric pressure and 20% water a typical water permeation value is 1.3X10 -2 m -2 s -1 (= 20 1 H 2 0 (liquid) m -2 day-l). Azeotropic points can be bypassed in this way with an alcohol concentration much higher than the azeotropic concentration. Similar results are given for mixtures of water and organic acids (acetic, propionic, acrylic) by Kitao and Asaeda [52] for rather thick (10 ~tm) silica membranes supported by 7-A1203 and made in a multi step process (up to 15 layers on top of each other). A permeation mechanism and a model for the pore
374
9 ~ TRANSPORT A N D S E P A R A T I O N PROPERTIES OF MEMBRANES W I T H GASES A N D V A P O U R S
structure is proposed by Kitao et al [53]. The pore shape is assumed to be conical, changing from rather wide on the support side to very small at the surface. Here the 'neck' diameter is suggested to be 0.4 nm. Equations for the (preferentially) permeating water flux are given. Near the surface an additional resistance to the flow builds up due to osmotic effects caused by rejection of organic molecules at the pore entrance. 9.4 PERMEATION A N D SEPARATION IN MICROPOROUS MEMBRANES
9.4.1 Introduction and Important Concepts Existing ceramic, mesoporous membranes (with a 4 nm pore diameter) perform most gas separations according to Knudsen diffusion. The obtainable separation factors (Section 9.3.2.) are usually not economical for most gas separations and provide incremental but limited conversion enhancement in catalytic membrane reactor applications. Capillary condensation and preceding surface flow yield economically interesting separation factors but this mechanism is limited to easily condensable gases and is limited to rather low pressure drops due to stability problems (Sections 9.2.3. and 9.3.3.). To enhance the separation factor the average pore diameter should be decreased considerably. According to Eqs. (9.9a) and (9.15) the contribution to the total gas flux of the gas (Knudsen) diffusion decreases and at the same time that of surface flow (diffusion) increases with decreasing pore radius. In recent years modification of existing membranes for improving their separation efficiency has been actively pursued especially by attempts to decrease the pore size of membranes. This resulted in different types of microporous membranes. According to IUPAC convention these are porous systems with a pore diameter below 2 nm. In the literature the name 'microporous' is frequently misused and this should be avoided. An overview of microporous membrane types is given in Table 9.4. The oldest microporous membranes are based on carbon and are reported by Koresh and Softer in a series of papers from 1980 to 1987 (see overviews in Refs. [6,42]). They are made by pyrolysis of a suitable polymer (hollow fibre) as reviewed by Burggraaf and Keizer [9]. More recently Rao and Sircar [42] developed a new technique. A macroporous graphite sheet was coated with a suitable polymer (latex) which was pyrolysed subsequently. This process was repeated 4-5 times and resulted in a total carbon layer thickness of 2.5 ~tm with an average pore diameter between 0.5 and 0.6 nm. The membrane has interesting properties (see Section 9.4.3). Finally, very recently Linkov and Sanderson et al. [55] modified and improved the method reported by Koresh and Softer and produced flat sheets as well as hollow-fibre systems.
9m TRANSPORTANDSEPARATIONPROPERTIESOFMEMBRANESWITHGASESANDVAPOURS
375
TABLE9.4 Microporous membrane types Type .
2. 3. 3.1 3.2 4.
Ref. Carbon hollow fibre, film on (C) support Porous silica glass (Vycor) Amorphous silica based systems Sol-gel techniques C.V.D. Zeolite films on supports (alumina, steel)
42,54,55 56 21,57-63,64 65-68 69-79
Mesoporous glass (Vycor type) can be produced by a combined heat-treatment and leaching procedure [9]. Modification of this process can lead to microporous hollow-fibre systems with interesting properties as discussed by Shelekhin, Ma et al [56]. For further discussion see Sections 9.4.2 to 9.4.4. The most promising results from the viewpoint of a combination of large separation factors and reasonable-to-large flux values are reported for supported silica based systems. Burggraaf and co-workers reported in a series of publications [21,57,63] the sol-gel, two-step synthesis of silica and silica-titania films supported by a composite membrane of mesoporous 7-A1203 and macro-porous (x-al203. The film has a thickness of 50-100 nm and is situated for about 50% within the mesopores of the y-A1203 and for the rest on top of it. The pore diameters are around 0.5 nm. A combination of large separation factors and large fluxes was reported for several gas combinations [60,61] (See sections 9.4.2-9.4.4). As described by de Lange et al., the precursor sol consist of a polymeric silica solution with low fractal dimension [59,62] and the support quality (roughness) is important to obtain defect-free membranes [59,60]. Asaeda et al. [64] produced a microporous film directly in a macroporous c~-A1203 support with a 15-step coating process starting with colloidal silica solution and ending with a polymeric silica solution. This means that a mesoporous intermediate silica layer was first produced. The final top layer was said to have a pore diameter < I nm. The system shows very interesting (isomer) separation properties (see Sections 9.4.3-9.4.4). In a series of papers (1989-1994) Gavalas and co-workers reported the synthesis of silica films in porous Vycor glass substrates with chemical vapour deposition (CVD) techniques [65]. A similar technique was used by Heung et al. [66]. The separation factors reported by Gavalas and by Heung are very high but the fluxes are low. In fact the silica layers are non-porous (no interconnected pore network). Wu et al. [67,68] improved the method used by Gavalas using a
376
9 ~ TRANSPORT A N D SEPARATION PROPERTIES OF MEMBRANES WITH GASES A N D VAPOURS
composite support of 0~-A1203with a 3-5 ~tm thick 7-A1203 in the top region of the 0~-A1203.A 1.5-3 ~tm thick silica film was deposited in the 4 nm pores of the 7-A1203. The smallest obtained effective pore diameter in the silica plugs was estimated to be -- 0.5 nm. A combination of large separation factors and reasonable fluxes was reported for H2/N 2 and H2-isobutane mixtures (see Section 9.4.3). Zeolite membranes form the most recent branch of the inorganic membrane field. It is only very recently that well characterised and properly described real microporous zeolite membranes have been reported [69,72-78,88,89]. Geus et al. [69,70] and Bakker et al. [70] described the synthesis of 50 ~tm thick silicalite (MFI) membranes on porous stainless steel supports; Vroon et al. synthesised 3 ~tm thick silicalite membranes on o~-A1203supports [72-74]. These membranes consist of very small crystals (100-200 nm). Jia and Noble and co-workers et al. reported a 10 ~tm thick silicalite membrane on a composite support of c~-A1203 [27,77]. Finally, Xiang and Ma [76] partially filled the pores of a microporous (~-alumina support with ZSM5 crystals. All the authors used an in situ hydrothermal crystallisation method to grow directly polycrystalline zeolite layers. The layers reported by Jia et al. and by Xiang and Ma contain a relatively large number of defects, in contrast to that of Geus/Bakker and Vroon, but nevertheless show interesting separation and flux properties provided that good condensable gases are present (e.g. methanol, xylenes). The microstructure of the layers plays an important role as shown by Vroon et al. [72,74] as well as does the support (compare clay with stainless steel) as shown by Geus et al. [69,75]. Examples of properties will be discussed in Section 9.4.3. Zeolite membranes on porous support with good to reasonable quality has been reported so far only for silicalite and (related) ZSM5 systems. In the literature since 1985 a number of other systems are reported including a series of patents. They are reviewed by Geus [69] and Vroon [72] and briefly by Matsukata et al. [78] and Burggraaf [79]. This older literature concerns either membrane systems which are not real (but very defective) membrane systems but sometimes have interesting properties for membrane reactors or concern single crystal work or very fragile non-supported membranes on which important fundamental studies have been performed. In the first category belong the pioneering work of Suzuki (patents 1985, 1987) and of I.M. Lachmann (patent 1989) yielding N a A / C a A and X or Y or mordenite zeolites. Unsupported ZSM5 layers were prepared by Haag and Tsikoyannis (1992) and Sano (1991/1992). Work on single crystals of NaX and silicalite were reported by Wernick and Osterhuber (1985) and Geus [69] respectively. For literature references see cited overview papers. It is not the place here to treat structural characteristics of zeolites. Nevertheless a very brief summary with a focus on silicalite/ZSM5 systems is necessary
9 m TRANSPORT A N D SEPARATION PROPERTIES OF MEMBRANES WITH GASES A N D VAPOURS
377
as a background for permeation/separation studies in Section 9.4.3. For details the reader is referred to books of e.g. Breck [80], Meier [81] and van Bekkum et al. [82]. Zeolites can be represented by the empirical formula [80]
M 2/nO"al 203"xSi O2"yH 20 in which n is the cation (M) valence, x/2 is the Si/A1 ratio (equal to or larger than two). The cations M are present to balance the negative charge introduced in the crystalline framework by the substitution of Si 4+ by A13+. These cations can be exchanged (in exchange reactions). The aluminium-rich zeolites are hydrophilic (high affinity for water), the silica-rich zeolites are hydrophobic (small affinity for water) a n d / o r organophilic. Also the thermal stability increases with increasing Si/A1 ratio. The crystalline framework consists of a three-dimensional network of SiO 4 and A104 tetrahedra, linked to eachother by sharing the oxygen atoms. The framework structures contain channels of voids interconnected by ring openings. These channels can be isolated from each other (one-dimensional) or are interconnected by ring openings and form two or three-dimensional network structures. More then 85 different framework structures are known [81]. Silica-rich zeolites are ZSM5 with a Si/A1 ratio of 11/1000 and silicalite (Si/A1 > 1000). Both have a similar structure (i.e. MFI type) but ZSM5 contains some cations and is more hydrophilic. The structure of MFI-type zeolites is given in Fig. 9.19. The structure has two sets of intersecting channels (10-membered oxygen rings, see Fig. 9.19b), one set consisting of straight channels with ring openings of 0.52x0.57 nm, the other set consists of sinusoidal channels of 0.53x0.56 nm (Fig. 9.19a). At the intersection points cavities are formed with a size of about 0.9 nm. The lattice of ZSM5 is stable up to 1175 K; that of silicalite to a somewhat higher temperature. Both zeolites have a good stability in strongly acidic environments, are relatively easy to prepare and have a low affinity for water, which is important for (gas) separation properties. In recent years zeolites with very large pores (supercages) and ring openings up to 0.6xl.32 nm (cloverite) have been synthesised.
9.4.2 Phenomenological Description of Single Gas Permeation The theory of transport in microporous solids is complex and involves many aspects and steps. Although many aspects has been treated separately (e.g., sorption, diffusion, simulation studies, mechanisms, etc.) there are no coherent descriptions of permeation and separation in microporous membranes covering all the important aspects. In this chapter an attempt is made to introduce such a description. It is useful to give a qualitative picture first (Section 9.4.2.1).
378
9 ~ TRANSPORT A N D SEPARATION PROPERTIES OF MEMBRANES W I T H GASES A N D VAPOURS
(.) (b) Fig. 9.19. Schematic picture of zeolite MFI structure. This will show that a quantitative description involving all the complexities in simple microporous membranes is not available (if possible). However a number of boundary cases can be described quantitatively, as in Section 9.4.2.2, and trends in more complex situations can be predicted in combination with the qualitative pictures based on mechanistic considerations.
9.4.2.1 Qualitative Description of Gas Permeation As discussed in Section 9.4.1, the contribution of Knudsen diffusion to the total flux decreases with decreasing pore radius of the membrane material. Initially the selectivity of binary mixtures of gases is constant and equal to the Knudsen value. Lin et al. [67] reported in the region between pore diameters of 3.0-2.0 nm small negative deviations for H e / N 2 mixtures, but with pore diameters < 2.0 nm a strong increase occurs to values above the Knudsen value. This is a typical phenomenon for microporous systems together with the onset of activated gas permeation. As will be shown, it is useful to distinguish microporous membranes in systems with relatively large, intermediate and small pores. This is discussed by de Lange and Burggraaf et al. [59,63] and is schematically shown in Fig. 9.20. Note that here the location of the minima and the shape of the potential as a function of z is given schematically and is not exact. Simulation results yielding pictures as given for region c2 are reported by Petropoulos and Petrou [83]. For mesopores the minimum in the potential curves is equal to the (isosteric) adsorption heat at 'free' surfaces with respect
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
deft
, B
,........,
1!1 a
r
379
nt bl
b2
- si c1
'i c2
2-
0
Z
A
II A A > B, A >> B A>B A>B
w and S are weakly or strongly adsorbing components respectively of mixtures: W-W, etc. SE is size exclusion. *B is large molecule.
flow). The best w a y to p e r f o r m this is d e t e r m i n a t i o n of p e r m e a t i o n values and ~ separation factors in binary mixtures of gases consisting of small, w e a k l y a d s o r b i n g and very large molecules at high temperature. This is the size exclusion regime u n d e r exclusion of strong a d s o r p t i o n on the external surface of the larger molecule.
9.4.2.2 Quantitative Description of Gas Permeation and Separation Single gas permeation The equations given below are derived for single wall or u n s u p p o r t e d m e m b r a n e s u n d e r similar conditions to those given in Section 9.2.4 a n d Fig. 9.11. These are h o m o g e n e o u s and u n i f o r m concentrations (well mixed) a n d pressures on the feed and p e r m e a t e sides of the m e m b r a n e and near equilibr i u m b e t w e e n concentrations in the bulk gas phase and in the m e m b r a n e surface. As discussed in Section 9.4.2.1, small and large micropores s h o u l d be distinguished. This treatment will start w i t h a general description w h i c h is applied to small micropores. Subsequently the consequences for larger micropores will be treated. U n d e r isothermal conditions it follows from irreversible t h e r m o d y n a m i c s [1-3] for the flux Ji in a mixture of k components:
Ji = - ~_~ Lik V ~tk k
(9.39)
9 n TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
383
_
So the real driving force is the sum of the gradients of the chemical potentials as is also implicit in the general Maxwell-Stefan formulation [87-89]. For a single gas this reduces to: 3 In P dqi
J i - - ~tDo,i 0 ha qi dz
(9.40a)
Here the term 31nP/31nq is the so-called thermodynamic factor (hereafter called F), Do(q) is the corrected or intrinsic diffusion constant and Ix is a correction term (see notes below). With q = qsat'0, Eq. (9.40a) becomes:
Ji--
~tqsat, i D 0 ( 0 ) F
d0i d---~-
(9.40b)
Notes on Eqs. (9.40a,b): (1) When qsat~ is expressed in m o l / k g the density ( k g / m 3) enters the nominator of Eqs. (9.40a,b) and ~t = p. (2) If the zeolite is supported with a support having porosity ~, the effective surface area of the zeolite available for transport is ~.m2/m 2 and the term ~ enters the nominator of Eqs. (9.40a,b). If the flux is measured on a supported system and one wants to calculate the intrinsic zeolite properties, ~ enters the denominator of Eqs. (9.40a,b). (3) The term Do(q).blnP/blnq is identical to the Fick diffusion coefficient DF, while the intrinsic diffusion coefficient Do(q) is identical to the Maxwell-Stefan diffusion coefficient Dms. The thermodynamic factor F corrects for differences in activities (chemical potentials) of different gases which can exist with similar concentration gradients. It is similar to the factor that has been described in solid state diffusion by Darken and is sometimes named after him. Equations (9.40a,b) can be integrated over the thickness L of the membrane to yield expressions for the flux of specimen i: qp
Ji dz = - ~ , D 0 (q) F dq
(9.41)
qf
with q - qf (feed) at z = 0; q = qp (permeate) at z = 1. Note that qp and qf are steady-state concentrations which are not necessarily equal to the equilibrium concentrations. Equation (9.41) can be integrated under a number of different boundary conditions which will be treated below.
The Langmuir and Henry adsorption regions In many cases single gas adsorption in zeolites can be adequately describe d by a Langmuir-type adsorption isotherm as given in Section 9.2.2.3.:
KiP I+KiP
O~= ~
(9.16b)
384
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
with Ki the equilibrium Langmuir adsorption constant (Pa-1). Inserting (9.16b) in the expression for F yields the equation for the thermodynamic factor in the case of Langmuir adsorption: F-
1 1-0
(9.42a)
At higher values of 0 small deviations from the Langmuir isotherm are corrected in (9.42a) by introduction of an empirical constant k [86b]" F-
k 1-0
(6.42b)
Substitution of Eqs. (9.16b) and (9.42a) in (9.41) and integration, assuming D is not dependent on q, yields an explicit equation for the single gas flux in the Langmuir regime in terms of sorption and diffusion parameters: Do, i" qsat,i In
L
Ji-
qsat, i - qf, i qsat, i - qp, i
_ ~ D o , i qsat, i In 1 + KPf, i
L
(9.43a)
(9.43b)
1 + KPp, i
Equation (9.43a) can easily be converted in terms of occupancies by dividing numerator and denominator of the In term by qsatd. Note again that Do,i is the intrinsic diffusion coefficient and that DFick = D0/(1 - 0 ) and so DF increases strongly when e assumes larger values. At low occupancy we are in the Henry regime and Eqs. (9.16b) and (9.43b) can be simplified because KiP < 1: Ji = ~ D ~
" qsat'i " K
L
(Pf, i - Pp, i)
(9.44)
The temperature dependency of Ji can be introduced using a van 't Hoff-type relation for K and an Arrhenius relation for D:
i: 0iexp Do, i = D~,i exp -
(9.46)
where Ed is the activation energy for diffusion in the micropores. Insertion of (9.45) and (9.46) in (9.44) yields the temperature dependency of the flux in the Henry regime of a supported zeolite (~t ~ 1):
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
Ji = -~ D~,i Ko, i qsat,i exp
(Ed, i - Qa, i) (Pf- Pp) RT
385 (9.47)
Here D~,i is the pre-exponential coefficient of the intrinsic diffusion coefficient. Note that K'qsat is the Henry constant b as given in Eq. (9.16a) (mol kg -1Pa -1 or mol m -3 Pa -1) which can be directly determined from experiments without separate knowledge of the value of qsat. Equation (9.47) shows that the flux Ji is activated with an apparent activation energy (Ed~- Qa~) which is determined directly from permeation experiments. Since both parameters are positive quantifies, positive as well as negative values can be expected and the flux can be increase as well as decrease with temperature depending on the relative values of Ed~ and Qa~. Equation (9.47) has been used by several authors to describe, analyse a n d / o r simulate permeation and diffusion in silica [59,63,92] and in zeolite membranes [69,72,75]. 1
- Some limiting cases and discussion: At high concentration (high 0, low temperature, relatively large pressure), but within the Langmuir regime, KP >> I and with (9.16b) and (9.43b) or (9.41) one finds d ln pi li = - ~t . Do, i(q) 9qsat,i dz (9.48a) and
bt ]i = -~ Do, i qsat.i
P Li
(9.48b)
i n /~p,i
and with (9.45) and (9.46) assuming qsat~is independent of F:
~t, ln Pf, i ~ Pp, i exp Ji = ~ D~ qsat'iln
E(_~I
(9.48c)
Equation (9.48c) shows that at high values of 0 (low temperature) the apparent activation energy of the permeation equals that of the diffusivity provided that intra-crystalline diffusion is still the controlling mechanism. Outside the Henry region calculation of the permeation from adsorption and diffusion data requires knowledge of the value of qsat~" Especially for weakly adsorbing gases the value is not always known nor can be easily determined from experiments. As discussed by Kapteyn et al. [88] the value of qsat can be estimated from the molar volume which is obtained from extrapolation of the liquid state [90] or from volume filling theory [91]. Some results will be discussed below (binary gas permeation). In the Henry regime separate values of qsat are not necessary as discussed above and the product K'qsat~= b (Henry coef.) 1
Ed~can be larger than Qa,i because molecules can penetrate pores directly without preceding adsorption.
386
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
can be directly measured and used in the permeation equation. To obtain Eqs. (9.43), (9.47) and (9.48) it was necessary to assume that the intrinsic diffusion coefficient D o should be independent of the concentration (occupancy). This is only correct when there are no intermolecular interactions, so for lower values of the occupancy 0. An extensive discussion has been given by Xiao and Wei [86]. Based on model calculations and analysis of experimental results, they showed that Do is approximately constant until 0 -- 0.5-0.6 and then starts to change in a way depending on the value and the character of the interaction energy. This is expressed by a parameter W = A E / R T being a non-dimensional energy change from the non interacting to the intermolecular interacting situation. Analysing their model results it is obvious that for W = 2 and until 0 = 0.85 the relation between DFickand Do is almost similar to that obtained from the Langmuir adsorption type of isotherm. This is equivalent to an occupancy independent D o until large values of 0. For 0 _ Tiso + AT and C T ~ C g the first term in Eq. (9.60) can be ignored. For T < Tiso - AT and C T ~ C a the last term in Eq. (9.60) is negligible. From adsorption isotherms the value of T~o was determined for a number of gases. For highly adsorbable gases like CO2, the magnitude of T~somay be as high as 160~ At p = 5 bar, and in the investigated membrane T~o equals -20, 30, 70 and 160~ for N2, 02, CH4 and CO2 respectively. Adsorption is negligible for He and H 2 at temperatures above ambient. With decreasing pore size the contribution of the gas phase decreases and that of the surface flow increases and interaction between gas molecules and pore wall increases (see Section 9.4.2.1.). This is the case in the small windows between the larger pores. This transition situation from Knudsen-like to configurational diffusion has been modelled by Xiao and Wei [86] for zeolite systems. The activation energy of the gas phase is calculated using LennardJones potentials for the interaction energy, the activation energy for the adsorbed molecules is determined as a difference between the potentials in the pores and in the necks. The ratio of molecular diameter om and pore diameter dm at which the transition takes place depends on molecular shape and zeolite pore characteristics but is situated in the region 0.6 < ~m/dm < 1. A maximum permeability coefficient was estimated by Shelekhin using Eq. (9.60) assuming porosity r and tortuosity 9 values equal to 0.3 and 2 respectively, a pore diameter of 1.5 nm and a micropore volume of 0.11-0.13 m3/g. For gases with T >> T~s~ and so in the regime where bulk gas diffusivity with AEg is dominant, the permeability is strongly dependent on the magnitude of AE. Permeability values for He at T = 90 K are estimated to be 5000 and 9000 Barrer for z~E = 10 and 6 kJ/mol respectively (note: for AE = 0 (Knudsen) this value is 3500 Barrer). With a membrane thickness of 30 ~tm, estimated permeation values for He are 5x10 -~ and 10 -7 mol m-2s -1Pa respectively. Hassan et al. [95] using porous Vycor glass with a pore diameter of about 0.8 nm reported a separation factor ~ equal to 11.5 for O 2 / N 2 at 298 K and of 0~ = 4.6 at 423 K which values are about 20% larger than the perm selectivities. This is due to competitive adsorption in which the relatively strongly adsorbing component (02) saturates the surface and blocks the transport of the weakly adsorbing component (N2). Similar results are reported for C O 2 / C H 4 mixtures (~ 186-122 in the same temperature range). This explanation seems qualitatively in accordance with sorption data of Shelekhin [92] giving a sorption of 2 c m 3 / c m 3 membrane for 02 which is a factor 100 larger than that of N 2 at 30~ -
- Diffusion coefficients and kinetic information: The simplest way to obtain kinetic information is to perform permeation measurements under transient conditions with a non-adsorbing gas in a Wicke-Callenbach experiment [3]. In this case the total amount of permeant qt that has passed through the membrane as a function of time is given by
9 m TRANSPORT AND SEPARATIONPROPERTIESOF MEMBRANES WITH GASES AND VAPOURS
qt 1 De' t L" c-----~= 6 g 2
2 ~2
(-1) n ~
/,/2
De" t exp
_//2/i;2
g2
391
(9.61a)
n=l
which for t --4 ~ approaches the asymptote:
OeC0/t - - ~'2e /
qt= L
(9.61b)
which yields a straight line with intercept (time lag) equal to L2/6Deon the time axis. A similar result is obtained by plotting any quantity which is directly proportional to qt. Here Co = c(z = 0,t) and D e is the effective diffusion coefficient (porosity and tortuosity effects are incorporated in De). If the upstream (high) pressure is constant and much larger than the downstream (low) pressure, the slope of the asymptote will correspond to the steady state and so it is possible to determine the diffusivity under both steady state and transient conditions from a single permeation experiment. With a narrow and unimodal pore size distribution both methods yield reasonable consistent values. Large discrepancies point to strong microstructural effects (bimodal broad distribution, many dead ends, many defects).
9.4.3.4 Illustrative Examples of Permeation and Separation with Microporous Membranes Usually membranes investigated in literature do not have the simple architecture assumed in the preceding theoretical treatment. This requires a number of corrections or modified equations before data of the separating layers can be compared and analysed. This problem is treated in Section 9.5 but results will be used in this section.
(a) Large micropores Shelekhin et al. [92, 56] reported some interesting results for Vycor type of hollow-fibre membranes (for membrane characteristics see Section 9.4.2.2). The theory of permeation of hollow fibre systems will be treated in Section 9.5.
- Pressure dependence of permeation: For He, H2, 02 and N2 a linear dependence of the transmembrane flux on the pressure gradient across the membrane was observed. So the permeation is constant and independent of pressure as expected for Knudsen diffusion and sorbed gases in the Henry regime (and accordingly to the sum of both mechanisms).
392
9 - - TRANSPORT AND SEPARATION PROPERTIESOF MEMBRANES WITH GASES AND VAPOURS
T'I~
~
It, in I
m
120
O~
T-3,0~
I e l
.a
g
U
t-
eq
o 3"heory imi
0
4
8 12 Pressure, s t m
16
Fig. 9.22. Pressure d e p e n d e n c y of the permeability of CO2. After Shelekhin et al. [92].
For C O 2 , which is a highly adsorbable gas, the permeability (Barrer) as a function of pressure at T = 30 and 100~ is given in Fig. 9.22. At 100~ there is a weak maximum above which the permeability slightly decreases with increasing pressure, at 30~ there is a continuous decrease. The two curves could be described with Eqs. (9.60) using Eqs. (9.59) and (9.56) with values for AEads and AEgasof 21 kJ/mol and 10 kJ/mole, respectively. These values were obtained from a best fit of the curves to the experimental results (note: AE=E in Fig. 9.22). The maximum was explained with Eq. (9.60) considering a pressure independent bulk gas term (second term in (9.60)), while the first term for highly adsorbable gases may initially increase or decrease and then decrease with increasing pressure.
- Temperature dependence of permeation: For He the theoretically predicted permeability (Barrer) using Eq. (9.60) exhibits a maximum as a function of temperature for AEgas-4 kJ/mole. Note that the adsorbed gas phase is hardly present here. For larger values of AEgas
9 ~ TRANSPORTANDSEPARATIONPROPERTIESOFMEMBRANESWITHGASESANDVAPOURS
393
TABLE 9.6 Activation energies of diffusion and molecular kinetic diameters for different gases in microporous silica and zeolite membranes Gas
(~m (nm)
Eperm (kJ/mol)
Ed (kJ/mol)
He H2 CO2 02 N2 CH 4 C~-I8 n-C4H8 iso-C4H8 benzene
0.26 0.289 0.33 0.346 0.364 0.38 0.43
22.5 (b)
0.52 (a)
3.68 (b) 13.0 (b) 13.0 (b) 23.4 (b)
9.9 (a) 10.6 (a) 18.2 (a) 28.7 (a)
Eperm (kJ/mol)
Ed (kJ/mol)
15-21 [63] --10
21 [63] 32 [63]
30 [88,89] 0.5 0.585
(a) Ref. [92]; (b) Ref. [56].
there is a c o n t i n u o u s increase of the permeability. N o discussion w a s g i v e n of the occurrence of this m a x i m u m . Such a c o n t i n u o u s (non-linear) increase of the p e r m e a b i l i t y as fiT) w a s o b s e r v e d i n d e e d for CO2, O~- N2 a n d CH4 in the t e m p e r a t u r e r e g i o n of 300520 K. The activation energies of the p e r m e a b i l i t y w e r e o b t a i n e d w i t h a non-linear least-squares fit to the exp. curves a n d are given in Table 9.6 t o g e t h e r w i t h the theoretical calculated ones. These activation energies reflect the activation e n e r g y of diffusion E d in the m e m b r a n e . This table illustrates that, g i v e n the d i a m e t e r of the p o r e there is an increase of the v a l u e of E d w i t h increasing kinetic (molecular) diameter. The reverse t r e n d is f o u n d in the p e r m e a b i l i t y values. Theoretically it is p r e d i c t e d that the v a l u e of B = D.(M/T)l/2should c o n v e r g e to a single v a l u e for all gases for T ~ co. A plot of literature d a t a of B v a l u e s as af(T) for a large n u m b e r of gases yields v a l u e s of 1.1x10 -4 for the Vycor m e m b r a n e a n d 2.2x10 -4 for ZSM5 (Shelekhin [92]) in g o o d a g r e e m e n t w i t h theory.
- Selectivity: Based on p e r m e a b i l i t y data, permselectivities (selectivity factor a = FA/FB) w e r e calculated for a n u m b e r of pairs of gases A-B. At 30~ s o m e typical v a l u e s are a -- 4190 for H2/CH4, (z = 2.5 for H 2 / C O 2 a n d a = 1675 for C O 2 / C H 4. All selectivity factors decrease w i t h t e m p e r a t u r e e.g. at 250~ ~ = 62 for C O 2 / C H 4.
394
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
(b) Small micropores - Silica membranes (permeation and separation): Silica microporous membranes combining high separation factors and high permeation values were first reported by Uhlhorn et al. [28,58] and were further developed and analysed by de Lange et al. [59-63]. More recently silica membranes made by a CVD process with similar qualities were reported by Lin et al. [67] and by Wu et al. [68]. The membranes, synthesised by Uhlhorn and by de Lange et al., were formed from polymeric silica solutions in an ultra-thin layer of about 100 nm thick partly on top (50 nm), partly within (50 nm) the pores of the y-A1203 support (pore diameter ~- 4 nm, thickness 3-8 nm) which was in turn supported by an ~-alumina support (disc) with a pore diameter of about 0.2 ~tm. The characteristics of the silica layer depend strongly on details of the synthesis procedure and a high quality supporting system is required (with low roughness and no or few defects) to obtain good quality membranes. The pore diameters were in the range 0.4-0.5 nm. Discussion of permeation and separation requires some characteristic parameter for the membrane quality. As shown below the apparent activation energy for H 2 permeation gives a good correlation with the separation factor and is used as a measure of quality. Furthermore, the total measured permeation has to be corrected for influence of the support to obtain permeation and activation energy values characteristic for the silica layer (see also Section 9.5). The experimental permeation results could be consistently described using Eqs. (9.43b) and (9.47) for Langmuir and Henry sorption respectively as shown by de Lange in a full analysis of sorption, permeation and separation results of five different gases [63]. This description requires knowledge of adsorption isotherms which could be measured only on unsupported membranes. To use these data for calculation of the permeation of supported membranes requires the assumption of equal pore characteristics in both cases. As discussed by de Lange et al. this is probably not correct in the case of silica layers. Based on sorption data a microporosity of about 30% and a pore size distribution with a peak at 0.5 nm is found. Analysis of permeation data point to a pore diameter of -- 0.4 nm and a considerably smaller porosity. Table 9.7 summarises the sorption data. H 2 and C H 4 have relatively low (isosteric) adsorption h e a t s (qSt) while CO2 and isobutane strongly adsorb. Henry behaviour in the pressure range up to 125 kPa exist at temperatures larger than the limiting temperature Z l i m i t , H e n r y given in Table 9.7. At ambient temperature (323 K) C H 4 showed Henry behaviour up to 8 bar while H 2 exhibited Henry behaviour to at least 15 bar. CO2 exhibited Langmuir behaviour at I bar (323 K).
9 ~ TRANSPORTAND SEPARATIONPROPERTIESOF MEMBRANESWITHGASESAND VAPOURS
395
TABLE 9.7 Henry constants (b), isosteric heats of adsorption qst and lower limiting temperature for sorption behaviour Zlimifl-Ienry for CH4, H2, CO2 and isobutane in microporous silica at P < 125 kPa. After de Lange et al. [59,63] Gas (--~)
CO2
CH4
H2
iso-C4Hlo
qSt (kJ/ mol)
22.3
10.3
6.1
22.9
(-~) b (mol kg-1 Pa-1)
348
273
194
T (K)
($)
~) 0.43 4.2x10-6 2.3x10--6
1.5x10-4 3.7x10-7 1.8x10-7
1.7x10-4
(~)
Zlimit~
Henry (K)
77 194 273 303 323 348 373 473
3.2x10-5 7.8x10-6 5.9x10-6 3.2x10--6 2.1x10-6 4.8x10-7
w
-
2.9•
-5
2.4x10 -5 9.4•
-6
TABLE 9.8 Typical values of permeation and activation energies of microporous silica membranes. Phigh~ 3 bar. After de Lange et al. [59,63] Permeation (10-7 tool m -2 8-1 Pa-1) Gas H2 CO2
50~ (H2) 28~ (CO2) 4.1 (4.5) 2.3 (3.0)
Apparent Eact* (kJ mo1-1) 200~ 21.7 (52.7) 6.8 (32.3)
14.9 (21.7) 6.1 (14.9)
*Values between brackets corrected for support influence.
T h e s u r f a c e c o v e r a g e (0) for C O 2 w a s m a x i m u m 20% at 273 K a n d 125 k P a a n d the isosteric h e a t w a s p r a c t i c a l l y i n d e p e n d e n t of c o v e r a g e . T h i s r e s u l t i n d i c a t e t h a t for all o t h e r g a s e s in the p r e s s u r e r a n g e u p to -- I b a r c o v e r a g e w a s also low. C o n s e q u e n t l y , Eq. (9.47) c a n be u s e d to d e s c r i b e t h e p e r m e a t i o n results. T y p i c a l p e r m e a t i o n r e s u l t s are g i v e n in T a b l e 9.8. T h e p e r m e a t i o n v a l u e s for H2, CH4 a n d CO2 at T > Tiso,Henry w e r e a p p r o x i m a t e l y i n d e p e n d e n t of p r e s s u r e (as e x p e c t e d for H e n r y b e h a v i o u r ) a n d in-
396
9 m TRANSPORT A N D SEPARATION PROPERTIES OF MEMBRANES WITH GASES A N D VAPOURS
creased with temperature from about 4.5-20x10 -7 m o l / m 2 s Pa (425-473 K) for H 2 and 2.3-7x10 -7 mol/m2 s Pa (273-473 K) for CO2 [63]. For membranes with lower quality the increase is less pronounced due to the smaller apparent activation energy. Typical values for iso-butane are 0,6-0.35x10 -7 m o l / m 2 s Pa at 50~ and 200~ respectively. Note that the permeation in this case decrease with increasing temperature. Similar conclusions were drawn by Wu et al. [68] who reported apparent activation energies in the range 11-20 kJ/mol for considerably lower H 2 permeation values. Table 9.8 shows that in the case where the flow resistance is not negligible, corrections should be applied on the total permeation value of the system to obtain the true permeation values of the silica separation layer. Consequently the true values of the activation energy may also differ considerably compared to the apparent ones. (See further Section 9.5). The conclusion of de Lange et al. [61] is that the activation energy of permeation of H2 exhibit a good correlation with the quality of the membranes (permeation, separation factor) and high quality membranes should have an apparent activation energy of at least 10 kJ/mol. Sometimes a weak maximum in the permeation of CO2 as a function of the feed pressure of a similar type as reported by Shelekhin et al. [92] has been observed by de Lange [61] also for small micropores. Separation factors (defined by Eq. (9.36)) obtained from mixtures are usually smaller than permselectivities obtained from the ratio of single gas permeation (see qualitative discussion in Section 9.4.2.1.). At hightemperature and lower concentrations the mixture separation approaches the permselectivities which in turn tend to approach the Knudsen value at high enough temperatures. Typical values for some gas mixtures in combination with permeation data (in the mixture) for different silica membrane systems are given in Tables 9.9a and 9.9b respectively. Several interesting conclusions can be drawn from Tables 9.8 and 9.9. The synthesis method and related membrane quality strongly determines the obtainable combination of permeation and separation values (as characterised by E~ct,H2).High quality membranes have activation energies for permeation (after correction for support influences) in the range 15-22 kJ/mol for H 2 and 10-15 kJ/mol for CO2 with typical permeation values at 200~ of 20x10-7 for H2 and 5x10-7 m o l / m 2 s Pa for CO2 respectively. The permeation value of isobutane at 200~ is very small which indicates a pore size close to that of the kinetic diameter of i-butane and the absence of (larger) defects. Separation factors are in the order of 20-30 for H2/CH4 and 150-200 for H2/isobutane. Lower quality membranes (lower values of Eapp,H2 tend to give larger permeation and smaller separation values for non-adsorbing gases. For strongly adsorbing gases (i-butane, CO2) even with moderate quality membranes ( E a p p , H 2 = 5-10 kJ/mol) good separation factors can be obtained up to about 200~
9 m TRANSPORTAND SEPARATIONPROPERTIESOF MEMBRANESWITH GASESAND VAPOURS
397
TABLE 9.9a Separation factors (defined by Eq. (9.36)) for some gas mixture-silica membrane combinations Membrane
Gas mixture Separation factor a at T (~ ~-25
A13Sil a A13Sil a a. A13Si2 b. A13Sil A13Si2 A13Si2Ti AllSi2
H2/C3I~ H2/CO 2 H2/N2 H2/Ct-I4 H2/CH4 H2/CH4 H2/CH4
AllSil AllSi2
H2-iC4H8 H2-iC4H8
CO2/CH 4
50
48 13 1.7 15-36 2 2 12
100 65 62 2.5
-
Eapp
150
4.5
200
>_250
kJ/mol
28 156 5.5
270 6.6
7-(10) 7-(10) 5-(7)
3 3 50 =9
5 5 150 10
8 10 200 11
11 18 165 11
80 80
105 130
110 170
110 180
12 30-40 11 110 170
5-(7) 7-(9) 12-(16) -
Values are taken from de Lange et al. [59-63], unless otherwise referred. aTaken from Uhlhom et al. [58]. bTaken from Shelekhin et al. [56]. Membrane code: AlxSiy with x and y are number of A1203 and SiO2 layers respectively. Eapp is the apparent activation energy of permeation for H2. Figure in parenthesis is corrected for support influence.
TABLE 9.9b Permeation values of some gas mixtures in different silica membranes Membrane
Gas mixture
A13Sil a A13Si2
CO2/CH4 H2/CH 4
A13Si2 A13SilTi
H2/CH4 H2/CH4
Permeation F at T (~ 50
100
4 (CO2) 3.7 (H2) 1.8 (CH4) 1.8 (H2)
10
2 (H2)
150
200
3 (H2)
50 (H2) 4 (H2)
Membrane code: see Table 9.9a. Permeation given in 10-7mol/m 2 s Pa.
For non-adsorbing or weakly adsorbing gases (H2, CH4, N2, 02) the permeation increases with temperature (for high quality membranes). This is in accordance with data of Wu et al. [68] who reported increasing permeation values (H2) for membranes with lower quality (characterised by positive values of Eap p for
398
9 - - TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES w r r H GASES AND VAPOURS
N2). Permeation values of Wu at 600~ are in the range 0.03-1.0x10 -7 m o l / m 2 s Pa for H 2 (for N 2 a factor of 20-70 lower) with the highest value for the lower quality membrane (Eact,H2-~ 11 kJ/mol). As will be discussed in Section 9.5 the different values cannot be compared directly because of the strong influence of pressure conditions and support effects.
- Silica membranes (diffusion data): Equation (9.47) can be used to calculate the activation energy for "intracrystalline" micropore diffusion Ed,/ of specimen i provided sufficient sorption data are available. The value of Ed,/ follows from: Edd = Eapp- Qa,i
(9.62)
where Qa,/is the isosteric heat of adsorption and Eapp is the measured apparent activation energy of permeation (Eq. (9.47)) after correction for support influences. With typical values of Eap p equal to 15 kJ/mol (H2) and 10 kJ/mol (CO2) and typical values of Qa equal to 6 (H2) and 23 kJ/mol (CO2) [63] the resulting calculated activation energies of the intra channel (micropore) diffusion are about 21 kJ/mol for H2 and 32 kJ/mol CO2 [59,93]. This is in accordance with the expectation that larger molecules will have a larger activation energy for diffusion than smaller ones [92,82]. Equation (9.47) is also used by de Lange [63] to calculate the value of the diffusion coefficient of several gases in silica membranes. The term ~t in (9.47) takes the form ~t = p(1 - ~)/~ with the skeletal density of silica p = 2.2 k g / m 3, the silica porosity ~ = 0.4 and the membranes thickness L = 100 nm. Taking all the sorption terms together in the Henry constant b (which can be directly measured) and substitution in (9.47) yields: J = 3.3 x 101SD x b
(9.63)
Typical values of b and D for a range of membranes are given in Table 9.10 together with some other parameter values. The range in D values reflects differences in membrane quality, the smallest D values being formed in high quality membranes. The diffusion coefficients become smaller in the same order as the kinetic molecular diameter (see Table 9.6) increase. The large differences in the D values indicate that the pore diameter is of the order of the molecular diameters (0.4~.5 nm). The differences in D values are much larger than the differences in permeation values and indicate the effect of the sorption term even for weakly adsorbing gases (compare H 2 and CH4). The absolute magnitude of the diffusion coefficient is rather uncertain, because all uncertainties concerning the value of ~t are reflected by the D values.
9 ~ TRANSPORT AND SEPARATIONPROPERTIESOF MEMBRANESWITH GASESAND VAPOURS
399
TABLE 9.10 Typical values for the diffusion coefficient of different gases in silica membranes at room temperature after de Lange et al. [63] Gas
b (289 K) ( m o l / k g Pa)
D (,298 K) (m'/s)
Ed,i (kJ/mol)
qst (kJ/mol)
H2
1.4x10 -7
5.25x10 -11
13-21
--6
CO2
1.4x10 -5
6-9x10 -13
--30-33
=23
CH4
1.6x10 -6
5-15x10 -13
--10
iC4H10
7.3x10 -5
3x10 -14
---23
The value of the porosity is taken from adsorption measurements on unsupported silica membranes and probably the porosity of supported silica membranes is considerably smaller and the calculated D values give a lower limit. A comparison with zeolite data and effects of surface reactions will be discussed below. - Zeolite membranes:
Permeation and separation data on well defined, high quality zeolite membranes are only reported for MFI (ZSM-5, silicalite) zeolites grown in situ directly from the precursor solution on top of a substrate. The experimental single gas permeation results could be in a number of cases consistently described using Eqs. (9.43)-(9.48) for the Langmuir and Henry regimes. Geus et al. [70] give a detailed description of the synthesis of a MFI layer with a thickness of about 50 ~tm on top of a porous steel support. Vroon et al. [72,74] synthesised thin (2-6 ~tm) MFI layers on a (x-A1203 support and varied the crystallite size (0.1-0.4 ~tm) in the layers by varying the synthesis temperature and using a very high pH (-~ 12.5). Both groups of authors investigated the quality of their membranes. Both groups of authors measured a very small flux of gas molecules which are much too large to pass the pores of the MFI structure indicating that some larger pores were present in the layer. The measured fluxes for iso-octane (or of 2-2-di-methylpentane) were more than 5 orders of magnitude smaller than that of C H 4 indicating a good membrane quality. This conclusion is supported by the large observed separation factors for e.g. H2/butane, CH4/butane and n-butane/i-butane (see below). Vroon had to apply two silicalite layers on top of each other in order to obtain this good quality. Typical single gas permeation data for relatively thick MFI membranes are given in Fig. 9.23 [71]. At 673 K all the gases show a linear dependence on the (feed) pressure (Henry behaviour) as is the case at 300 K for the noble gases and for C H 4 , whereas butane and ethane exhibit saturation at low and higher
400
9 - - TRANSPORTAND SEPARATION PROPERTIESOF MEMBRANESWITH GASESAND VAPOURS
10
22
(a)
20 8 18
N
16
s
~4
6
E E
~2
5
~'~
10
0
ft.
-~
@
T
~W"
.,..m
O.
W
0
E 0
.E 0
4
e
~" , . b u t a n e * 100 A
W W
o
3
O. t
0
20
40
60
80
I O0
Partial pressure in feed (a)
(kPa)
Fig. 9.23. Steady-state flux and permeate pressure as a function of partial feed pressure for different gases at 300 K (a) and at 673 K (b). After Bakker et al. [71]. (0) neon, (+) argon, (V) krypton, ( 9) methane, (A) ethane, ( I ) n-butane, (&) isobutane, (O) CFC-12.
pressure (30 and 80 kPa) respectively. The permeation increases with temperature for all gases except krypton and CH4 which were almost independent of temperature. The maximum observed permeation values (673 K) of noble gases and CH 4 are about the same and correspond with a permeation of 1.6-2.3x10 -7 m o l / m 2 s Pa. Permeation values of 1.2 and 0.9x10 -7 m o l / m 2 s Pa are found for n- and i-butane respectively. Similar results concerning the trends in the permeation values as a function of pressure are reported by Vroon et al. [72,73,74] for CH4, ethane, propane and butane. The absolute values of the permeation reported by Vroon et al. for these
9 -- TRANSPORTANDSEPARATIONPROPERTIESOFMEMBRANESWITHGASESANDVAPOURS
401
10
22
(b)
20
8
18
O.
..~ Q)
A(/)
16
7
E
~4
6
12
S
.E
4
~
3
ID ~" t2.
2
"~.
0
E E x
~..
10
8 II
0
20
40
60
80
m
E
~00
Partial pressure in feed ( k P a ) (b) Fig. 9.23 (continued). Caption opposite. MFI m e m b r a n e s on (x-A120 3 supports showed for CH 4 a decrease of the permeation from lx10 -7 (298 K) to 0.6x10 -7 m o l / m 2 s Pa (473 K) with increasing temperature. Plots of the flux of butane (Fig. 9.24), propane and ethane versus temperature exhibit a (weak) m a x i m u m which values shifts from 440 K for n-butane to 350 K for ethane at 100 kPa. This m a x i m u m depends on the (partial) pressure of the gas (e.g. for n-butane at 8 kPa pressure the m a x i m u m is situated at about 390 K). Similar maxima are found [72,74] in the curves of H 2 and CO2 vs temperature as s h o w n in Fig. 9.25 and are also reported by Kapteyn for n-butane [88,89]. The absolute values found for CH4, CO 2 and n-butane can be compared with that obtained by Bakker [71], Geus [75] and Kapteyn et al. [88]. At 473 K the values obtained by Vroon are lower by a factor of about 2.5 compared to that of
402
9
-
TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
-
15 0
0 0
A
qll I
E
,/.
0 o
00 O
10
0 o
0
0
0@0000
m m
0
E E v
X
0
O O0
o
vvVVVVVVVv
vvvvVVVVYV
m
u
vvVt~ vvv vV V
0 273
i
,
I
,
I
373
473
Temperature (K) Fig. 9.24. Comparison of the methane and n-butane flux measured by the dead-end method (O,V) and the Wicke-Kallenback method (o,v). After Z. Vroon [72-74].
20
,..
15
oO~176176176176176176176176176176
10r
oO
~1 v v v V V F
0
-vvvvvvvvvvvv
+ "l" "1""1"+ + + + + + + + ,1" + +
50
100
150
V
l /
+++t 200
Temperature (~ Fig. 9.25. Flux of hydrogen (O), helium (V), carbon dioxide (+) and sulphur hexafluoride A as a function of temperature at a feed pressure of 100 kPa. After Vroon et al. [72-74].
9 -- TRANSPORT AND SEPARATIONPROPERTIESOF MEMBRANESWITH GASESAND VAPOURS
403
Bakker. This means that the permeation values corrected for thickness differences (permeability) of the thin membranes with small crystallites on a support with relatively narrow pores (0.2 ~tm) are much lower than of the thicker layers on a steel support. This point will be discussed later. Kapteyn et al. [88,89] and Vroon et al. [72,74] could model and describe their single gas permeation measurements for C H 4 and n-butane rather well with Eqs. (9.41)-(9.48) taking the thermodynamic factor in the Langmuir regime from adsorption measurements. Using Eqs. (9.40) and (9.43), Kapteyn showed that the corrected (intrinsic) value of the diffusion constant Do at 300 K of n-butane is independent of pressure (up to 1 bar) and Do (300 K) equals 0.4x10-6 c m 2 / s (contrary to the Fickian diffusion constant which is strongly increasing with increasing 0). The maximum in the flux versus temperature for n-butane could be correlated with a strong decrease of the occupancy 0 at higher temperatures. Occupancies at p = 0.5 bar vary almost linearly from 0 = 0.8 at 350 K to 0 = 0.2 at 450 K and the maximum in the c u ~ e is situated at 0 -- 0.4. Initially the change in 0 is less. So at lower temperature the diffusion coefficient increases more rapidly than the concentration (occupancy) decreases, at higher temperature the reverse is true and this give rise to the observed maximum asflT). A similar result is reported by Vroon et al. [72,74] who calculated the flux of C H 4 and n-butane using also Eqs. (9.40) and (9.43) and using diffusion constants taken from literature [94] and measured on twinned single crystals by the membrane method. The sorption data for methane taken from literature agree within 20% from data obtained by Vroon, for n-butane not sufficient literature data are available and measured data (gravimetric method) are used. The set of data used in the calculations is given in Table 9.11 and the calculation results in Figs. 9.26 and 9.27. TABLE 9.11 Henry constants and saturation concentrations obtained from the gravimetric sorption measurements on silicalite particles and diffusion constants obtained by the membrane method of methane and n-butane. After Vroon et al. [72-74] Gas
Temperature (K)
Henry constant (mol Pa -1 m -3)
Saturation concentration (mol m -3)
Diffusion coefficient (m 2 s -1)
Methane
298 323
8.6x10 -3 5.4x10-3
-
0.7x10-1~ 1.0x10-1~
n-Butane
298 323 348
17.5 4.0 1.0
2.2x10 3 1.8x103 1.4> I for the S / W mixture at lower temperature to values ~ < 1 at high temperature (note a > I for the W / S combination is equivalent to ~ < 1 for the S/W). This can be explained by preferential sorpfion of the strongly adsorbing component which excludes (or decreases) the concentration of the S component. With increasing temperature the concentration of the S component decreases much more strongly than that of the W component, the "blocking" effect decreases and finally vanishes and at high temperature the mixture starts to behave in a similar way to a mixture of two W components.
408
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
6
A
t
4
I
o
E E I
V v
2
N l
V
la,
0
,
9 ,
,
,i,,,
9
i
Temlmrature (K)
Fig. 9.28. Permeation and separation behaviour of a mixture of 50 EPa C1-~ ( 9and 50 EPa n-butane
(V) as a function of temperature of a MFI membrane. Single gas permeation values are added: CH4 (O), n-butane V). After Vroon et al. [72-74]. 25
-'" ?
E
,,,,,=
O
E E
f
2o
~s n-Butane
/
10
"0 e ~
X
ft.
s
Hydrogen/ 9
300
,
~
&
400
500
,,i
600
Temperature (K) Fig. 9.29. Separation b e h a v i o u r of a H 2 / n - b u t a n e m i x t u r e (1:1) as a function of t e m p e r a t u r e of a MFI (silicalite) m e m b r a n e at 100 kPa. After K a p t e y n et al. [99].
A q u a n t i t a t i v e t r e a t m e n t of this complex b e h a v i o u r is not yet p u b l i s h e d . The case of a m i x t u r e of t w o S c o m p o n e n t s is e v e n m o r e c o m p l e x a n d general qualitative descriptions h a v e not yet b e e n published. E x a m p l e s of p e r m e a t i o n
9 -- TRANSPORTAND SEPARATIONPROPERTIESOF MEMBRANESWITH GASESAND VAPOURS
10-' V V
A (N I
E
T I m o
E E
x =l m u,,,
V'
V
V
V
V
409
V V
T
10"
V V
A
A
10-*
A
A
A
of equipment
10" 273
373
473
Temperature (K) Fig. 9.30. Separation and permeation behaviour of a mixture of 0.31 kPa paraxylene (V) and 0.26 kPa o-xylene (A) as a function of temperature. Single gas permeation data are also given: 0.62 kPa px (A) and 0.52 kPa ox (V). The total pressure was 100 kPa, the balance being He. After Vroon et al.
[72-74]. and separation results of these combinations are reported by Vroon et al. [72,74] for n-butane/i-butane, benzene/cyclohexane, methane or hexane/2,2-dimethylbenzene and p/o-xylene mixtures. The separation behaviour of a p / o xylene mixture is given in Fig. 9.30. The permeation of the paraxylene is much larger than that of the o-xylene at higher temperature, the last one has a permeation which is at the detection limit of the equipment used. The molecule has a diameter which is larger than that of the pore diameter of the MFI and so we have here an example of separation by size exclusion. The flux of p-xylene shows a weak maximum as a fiT) and consequently the separation factor does the same with a peak value of c~ = 100 at ~400 K under the given conditions. The separation factors and the permselectivities are equal as expected for the size exclusion mechanism. Xiang and Ma [76] reported a value of ~ = !5 for p/meta-xylene separation with a flux of 35 ml m -2 h -1 (=--4.3x10-7molm -2 s-1) for the m-xylene at room temperature. An even more straightforward example of size exclusion is exhibited by the mixture of n-hexane and 2-2 dimethylbutane where the flux of the hexane is three to four orders of magnitude larger than that of the 2-2 dimethylbutane up to the highest temperature measured (473 K) and the flux of the 2-2 dimethylbutane
410
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
is of a similar magnitude of that of o-xylene Vroon [72,74]. Obviously a small number of small defects accounts for the remaining flux of the large molecule. Finally it should be noted that isomers like n-i-butane [73,89] and cis-butane/ 2-trans-butanes [76] exhibit different permeances indicating the shape selective properties of the zeolite membranes. - Diffusion in zeolite membranes:
Diffusion data can be obtained by a wide variety of different techniques which yield diffusion data which, for the same material, can differ by more than 4 orders of magnitude. So it is outside the scope of this paper to give a full discussion of diffusion data. Some relevant aspects for membrane permeation will be mentioned. A comparison of diffusivities of n-butane in silicalite obtained by different techniques is given by Kapteyn et al. [88]. Compared with his experimental results obtained from steady-state permeation measurement using Eqs. (9.50) and (9.51), values obtained by single crystal (membrane) measurements [94] are too low by more than two orders of magnitude. These single crystal data however reproduce reasonably the permeation results of Vroon et al. [72,74] as discussed in the preceding sections. The diffusion data of Kapteyn et al. agree well with diffusion data obtained by frequency response (FR) and square wave (SW) methods. Kapteyn argues that the diffusivity of n-hexane in silicalite is not influenced by the fact that the crystals in the membrane are intergrown and assumes that the same holds for n-butane. The intrinsic (corrected, Maxwell-Stefan) diffusion coefficient Do of n-butane in silicalite can be described by an Arrhenius equation with a pre-exponenfial coefficient D~ -- 0.053 c m 2 s -1 and an activation energy for diffusion E d = 29.8 kJ mo1-1. At 300 K this gives a value of Do = 4x10 -7 cm 2 s-1. It should be noted that the absence of effects due to intergrown particles does not mean that grain boundaries do not play a role, as has been shown by Vroon [72,74] and discussed in the preceding section. Vroon et al. report values obtained by transient measurements on their silicalite membranes using Eq. (9.61b) and find a good agreement with values obtained from steady-state membrane measurements. Values obtained from transient measurements in sorption experiments on powdered material are two orders of magnitude smaller. Geus et al. [75] reports diffusion data at 21 and 145~ for H2, N2, C H 4 , C O 2 and CF2C12 in silicalite membranes on a clay support which are obtained with the similar transient permeation technique as used above by Vroon. The diffusion coefficients for methane are about two orders of magnitude smaller than those obtained by PF-NMR methods. Usually this last technique gives relatively large diffusion coefficient values, which in the case of n-butane are of the same order of magnitude as reported for FR techniques and membrane techniques as reported by Kapteyn.
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
411
Geus ascribes his low values to the influence of the support which has a low porosity. Indeed, uncertainties in geometric aspects of the separation layer and the membrane system affect the value of the measured diffusion coefficients. The conclusion so far must be that synthesis and sample preparation techniques play an important role. Diffusion data to be used in permeation experiments should be measured on membranes with techniques which reflect as closely as possible the transport phenomena during permeation. This also minimises heat effects due to adsorpfion/desorption which play an important role in diffusion experiments based on large crystals, but is of minor importance in membrane experiments [101].
9.4.4 Surface Effects on Permeation in Microporous Membranes In the preceding discussion it was assumed that the transfer from molecules from the gas phase to the solid (porous) membrane was not the rate-determining step in the permeation. This assumption will be evaluated in this section because in oxygen permeation of dense oxidic membranes surface reactions become clearly rate determining for several groups of materials (see also Chapter 10). For the best permeating dense materials (perovskites) with relatively large exchange coefficients, surface reactions become rate determining with membrane thicknesses in the range 0.3-1.0 mm corresponding with flux values in the range 0.4-4.0x10-6 mol m -2 s -1 (corresponding to a permeation value of 0.44.0x10-7 mol m -2 s -1 Pa -1 with pressures of I bar and about zero at the feed and permeate side respectively). This high oxygen permeation is comparable with or somewhat lower than many of the permeation values for microporous membranes. De Lange [63] used a gas kinetic expression to estimate the total number of molecules Zwan colliding per second with the walls of a volume: 1N_
(9.64)
Zwall -" 4 V v
where V is the molar volume, v the mean molecular velocity and N / V is the number of molecules per unit volume. At 1 atm and 300 K the calculated value of Zwall is approximately 1.8 mol c m -2 s -1. A typical hydrogen flux through the microporous membranes is 10x10-6 mol c m -2 s -1 (calculated from a permeation of 1 0 X 1 0 -7 mol m -2 s - 1 P a -1 at a pressure difference of I bar). Not every collision leads to penetration of the molecule into the membrane. This is expressed by the sticking factor t as defined by Eq. (9.65)"
P 1
R a = t [ (2~MRT)I/2 e x p -
E('R-T-/
(9.65)
412
9 a TRANSPORT A N D SEPARATION PROPERTIES OF MEMBRANES WITH GASES A N D VAPOURS
The sticking factor gives the ratio of the number of activated collisions divided by the total number of collisions, whereas Ra in Eq. (9.65) gives the rate of adsorption (in mol cm -2 s-l), with an activation energy Earsfor adsorption at the external surface, the other parameters having their usual meaning. According to Turkdogan [5] the maximum value of R a in Eq.(9.65) is obtained by setting t equal to unity and zero activation (adsorption) energy (Ea = 0). Equation (9.65) then transforms to the classical Hertz-Knudsen equation for the number of moles striking a unit surface area per unit time Rmax: Rmax =
P
2 ~ ( M R T ) 1/2
(9.66a)
and with p given in atmosphere: Rmax - 44.3p
- (MT)I/2
(9.66b)
with Rmax in mol cm -2 s -1. Note that Eqs. (9.66b) and (9.64) will produce figures of similar orders of magnitude and that Eq. (9.66a) also gives the maximum rate of vaporisation from a non-contaminated surface at low pressures. For microporous membranes only the porous part of the surface (~) is available for penetration; the solid is assumed n o t to accept molecules. For small molecules hitting the surface under not too low angles it is reasonable to assume a low value of the activation energy for pore penetration (this is process F1 i n Fig. 9.21). A pessimistic estimate for microporous silica membranes using values of ~ = 0.01 and t = 0.01 yields at 300 K and I atm a collisional flux (of H2) which is at least one order of magnitude larger than the permeation (flux) values found by de Lange et al. [63]. The conclusion is that for relatively small molecules (H2, CO2, etc.), permeation in microporous (silica) membranes is not limited by surface reactions and direct penetration in the pores is the dominant mechanism in a wide range of temperature and pressure conditions [63]. This conclusion does not hold for large non-spherical molecules. Here sorption is necessary, the sticking coefficient becomes very important and surface reactions probably will limit the permeation as soon as bulk permeation becomes appreciable. To the knowledge of the present author, no investigations of this phenomenon in microporous membranes have yet been reported. In dense, non-porous membranes, surface limitations to oxygen permeation are a common phenomenon as can be understood from the very low adsorption levels and large activation energies on the dense membrane materials (see Chapter 10). For hydrogen permeation in dense metal membranes estimates have been made by Govind [105].
9 - - TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
413
The implication of the theoretical considerations given above is that the permeation can be increased in cases of low adsorption and sticking coefficients by application of a mesoporous top layer with better sorption properties on top of the microporous membranes. Selective sorption should then also lead to an enhanced separation factor (see Eq. (9.71)). Indications for this effect are reported for dense membranes by Deng et al. [106] and for microporous silica membranes by Nair [107]. 9.5 PERMEATION A N D SEPARATION IN MORE COMPLICATED SYSTEMS
Real membrane systems to be used in practice usually do not have the simple architecture assumed in the preceding quantitative treatments (single-wall, non-supported) nor do they fulfil basic boundary conditions, i.e. well mixed gas mixtures, homogeneous gas compositions and pressure (no gradient) across the membrane length (flow direction of feed/permeate). In those cases the aerodynamic conditions of the feed and permeate flow, the precise design and the type of permeate removal (sweep gas, vacuum suction) are important. In the case of supported membranes the effect of the support has always to be evaluated, and if not negligible, corrections for support effects should be applied even with simple membrane architectures. A full description of permeation and separation in practical systems is out of the scope of this paper. Two important cases will be treated for illustration because of their importance for laboratory experiments.
9.5.1 Hollow Fibres In the case of hollow fibres, or long cylindrical tubes, the pressure drop across the membrane length is not negligible. In the case of hollow fibres with a characteristic ratio of length-to-inside-diameter of 104 this pressure drop is very large and the gas densities at inlet and. outlet differ considerably. Then the gas flow is a compressible flow. Shelekhin et al. [56] derived a set of three expressions to describe the permeation of single gases through a micro porous hollow fibre (Vycor type) which, in the general case, should be solved numerically. In the special case of a relatively low permeable gas, the pressure drop along the fibre becomes again negligible and the permeation Fp (mol m -2 s -1Pa -1) can be calculated directly: 9 T . r i In Fp =
S(po-
(ro/ri)
(9.67)
P3)
with (I) T the transmembrane flow rate, ro and ri the outer and inner radii of the fibre, S the membrane surface area (m 2) and P0 and P3 the inlet pressure and the pressure on the permeate side of the membrane, respectively.
414
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
The pressure distribution along the fibre was expressed as P/Po and was calculated with the complete set of equations. Values of P/Po across the complete fibre length (so Po/Pl with Pl is outlet pressure) for He at 30 and 250~ are equal to 0.8 and 0.5 respectively and use of Eq. (9.67) gives wrong results. This is illustrated by a comparison of the permeation versus temperature curves of He which give a maximum if Eq. (9.67) is used but give a continuously decreasing function when the pressure drop is taken into account. In the case of gas mixtures the gas composition of both feed and permeate flows changes along the membrane length. There is a difference in behaviour between co-current and counter-current flow of feed and permeate streams. A brief description for separation in a single stage module with ideal mixing and of a coupling of modules to form cascades or membrane rectification units is given by Eichmann and Werner [18]. It is illustrated that the concentration at the permeate side changes across the membrane length. An implicit expression to calculate the concentration of a binary mixture at the outlet of the membrane system as a function of the inlet concentration is given and so the separation factor can be calculated. This equation gives a good description of the actual behaviour of gases as illustrated e.g. for H2/CO 2 mixtures. The effect of several important parameters (e.g. average pressure, feed or permeate pressure, feed or permeate pressure at outlet, temperature) is illustrated and the necessity to select an optimum set of parameters, given economical boundary conditions, is shown. An extensive treatment of this type of problem is given by Sengupta and Sirkar [114]
9.5.2 Multilayered, Asymmetric Supported Systems The use of supports in asymmetric, supported membranes introduces a number of complications in the interpretation of permeation and separation data as well as in the optimalisation of membrane systems. If the flow resistance of the support is not negligible, there is a pressure drop across the support. This implies that the pressure and so the occupancy at the interface of separation layer and support is different from the (directly accessible) pressure at the support surface, usually the permeate side. Consequently, the driving force for permeation through the separation layer is different from the total driving force across the membrane system. In cases where one wants to calculate or compare transport properties of the separation layer material, it is necessary to correct for this effect (for illustration see below). Expressions to calculate the pressure Pint at the interface of top layer and substrate and thus to calculate the pressure drop across the top layer only are originally derived by Uhlhorn et al. [21] and further developed and used by Lin et al. [103,104] and de Lange et al. [59,60]. More recently Uchytil [102] used and refined this method for different cases. De Lange [60] gives an illustration of the
9 ~ TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
415
calculation on a typical, supported mesoporous y-A1203 membrane. It is assumed that the flow resistance in single gas permeation experiments is a series combination of the flow resistances of support and top layer, respectively, where the permeation is a reciprocal resistance. The support permeation can be expressed as (see also Section 9.2.3.): F0,s = gs + ~ls" Pav
(9.68a)
where gs expresses the Knudsen component, [3s- Pay expresses the Poiseuille (Laminar) flow component and Pav is the average pressure (Phigh- P!o:w). For gs and as gas kinetic expressions can be given [103]. A similar expression to Eq. (9.68) usually fits very well the behaviour of the membrane (F0,m: support + top layer) so: F0,m = g m + ~ m " P a v
(9.68b)
The values of gs and [3s are calculated from measured permeation data for non-adsorbable gases (He, Ar, H2) using Eq. (9.68a). The permeation or permeability properties of the top layer are calculated now by subtracting the permeation data of the support only from the measured permeation data of the membrane using the series model. Note that Ph (high pressure) and P1 are measured at the interfaces of gas/top layer and gas (permeate side) / support respectively. The pressure Pi at the interface of top layer and support can be calculated by
Pi= 1~s -gs +
2 + [3s. P~ + 2gs" P1 + 2
(9.69)
where ~ is the flow rate (tool or m 3 s -1) and A the membrane surface area. The support permeation for the actual experiment is given by: F~
-
A(Pi-
P1)
(9.70)
The theoretical validity of Eq. (9.68b) is discussed by Lin et al. [104] and it is shown that this equation is a special simplified case of a more general, but very complicated expression which strictly holds for the case that ~m/gm = ~s/gs. Uchytill [102] also devotes an extended discussion to this problem. Typical examples of the value of Pi and of the magnitude of the corrections are given in the cited literature. For ~/-A1203 top layers (thickness -- 4 ktm, pore diameter 4 nm) on an ct-A1203 support (thickness -- 2 ram, pore diameter -- 0.2 ktm) Uhlhorn [21] reports a value of Pi = 55 kPa with Ph = 80 kPa and P1 = 7 kPa and H 2 as the permeating gas. This means that in this case only 30% of the total pressure drop is across the 7-A1203 top layer; the remainder is across the support. De Lange et al. [60] applied a
416
9 --TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
microporous silica layer on top of supporting system similar to that used by Uhlhorn. The silica layer has a thickness of about 100 nm and a pore diameter of 0.4-0.5 nm. At 50~ the relative pressure drop across the support is about 4%, while at 250~ this is 15% (with the same gas flow 9 of 7.1x10-6 mol s-1) and the main pressure drop being across the silica top layers. At 250~ the value of Ph = 0.71 bar and of Pi = 0.12 bar. So in this case the correction increases with increasing temperature and with increasing total permeation values due to a decreasing contribution to the permeation of the silica top layer.
Effects of the support on separation The low pressure PI is measured and can be manipulated only at the interface gas(permeate)/support. This implies that when the support resistance is not negligible the value of Pi at the interface of support and top layer can be considerably larger than P1. Especially important is the case with strongly adsorbing gases where even a small increase of Pi c a n lead to a large increase of the occupancy at this interface and consequently to strong effects on the relative permeation contributions (separation) in gas mixtures. According to Eq. (9.38) this also means that the real separation factor of the top layer is decreased with respect to the ideal separation factor by back diffusion from the support (see Sections 9.3.1 and 9.3.2). This becomes especially serious when the conditions are such that the support is in the viscous flow or in transition regime from viscous flow to Knudsen flow. (This means that the support has no or hardly any separation properties itself.) Even relatively small amounts of non-Knudsen contributions in the diffusive transport (which hardly affects the permeation) can decrease the separation factor considerably (see Eqs. (9.38) and (9.34)). This implies that to obtain maximum separation factors the support resistance should be as small as possible and vacuum suction is preferred above use of a sweep gas to remove the permeate (from the permeate side). If the conditions are such that the mesoporous support is in the Knudsen regime, and so has some separation properties, the separation factor can be enhanced when the feed is applied from the support side. In this case the gas composition at the interface between support and separation layer is enriched already somewhat. This effect is reported by Keizer et al. [20] and could be described by the empirical relation: Or
= O~supp- O~oplayer
(9.71)
9.6 OVERVIEW OF I M P O R T A N T RESULTS
In this section a brief overview will be given of the most important results of permeation and separation. It is not the intention to give a complete review of
9 ~ TRANSPORT A N D S E P A R A T I O N PROPERTIES OF MEMBRANES W I T H GASES A N D VAPOURS
417
all available literature but merely to illustrate the state of the art, to show possibilities and to compare results with porous systems with competing dense membranes.
9.6.1 Introductory Remarks Permeation and separation data reported in the literature are difficult to compare directly. This is due to the variety of parameters which influence the absolute value of permeation and separation data and which are usually badly described and sometimes cannot even be adequately described. As is shown in the preceding sections the pressure conditions and the flow dynamics (aerodynamic conditions) play a very important role. These pressure conditions are not always adequately described and data describing the external flow conditions do not directly reflect flow conditions in the membrane (model design a n d / o r membrane architecture playing a role). Flux data in mol (or m 3) per unit of time and surface area are the preferred data. To obtain data reduction and to make comparison easier permeation (permeance) data are usually given. One should realise however that this is only meaningful if the flux is a linear function of pressure (difference), so in the Henry region. Permeation data given as permeation (permeance) must be accompanied by information concerning the validity region (pressure boundaries) and the form of the pressure dependency. In the latter case this leads generally to a dimension of mol m -2 s -1 Pa -x with 0 < x < 1. A membrane material with a high permeation which is valid only in a small pressure range and which "saturates" at low pressure is inferior compared with a membrane material with lower permeation which is valid in a wide pressure range. Data given in the form of permeability (mol m / m 2 s Pa) are usually meaningful only in symmetric membranes (single, homogeneous wall, non-supported). In asymmetric supported membranes the use of permeability data can give rise to much confusion and erroneous conclusions for several reasons. In most cases the layer thickness is not precisely known and usually it is not known whether this layer is homogeneous or has property gradients (e.g. a "skin" and a more porous part). In many cases the material of the layer penetrates the support to some extent and so it is not possible to separate properties of separation layer and support without giving account of the interface effect. Finally, even if all these complications can be avoided, a comparison based on separation layer properties expressed in terms of permeabilities can give a completely wrong impression of the practical possibilities (as done in e.g. Ref. [109]). This is illustrated by comparison of hydrogen permeabilities of ultra-thin silica layers (see Tables 9.14-9.16) with other materials such as zeolites and metals. The "intrinsic" material properties of these silica layers are not impressive;
418
9 - - TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
nevertheless m e m b r a n e s give the highest permeation values reported in literature in combination with good separation properties. This is due to a "technology factor", i.e., the possibility to make them extremely thin. This cannot be obtained so far with other materials. Comparison per unit of thickness ~ which is the essence of p e r m e a b i l i t y - gives the impression that equal thicknesses for all materials can actually be obtained, which is not the case. Furthermore, limitations imposed by surface reactions becoming rate limiting at a given thickness are not taken into account.
9.6.2 Typical Permeation and Separation Data for Porous Membranes Most of the data are taken from an overview of Burggraaf [108] which has been u p d a t e d with results reported later. Some typical results obtained with capillary condensation a n d / o r surface diffusion as transport mechanism are given in Table 9.13. A discussion of these data is given in Section 9.2.3.3. As is shown, interesting combinations of high to very high separation factors with reasonable to good flux values can be obtained. Typical results for supported microporous silica m e m b r a n e s are given in Table 9.14 and are partly discussed in Section 9.4.3. The data given by de Lange (see Table 9.14) are all in the Henry regime and the permeation of H 2 and CO2 is in the range of 10-25 and 3-6x10 -2 mol m - 2 s - 1 bar -1 TABLE 9.13 Some typical results with capillary condensation and surface diffusion in meso- and macroporous membranes Membrane
Thickness Pore dia- Separation (~Jm) meter factor (nm)
Permeation (mol m-2 S-1 Pa-1)
T-A1203modified with Ag
10
3--4
H2/N2:8
H2:35x10-6
T-A1203
4
3
C3H6/N2:
T-A1203 m o d i -
C3H6/N2:
fied with MgO T-A1203 5--8
4
T-A1203 + silicate
4-5 0.5
10
26 84
Temp.(~ Ref. Press.(bar)
230 0.1 (H2) C3H6:30x10-6 -10 C3H6:1.6x10-6 -10
Methanol/H2: Methanol: 680 =20x10-6 Methanol/H2: Methanol: 110 =1.6x10-6 Methanol/H2: Methanol: >1000 5.6x10-6
100 2.2 200 23 100 7.7
Ulhorn et al. [28] Ulhom [37] Ulhorn et al. [37] Sperry et al. [39] Sperry et al. [39] Bai/Jai Noble [27]
Permeation is measured at relative pressure p/po where pore is filled (capillary condensation). p0 is condensation (saturation pressure) of free liquid.
9 -- TRANSPORT AND SEPARATIONPROPERTIESOF MEMBRANESWITH GASESAND VAPOURS
419
TABLE 9.14 Overview of typical flux and separation data of supported microporous silica membranes m a d e by different processes Material / gas
o~
Flux (10-2 tool m -2 s -1)
Temp. (~
SiO2 (microporous) on 3'-(x-A1203 support
Pressure / thickness de Lange [5963] d=100 n m
H2/CO 2
4-10
H2: CO2:
10-25 3-6
200
I bar (-~ small
H2/CH 4 H2/n/i-butane
30--40 160
CH4: i-but:
0.9-2.5 0.3--0.6
200 250
" "
O2/N 2
2-4
02:
2-6
200
"
O2/N2
11.5 4.5
? ?
H2/N2
15.7 4.2
H2: H2:
H2/i-butane
243 40
H2: i-but: i-but:
75 21
1.6 3.2
C3H6/C3H8
(C3H6) (C3I-I6)
25 150
Hassan [95] ?
1 0.3
620 340
Wu [68]
2.7 --0.010 --0.1
300
Pfeed: 0.3-0.8 Pperm:small
300 35 150
C3H 8 (single)
1.5
50-100
Asaeda [64] d < 1 ~tm pfeed: 6 bar Pf = 2 bar
C3H 6 (single)
2.5
50-100
Pperm = 1 bar
respectively in combination with reasonable separation factors. High pressure data up to 20 bar for H 2 (de Lange et al.) indicate the possibility of very high permeation values. H 2 / N 2 mixtures are investigated by Kim and Gavalas [65] using Vycor glass supports with silica deposited partly in and partly on top of the support. At 500~ they report a separation value of c~ equal to 1000 and a permeation of 3.6x10 -3 mol m -2 s-1 bar -1 for H2. More recently Wu et al. [68] improved this method and at 600~ reported (x = 12-16 with a permeation of 1 x 1 0 -2 mol m -2 s -1 bar -1 which is about one order of magnitude smaller than that reported by de Lange. This is partly explained by the rather thick plugs (2.0-2.5 ~tm) of silica, completely deposited within the support pores. Interesting results are reported by Hassan et al. [95] using hollow fibre silica with an estimated pore size of 1.3 nm. For O 2 / N 2 the separation factor c~ = 11.5 at 298 K and c~ = 4.6 at 423 K. Permeation data are not given.
420
9 ~ TRANSPORTAND SEPARATIONPROPERTIESOF MEMBRANESWITH GASESAND VAPOURS
De Lange (Table 9.14) reports a value of (~ = 2 at 473 K with a permeation of 2-6x10-2 mol m -2 s -1 bar -1, which is reasonable in agreement with Hassan's data considering the strong decrease of c~with increasing temperature. These results indicate that interesting combinations of flux and separation factor in air separation can be obtained with silica membranes. A similar conclusion can be drawn for separation of saturated-unsaturated hydrocarbons as shown by Asaeda et al. [64] for propane-propene (see Table 9.14). In this case permeation values cannot be calculated from the flux values due to non-linear behaviour of the flux as function of pressure. Finally Rao and Sircar [42] report data for microporous carbon layers (thickness 2.0-2.5 ~tm) deposited on carbon supports. For C4H10/H2 mixtures, 0~ equals 94 (at 295 K) which is much larger than the permselectivity (for explanation see Sections 9.4.2 and 9.4.3). The permeability for C4H10 is reported to be 112 Barrer. The author of this chapter recalculated this value to a permeation of 1.4x10 -3 mol m -2 s -1 bar -1. It is, however, very questionable to assume a linear TABLE 9.15a Overview and separation data of typical supported microporous zeolite (MFI) membranes Gas
o~
Flux (10-2 tool m -2 s -1)
Temp. (~
Pressure/ thickness 50 kPa" 50 kPa
H2/CO 2
1-2
H2: CO2:
2-3 1.7
200-350
d=50 ~tm (Bakker) stainless support
1.25
CO2: H2:
0.1 0.12
200
d=3.0 ~tm (Vroon) A1203 support
CO2/H 2
10
H2: CO2:
0.18 1.8
25
d=50 ~tm (Bakker) [71]
14
CO2:
0.15--0.5
25
d=3.0 ~tm
H2/CH 4
low 1.9
CH4: CH4:
? 0.5-0.7
200 200
Bakker [71] Vroon [72-74]
H2/n-butane
2.5 1.0 >100
n-butane: n-butane: n-butane:
1.0 1.5 0.5
350 200 25
Bakker Bakker Bakker
n/i-butane
50 27
n-butane: ?
0.2
25 25
Vroon Bakker
O2/N2 p / o xylene
1 1 =60 25
02: p-xylene 3.5x10 --6 3.5x10 -6
0.5-0.35
25-200 25 100-150 200
Vroon Vroon [72-74] 0.36:0.26 kPa 0.36:0.26 kPa
25 200
Vroon 4.6:4.6 kPa 4.6:4.6 kPa
n-butane/H2
Benzene / cyclohex. 5 4.5
2.6x10 -7 18x10 -7 (benzene)
9 ~ TRANSPORTANDSEPARATIONPROPERTIESOFMEMBRANESWITHGASESANDVAPOURS
421
TABLE 19.15b Overview of typical flux and separation data of supported zeolite (MFI) membranes Material
0~
Flux (10-2 tool m -2 s-1)
Temp. (~
p / m xylene (triisopropyl benzene)
15
35 ml m -2 h* (m-xylene)
25
n-butane i-butane
-
5.5** 4.5
25 25
CH3OH/CH4
190
1.7 (CH3OH) 0.001 (CI-~
100
CH3OH/CH 4
29
1.8 (CH3OH) 0.007 (CH4)
100
n/i-butane
3-6
?
?
Pressure / thickness Xiang [76]: Pfeed" 17 bar Pperm: 10 Torr ratio 1:1:1 Pfeed' I bar Pperrn' zero Jia [27,97]: Ptotal:1100kPa Pfeed(CH3OH): 440-220 Ptotal:1500kPa Pfeed (CH3OH): 400-165 kPa ? thickness: 10.0 ~tm
*Not clear whether this is ml liquid or gas. **Calculated from given permeance assuming linear relation with partial pressure (=40x10-2. relationship b e t w e e n the flux a n d the p r e s s u r e for b u t a n e at 295 K. In other m i c r o p o r o u s s y s t e m s such as silica a n d zeolites this is not the case. In a n o t h e r p a p e r Rao a n d Sircar [Gas Separation and Purification (1993) 279-284] r e p o r t e d for C O 2 / H 2 m i x t u r e s at 296 K, c~ = 5 (Phigh = 2.36 bar) to 0~ = 20 (Phigh = 3.7 bar) w i t h a p e r m e a b i l i t y of 1200 Barrer in b o t h cases. This indicates an increase in the s e p a r a t i o n factor w i t h p r e s s u r e , w h e r e a s the p e r m e a t i o n r e m a i n s constant. Typical results for zeolite (MFS) m e m b r a n e s are collected in Table 9.15 a n d p a r t l y d i s c u s s e d in Section 9.4.3. As is s h o w n in the table the s e p a r a t i o n factor of m i x t u r e s of w e a k l y a n d s t r o n g l y a d s o r b i n g gases (see Section 9.4.3) s h o w s a c o n v e r s i o n as a function of t e m p e r a t u r e . Interesting s e p a r a t i o n values can be o b t a i n e d for C O 2 / H 2 a n d n - b u t a n e / H 2 m i x t u r e s at low t e m p e r a t u r e . These are in the s a m e r a n g e as those o b t a i n e d in carbon m e m b r a n e s ; for the flux v a l u e s a similar conclusion holds. I s o m e r s e p a r a t i o n is d e m o n s t r a t e d b y several a u t h o r s (see Table 9.15). G o o d s e p a r a t i o n factors (27-50) are r e p o r t e d for m i x t u r e s of n- a n d i s o b u t a n e b y V r o o n et al. [72,74] a n d Bakker et al. [71] with, h o w e v e r , m o d e s t flux values. S e p a r a t i o n of para- f r o m ortho-xylene is r e p o r t e d b y V r o o n et al. [72,74] w i t h c~ equals 60 in the t e m p e r a t u r e r a n g e of 100-150~ a n d (x = 25 at 200~ a n d a flux of 3.5x10 -6 m o l / m -2 s q for the fastest p e r m e a t i n g p-xylene (100-150~ with h o w e v e r a v e r y small d r i v i n g force.
422
9 - - TRANSPORT AND SEPARATION PROPERTIESOF MEMBRANES WITH GASES AND VAPOURS
The partial pressure at feed side (high pressure side) is only 0.36 kPa. Using higher partial pressures and increasing the temperature might bring the flux in the range 10-3-10-4 m o l / m -1 s-1. Xiang and Ma [76] reported results for mixtures of para- and meta xylene. At room temperature the value of (x equals 15 with a flux for m-xylene of 35 c m 3 m -2 h -1. Assuming that the permeating gas volume is expressed a s c m 3 gas under standard conditions (this is not defined) this permeation value corresponds to -- 8x10-s mol m -2 s -1 for m-xylene and with 3x10 -7 mol m -2 s-1 for p-xylene which is one order of magnitude smaller compared to values reported by Vroon et al.
9.6.3 Comparison of Permeation and Separation Data of Porous and Dense Membranes Typical data for dense membranes are collected in Table 9.16. A full discussion of these data is outside the scope of this chapter. Using permeation values the reader should be aware of the fact that the pressure dependence of the flux is usually strongly non-linear, but takes the form of a power law with values for the exponent around 0.5. This makes direct comparison on the basis of permeance or permeability not meaningful. Furthermore, the permeation value is limited by surface reactions with a critical thickness varying between 0.1 and 2 mm depending on material and condition. Finally, dense (i.e. non-porous) membranes permeate 02 o r H 2 o n l y and so are important only in applications where these gases play a role such as in air TABLE 9.16a C o m p a r i s o n of typical flux data of microporous and d e n s e m e m b r a n e s Hydrogen
Permeation (mol m -2 s-1)
Temp. (~
SiO2 a m o r p h o u s silica (measured)
6-20x10 -2
25-250
Calculated
>300x10 -2
Zeolite (silicalite) on steel
1-3x10 -2
100--400
Bakker et al. [71] Thickness 50 ~tm
Zeolite (silicalite) on alumina
0.5-0.85x10 -2
25-250
Vroon et al. [72-74] thickness 3-4 ~tm AP = 1 atm. ( 1 ~ 0 )
Pd resp. P d / A g films on alumina
3.0-4.5x10 -2
400-900
A r m o r [115]: AP = 2 bar; H2 thickness: 4.5; resp. 22 ~tm
Pd film
0.1x10 -2
100
N a g a m o t o [116] AP = I bar H 2
Pd film within pores of 0r
10-40x10 -2 0c > 1000 H2/N2
300
Y a n / M o r o o k a [113]; AP = 1 bar H 2 thickness 2 p m
AP = 1 bar (1---~0) AP> 1 5 b a r
9 ~ TRANSPORTAND SEPARATIONPROPERTIESOF MEMBRANESWITHGASESAND VAPOURS
423
TABLE 9.16b Comparison of typical flux data of microporous and dense membranes Oxygen
Flux (mol m -2 s-1)
Temp. (~
La0.6Sr0.4Co3_~
4.0x10-2
900
Teraok~ [110]
La0.3Sr0.?Co3_8
0.3-0.4x10 -2
900
v. Doom/Bouwmeester [119] Thickness I mm air vs. 10-2 bar
Y0.05BaCo0.9bO3.6
0.4x10- 2
900
Brinkman et al. [118] air vs. 10-2 bar
La0.2Sr0.8Fe0.6Co0.403-8
0.2x10-2
850
Balachandral [117] Thickness 0.25-1.2 mm air vs. CH4/H2 (4:1)
ZY-Pd (40 vol%)
0.1-0.2x10 -2 0.2-0.5x10 -3
1100 900
0.6x10-2
1100
Chen et al. [112] air vs. 10-2 bar Thickness 0.5 mm air vs. CO/CO2
BiEr-Au (40 vol%) BiEr-Ag (40 vol%) BiEr-Ag (40 vol%)
0.68x10-3 0.17x10-2 0.85x10-3
850 850 750
Chen et al. [112] Thickness 1-1.5 mm air vs. --2x10-2bar
BiY-Ag (35 vol%)
1.0x10-2
750
Shen et al. [111] Thickness 90 ~tm air vs 6x10-5 bar
si02 microporous film on alumina
2.0-5.0x10 -2
35-200
de Lange [59--63] thickness 100 nm AP = I bar
s e p a r a t i o n a n d d e h y d r o g e n a t i o n or p a r t i a l o x i d a t i o n r e a c t i o n s in m e m b r a n e reactors. A s is s h o w n in T a b l e 9.16 a n d b y c o m p a r i s o n of T a b l e 9.16 w i t h T a b l e 9.14 t h e v a l u e s of o b t a i n a b l e s e p a r a t i o n f a c t o r s of m i c r o p o r o u s m e m b r a n e s is m u c h l o w e r t h a n t h o s e o b t a i n e d w i t h d e n s e m e m b r a n e s ( w h i c h s h o u l d b e i n f i n i t e in t h e case of c o m p l e t e l y d e f e c t - f r e e d e n s e m e m b r a n e s ) . V e r y r e c e n t l y o x y g e n p e r m e a t i o n v a l u e s r e p o r t e d b y S h e n et al. [111], C h e n et al. [112] a n d T e r a o k a et al. [110] s h o w t h a t t h e o b t a i n a b l e flux v a l u e s at high
424
9 - - TRANSPORT A N D S E P A R A T I O N PROPERTIES OF MEMBRANES W I T H GASES A N D V A P O U R S
temperature (>600~ are at least a factor of 5-10 lower than those obtainable with microporous membranes at ambient or somewhat increased temperature (200~ For hydrogen a similar situation exists, except for the results reported by Yan and Morooka [113]. In this case the flux data are comparable with those obtained by de Lange et al. but with c~> 1000 for H2/N2 mixtures.
9.7 CONCLUSIONS AND EVALUATION A general description of gas transport properties of inorganic membranes with complex architecture and for multicomponent gas mixtures is not yet available. Quantitative descriptions based on phenomenological (thermodynamic) equations a n d / o r microscopic models can be given in a number of limiting cases like single gases or binary gas mixtures and single wall, unsupported membranes or small plate shaped, asymmetric supported ceramic membranes. In the latter case the support properties are important and must be taken into account in the description of the membrane system and of the separating top layer. In mesoporous membranes the maximum obtainable separation factor for non-condensable gases is limited to the Knudsen separation factor. For adsorbing gases below their critical point, surface flow can play an important role and high values of the permeation and of the separation factor can be obtained in some cases up to temperatures of 300~ In the case of macro- and mesoporous supports their flow resistance should be as small as possible. If the transport resistance is not negligible corrections must be applied in the study of the separation properties of the separating layers. It is shown that even small pressure gradients across the support can cause a considerable decrease of the permeation and of the separation factor of the top layer, especially in the case of adsorbing gases. The absolute value of the permeate pressure is important in addition to the pressure ratio of feed and permeate streams. Increasing support resistance causes an increase of the permeate pressure on the interface between support and separation (top) layer in the case of supported membranes. High separation factors can be obtained with microporous membranes with a pore diameter smaller than 2 nm and are realised with carbon, silica and zeolite membrane systems. The description of these systems is still in its infancy. In some cases reasonable agreement is obtained between calculated and measured permeation and separation properties. Permeation values of a single gas and of that gas in a mixture are generally different and so the separation factor of binary mixtures and the permselectivity (ratio of single gas permeation values) is also different.
9 - - TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
425
The permeation of a gas is strongly affected by the sorption properties of the combination of gas and membrane and by the ratio of the molecular diameter of the gas molecule and the pore diameter. Mixtures must be classified on the basis of these two properties and the transport properties of these classes differ considerably. The highest separation factors are obtained in the case of: (i) mixtures of strongly (S) and weakly (W) adsorbing gases at intermediate temperature and pressure values, and (ii) the size exclusion regime; here one of the gases in the mixture has a molecular diameter which is larger than the pore diameter. Typical values for permeation and separation factors of microporous membranes are given in Tables 9.14 and 9.15. A comparison is also made with dense membranes in Table 9.16.
LIST OF SYMBOLS
A A* b B0 B* sat
Ci C dp
D D* Ed
;Co: F Fp
gs
J k K
Kn L M Mx,M* FIk
Surface area (m 2) Coefficient in Eq. (9.34). Subscript o: per mol. absorbed Henry constant (mol/kg Pa) Permeation coefficient (m) Coefficient in Eq. (9.34). Subscript o: per mol. absorbed Saturation concentration in material (mol/kg or mol / m 3) Concentration. Subscripts: s, surface; sat, saturated Pore diameter Diffusion coefficient (m2/s) Pre-exponential coeff, in Arrhenius equation Activation energy for diffusion Coefficient in Eq. (9.34) Permeation (mol/m2 s Pa) Permeability (mol m / m 2 s Pa) Fitting parameter in Eq. (9.65) Molar flux (mol/m 2 s). Subscripts" v, viscous; k, Knudsen; c, capillary condensation Correction term in Eq. (9.41b) Langmuir constant (Pa -1) Knudsen number Thickness (m) Molecular mass or molecular weight (kg/mol) Eq. (9.34g-h)) Fitting parameter in Eq. (9.34a) (m -2)
426
9 - - TRANSPORT AND SEPARATION PROPERTIES OF MEMBRANES WITH GASES AND VAPOURS
P P Pd Pr q Q Qa AQa Y Ra
R Sc
t U V
Vm x Z Zwall
Partial pressure (Pa) Total pressure (Pa) Dimensionless pressure: P/Pref Ratio Plow/Phigh Amount absorbed gas (mol/kg) or (mol/m 3) Molar flow (tool/s) Heat of absorbtion (kJ/mol) Activation energy for surface diffusion (kJ/mol) Pore radius (m) or particle radius (m) Rate of adsorption Eq. (9.65) Gas constant (8.314 J/mol K) Stage cut (Qp/Qf) Sticking factor, defined by Eq. (9.65) Potential energy (kJ/mol). Subscript r: relative Molecular velocit~ (m/s) Molar volume (m/mol) Mol. fraction Distance coordinate (m) Number of molecular collisions with the walls of a volume ( c m - 3 S -1 )
Greek letters
F
0
Ok
V
Gs
Separation factor. Subscript 0: ideal separation factor (Eqs. (9.36) or (9.38)) Fitting parameter in Eq. (9.34b) (-) Fitting parameter in Eq. (9.68) Affinity coefficient in Eq. (9.56) (J/mol) Thermodynamic factor (-), defined in Eq. (9.40) Porosity (-) Dynamic viscosity (Pa s) Occupancy (c/cs) (-) Reflection factor in Eqs. (9.6) and (9.9) Molecular mean free path length (m) Geometric constant of pore structure (-) Jump probability, Eq. (9.57) Collision diameter (m 2) Surface tension (J/ m 2) Tortuosity (-) Volume flow (m3/s) or mol. flow (mol/s) Contact angle (-)
9 ~ TRANSPORTANDSEPARATIONPROPERTIESOFMEMBRANESWITHGASESANDVAPOURS
427
REFERENCES
1. R.E.Cunningham and R.J.J. Williams, Diffusion in Gases and Porous Media. Plenum Press, New York 1980. 2. F.A.L. Dullien, Porous Media: Fluid Transport and Pore Structure. Academic Press, New York, 1979. 3. J. Karger and D.M. Ruthven, Diffusion in Zeolites and Other Microporous Solids. Wiley, New York 1992. 4. J.O. Hirschfelder, C.F. Kurtiss and R.B. Bird, Molecular Theory of Ga~es and Liquids. Wiley, New York 1965. 5. E.T. Turkdogan, Physical Chemistry of High Temperature Technology. Academic Press, New York, 1980. 6. R.R. Bhave (Ed.), Inorganic Membranes, Synthesis, Characteristics and Applications. Van Nostrand Reinhold, New York, 1990. 7. E. Wicke and R. Kallenbach, Koltoid Z., 97 (1941) 34-41. 8. J. Rouquerol, D. Avnir, C.W. Fairbridge, D.H. Everett, J.H. Haynes, N. Pernicone, J.D.F. Ramsay, K.S.W. Sing and K.K. Unger, Recommendations for the characterisation of porous solids, Pure Appl. Chem.,66 (1994) 1739-1758. 9. A.J. Burggraaf and K. Keizer, Synthesis of inorganic membranes, in: R.R. Bhave (Ed.), Inorganic Membranes: Synthesis, Characteristics and Applications. Van Nostrand Reinhold, New York, 1991, pp. 10-63. 10. R.B. Bird, W.E. Stewart and E.N. Lightfoot, Transport Phenomena. Wiley, New York 1960. 11. (a) E.A. Mason and A.P. Malinauskas, Gas Transport in Porous Media: The Dusty Gas Model. Chem. Eng. Mon., 17. Elsevier, Amsterdam, 1983. (b) R.B. Evans, G.M. Watson and E.A. Mason, Gaseous diffusion in porous media at uniform pressure, ]. Chem. Phys., 35 (1961) 2076-2083; Effect of pressure gradients, I. Chem. Phys., 36 (1962) 1894-1902. 12. E.A. Mason, A.P. Malinauskas and R.B. Evans, Flow and diffusion of gases in porous media. J. Chem. Phys., 46 (1967) 3199-3216. 13. H.P.G. Darcy, Les Fontaines Publiques de la Ville Dyon. Victor Dalmont, Paris, 1856. 14. J.L. Poiseuille, Comptes Rendus, 11 (1840) 961. 15. M. Knudsen, Ann. Physik (Leipzig), 28 (1909) 75. 16. R.W. Schofield, A.G. Fane and C.J.D. Fell, Gas and vapour transport through microporous membranes. 1, Knudsen-Poiseuille transition. J. Membr. Sci., 53 (1990) 159-171. 17. (a) A. Leenaars and A.J. Burggraaf, The preparation and characterisation of alumina membranes with ultra fine pores. Part 3: The permeability for pure liquids. J. Membr. Sci., 24 (1985) 245-260. (b) A. Leenaars and A.J. Burggraaf, Part 2: The formation of supported membranes. I. Colloid Interface Sci., 105 (1985) 27-40. 18. W. Eickmann and U. Werner, Gas separation using porous membranes. Ger. Chem. Eng., 8 (1985) 186-194. 19. J.C.S. Wu, D.F. Flowers and P.K.T. Liu, High temperature separation of binary gas mixtures using microporous membranes. ]. Membr. Sci., 77 (1993) 85-98. 20. K. Keizer, R.J.R. Uhlhorn, R.J. van Vuren and A.J. Burggraaf, Gas separation mechanisms in microporous modified ~'A1203membranes. ]. Membr. Sci., 39 (1988) 285-300. 21. (a) R.J.R. Uhlhorn, Ceramic membranes for gas separation. Ph.D. Thesis, University of Twente, Enschede, The Netherlands, 1990.
428
22. 23. 24.
25. 26. 27. 28. 29.
30. 31. 32. 33. 34. 35. 36. 37.
38.
39. 39a. 40.
9 ~ TRANSPORTANDSEPARATIONPROPERTIESOFMEMBRANESWITHGASESAND VAPOURS (b) R.J.R. Uhlhom and A.J. Burggraaf, Gas separation with inorganic membranes, in: R.R. Bhave (Ed.), Inorganic Membranes. Van Nostrand Reinhold, New York 1990, pp. 155-176. R.L. Rowell, S.A. Carrano, A.J. de Bethune and A.P. Malinauskas, Gas and vapor permeability. ]. Colloid Interface Sci., 37 (1971) 242-246. R.W. Tock and K. Kammermeijer, Temperature-separation factor relationship in gaseous diffusion. AIChE J., 15 (1969) 715-718. K. Haraya, Y. Shindo, T. Hakuta and H. Yoshitome, Separation of H2-CO2 mixtures with porous glass membranes in the intermediate flow region. J. Chem. Eng. Jpn., 19 (1986) 186-190. Y. Shindo, T. Hakuta, H. Yoshitome and H. Inoue, Separation of gases by means of a porous glass membrane at high temperature. ]. Chem. Eng. Jpn., 17 (1984) 650-652. M. Okazaki, H. Tamon and R. Toei, Interpretation of surface flow phenomena of adsorbed gases by a hopping model, AIChE ]., 27 (1981) 262-270. C. Bai, M. Dong Jia, J.L. Falconer and R.D. Noble, Preparation and separation properties of silicalite composite membranes. ]. Membr. Sc/., 105 (1995) 79-89. R.J.R. Uhlhom, K. Keizer and A.J. Burggraaf, Gas and surface diffusion in modified Talumina systems. ]. Membr. Sci., 46 (1989) 225-241. (a) H.J. Sloot, A non-permselective membrane reactor for catalytic gas phase reaction. Ph.D. Thesis, University of Twente, Enschede, The Netherlands, 1991. (b) H.J. Sloot, C.A. Smolders, W.P.M. van Swaaij and G.P. Versteeg, Surface diffusion of hydrogen sulfide and sulfur dioxide in alumina membranes in the continum regime. J. Membr. Sci., 74 (1992) 263-278. E.A. Flood and M. Huber, Can. J. Chem., 33 (1955) 203. E.R. Gilliland, R.F. Baddour and J.L. Russel, Rates of flow through microporous membranes. AIChE J., 4 (1958) 90-96. E.R. Gilliland, R.F. Baddour, G.P. Perkinson and K.J. Sladek, Diffusion on surfaces. Ind. Eng. Fundam., 13 (1974) 100-105. H. Tamon, S. Kyotani, H. Wadu, M. Okazaki and R. Toei, Surface flow phenomenon of adsorbed gases on activated ~-alumina. ]. Chem. Eng. Jpn., 14 (1981) 136-141. H. Tamon, M. Okazaki and R. Toei, Flow mechanism of adsorbate through porous media in capillary condensation. AIChE J., 27 (1981) 271-277. K.H. Lee and S.T. Hwang, The transport of condensable vapors through a microporous Vycor glass membrane. ]. Colloid Interface Sci., 110 (1986) 544-555. H. Rhim and S.T. Hwang, Transport of capillary condensate. J. Colloid Interface Sci., 52 (1975) 174-181. R.J.R.Uhlhom, K. Keizer and A.J. Burggraaf, Gas transport and separation with ceramic membranes. Part 1: Multilayer diffusion and capillary condensation. J. Membr. Sci., 66 (1992) 259-269. M. Asaeda and L.D. Du, Separation of alcohol/water gaseous mixtures by thin ceramic membranes. I. Chem. Eng. ]pn., 19 (1986) 72-77. (a) M. Asaeda, L.D. Du and M. Fuji, Separation of alcohol/water gaseous mixtures by an improved ceramic membrane. J. Chem. Eng. Jpn., 19 (1986) 84-86. D.P. Sperry, J.L. Falconer and R.D. Noble, Methanol-hydrogen separation by capillary condensation in inorganic membranes. J. Membr. Sci., 60 (1991) 185-193. B. Abeles, L.F. Chen, J.W. Johnson and J.M. Drake, Capillary condensation and surface flow in microporous Vycor glass. Israel J. Chem., 31 (1991) 106. T. Okubo and H. Inoue, Introduction of specific gas selectivity to porous glass mem-
9 -- TRANSPORTAND SEPARATIONPROPERTIESOF MEMBRANESWITHGASESAND VAPOURS
429
branes by treatment with tetra ethoxy silane. J. Membr. Sci., 42 (1989) 109-117. 41. T. Okubo and H. Inoue, Single gas permeation through porous glass modified with tetra ethoxysilane. AIChE ]., 35 (1989) 845-848. 42. M.B. Rao and S. Sircar, Nanoporous carbon membranes for separation of gas mixtures by selective surface flow. J. Membr. Sci., 85 (1993) 253-265. 43. L.D. Rothfeld, Gaseous conterdiffusion in catalyst pellets. AIChE J., 9 (1963) 19-24. 44. N. Wakao, J. Otani and J.M. Smith, Significance of pressure gradients in porous materials, 1. Diffusion and flow in fine capillaries; 2. Diffusion and flow in porous catalysts. AIChE l., 11 (1965) 435-439; 439-445. 45. D.S. Scott and F.A.L. Dullien, Diffusion of ideal gases in capillaries and porous solids. AIChE J., 8 (1962) 113-117, 293. 46. (a) J.W. Veldsink, A catalytically active, non permselective membrane reactor for kinetically fast, strongly exothermic heterogeneous reactions. Ph.D. Thesis, University of Twente, Enschede, The Netherlands, 1993. (b) J.W. Veldsink, G.F. Versteeg and W.P.M. van Swaaij, An experimental study of diffusion and convection of multicomponent gases through catalytically and non catalytically membranes. ]. Membr. Sci., 92 (1994) 275-291. 47. R. Taylor and R. Krishna, Multicomponent Mass Transfer. Wiley, New York, 1993. 48. R.D. Present and A.J. De Bethune, Separation of a gas mixture through a long tube at low pressure. Phys. Rev., 75 (1949) 1050-1057. 49. J.J.M. Oude Egberink, Gas transport in a porous medium. M.Sc. Thesis, Faculty of Chemical Technology, University of Twente, Enschede, The Netherlands, 1993. 50. K. Keizer, R.J.R. Uhlhorn and A.J. Burggraaf, Gas separation using inorganic membranes, in: R.D. Noble and S.A. Stern (Eds.), Membrane Separations Technology: Principles and Applications. Elsevier, Amsterdam, 1994, pp. 553-588. 51. R.J.R, Uhlhorn, V.T. Zaspalis, K. Keizer and A.J. Burggraaf, Synthesis of ceramic membranes. Part II: Modification of alumina thin films: reservoir method. J. Mater. Sci., 27 (1992) 538-552. 52. S. Kitao and M. Asaeda, Separation of organic acid-water mixtures by thin porous silica-membranes. ]. Chem. Eng. Jpn., 23 (1990) 367. 53. S. Kitao, M. Ishizaki and M. Asaeda, Permeation mechanisms of water through fine porous ceramic membranes for separation of organic solvent-water mixtures. Key Eng. Mater., 61 & 62 (1991) 175-180. 54. J.E. Koresh and A. Softer, The carbon molecular sieve membranes. General properties and the permeability of CH4/H2 mixtures. Sep. Sci. Technol., 22 (1987) 972-982. 55. V.M. Linkov, R.D. Sanderson and E.P, Jacobs, Preparation of inorganic hollow-fibre membrane and composite hollow-fibre carbon membranes, in: Y.H. Ma (Ed.), Proceed-
ings of 3rd International Conference on Inorganic Membranes (ICIM3), July 10-14, 1994, Worcester, MA, USA. Distributed by Worcester Polytechnic Institute, 100 Institute Rd. Worcester, MA 01609, USA. pp. 471-481. 56. A.B. Shelekhin, A.G. Dixon and Y.H. Ma, Adsorption, permeation and diffusion of gases in microporous membranes. II. Permeation of gases in microporous glass membranes. ]. Membr. Sci., 75 (1992) 233-244. 57. R.J.R.Uhlhorn, M.H.B.J. Huis in 't Veld, K. Keizer and A.J. Burggraaf, High permselectivities of microporous silica modified ~'alumina membranes. J. Mater. Sci. Lett., 8 (1989) 1135-1138. 58. R.J.R.Uhlhorn, K. Keizer and A.J. Burggraaf, Gas transport and separation with ceramic membranes. II. Synthesis and separation properties of microporous membranes.
430
9 -- TRANSPORTANDSEPARATIONPROPERTIESOFMEMBRANESWITHGASESAND VAPOURS
J. Membr. Sci., 66 (1992) 271-287. 59. R.S.A. de Lange, Microporous sol-gel derived ceramic membranes for gas separation synthesis, gas transport and separation properties. Ph.D. Thesis, 1993, University of Twente, Enschede, The Netherlands. 60. R.S.A. de Lange, J.H.A. Hekkink, K. Keizer and A.J. Burggraaf, Formation and characterisation of supported microporous ceramic membranes prepared by sol-gel modification techniques. J. Membr. Sci., 99 (1995) 57-75. 61. R.S.A.de Lange, J.H.A. Hekkink, K. Keizer and A.J. Burggraaf, Permeation and separation studies on microporous sol-gel modified ceramic membranes. Microporous Mater., 4 (1995) 169-186. 62. R.S.A. de Lange, J.H.A. Hekkink, K. Keizer and A.J. Burggraaf, Polymeric silica based sols for membrane modification applications sol-gel synthesis and characterisation with SAXS. ]. Non-Cryst. Solids, 191 (1995) 1-16. 63. R.S.A. de Lange, K. Keizer and A.J. Burggraaf, Analysis and theory of gas transport in microporous sol-gel derived ceramic membranes. J. Membr. Sci., 104 (1995) 81-100. 64. M. Asaeda, A. Yamamichi, M. Satoh and M. Kamakura, Preparation of porous silica membranes for separation of propylene/propane gaseous mixtures, in: Y.H. Ma (Ed.), Proceedings of 3rd International Conference on Inorganic Membranes (ICIM3), 10-14 July 1994, Worcester MA, USA. Distributed by Worcester Polytechnic Institute, 100 Institute Rd., Worcester, MA 01609, USA. pp. 315-325. 65. M. Tsapatsis and G. Gavalas, Structure and aging characteristics of H2 permselective SiO2-Vycor membranes. ]. Membr. Sci., 87 (1994) 281-296. 66. Heung Yong Ha, Suk Woo Nam, Seong Ahn Hong and Won Kook Lee, Chemical vapor deposition of hydrogen permselective silica films on porous glass supports from tetra ethyl orthosilicate. J. Membr. Sci., 85 (1993) 279-291. 67. C.L. Lin, D.L. Flowers and P.K.T. Liu, Characterisation of ceramic membranes II. Modified commercial membranes with pore size under 40 ~. J. Membr. Sci., 92 (1994) 45-48. 68. J.C.S. Wu, H. Sabol, G.W. Smith, D.L. Flowers and P.K.T. Liu, Characterisation of hydrogen permselective microporous ceramic membranes. J. Membr. Sci., 96 (1994) 275-287. 69. E.R. Geus, Preparation and characterisation of composite inorganic zeolite membranes with molecular sieve properties. Ph.D Thesis, 1993, Technical University of Delft, Delft, The Netherlands. 70. E.R.Geus, H. van Bekkum, W.J.W. Bakker and J.A. Moulijn, High temperature stainless steel supported zeolite (MFI) membranes: preparation, module construction and permeation experiments. Microporous Mater., 1 (1993) 131-147. 71. W.J.W. Bakker, G. Zheng, F. Kapteyn, M. Makkee and J.A. Moulijn, Single and multicomponent transport through metal supported MFI zeolite membranes, in: M.P.C. Weynen and A.A.H. Drinkenburg (Eds.), Precision Process Technology. Kluwer, 1993, pp. 425-434. 72. Z.A.E.P. Vroon, Synthesis and transport properties of thin ceramic zeolite MFI membranes. Ph.D thesis, 1995, University of Twente, Enschede, The Netherlands. 73. Z.A.E.P. Vroon, K. Keizer, M.J. Gilde, H. Verweij and A.J. Burggraaf, Transport properties of alkanes through ceramic thin zeolite MFI membranes. J. Membr. Sci., 113 (1996) 293-300. 74. (a) Z.A.E.P. Vroon, K. Keizer and A.J. Burggraaf, Preparation and characterization of thin zeolite MFI membranes. J. Membr. Sci., in prep.
9 ~ TRANSPORTANDSEPARATIONPROPERTIESOFMEMBRANESWITHGASESANDVAPOURS
75.
76.
77. 78.
79.
80. 81. 82. 83. 84. 85. 86.
87.
88.
89.
90.
431
(b) Z.A.E.P. Vroon, K. Keizer, M.J. Gilde, H. Verweij and A.J. Burggraaf, Multi-component permeation through thin zeolite MFI membranes. Microporous Mater., in prep. (c) Z.A.E.P. Vroon, K. Keizer, M.J. Gilde, H. Verweij and A.J. Burggraaf, Single gas permeation through thin zeolite MFI membranes. Microporous Mater., in prep. E.R. Geus, M.J. den Exter and H. van Bekkum, Synthesis and characterisation of zeolite (MFI) membranes on porous ceramic supports. ]. Chem. Soc. Faraday Trans., 88 (1992) 3101-3109. Shouhe Xiang and Yi Hua Ma, Formation and characterisation of zeolite membranes from sols, in: Y.H. Ma (Ed.), Proceedings of 3rd International Conference on Inorganic Membranes (ICIM3), July 10-14, 1994, Worcester, MA, USA. Distributed by Worcester Polytechnic Institute, 100 Institute Rd. Worcester, MA 01609, USA. pp. 95-105. Meng Dong Jia, Baoshu Chen, R.D. Noble and J.L. Falconer, Separation of vapor/gas mixtures using ceramic zeolite composite membranes. J. Membr. Sci.,90 (1994) 1-10. M. Matsukata, N. Nishiyama and K. Keyama, Preparation of a thin zeolite membrane, in: J. Weitkamp, H.G. Karge, H. Pfeiffer and W. H61derich (Eds.), Zeolites and Related Microporous Materials: State of the Art 1994. Studies in Surface Science and Catalysis, Vol. 84, Elsevier, Amsterdam, 1994, pp. 1183-1190. A.J. Burggraaf, K. Keizer, R.S.A. de Lange, Z.A.E.P. Vroon and V.T. Zaspalis, Ceramic membranes for separation and reactions, in: R. von Ballmoos (Ed.), Proceedings of the 9th International Zeolite Conference, Montreal 1992. Butterworth-Heinemann/Reed, USA, 1993, pp. 47-70. D.W. Breck, Zeolite Molecular Sieves: Structure, Chemistry and Use. Wiley, New York, 1974. W.M. Meier and D.H. Olson, Atlas of Zeolite Structures, 3rd Edn., Butterworth Heinemann, 1992. H. van Bekkum, E.M. Flanigan and J.C. Jansen (Eds.), Introduction to Zeolite Science and Practice. Studies in Surface Science and Catalysis, Vol. 58, Elsevier, Amsterdam, 1991. J.H. Petropoulos and J.K. Petrou, Simulation of molecular transport in pores and pore networks. J. Chem. Soc. Faraday Trans., 87 (1991) 2017-2022. D.E. Everett and J.C. Powl, Adsorption in slit-like and cylindrical pores in the Henry's law region. J. Chem. Soc. Faraday Trans., 72 (1976) 619. R.M. Barrer, Porous crystal membranes. J. Chem. Soc. Faraday Trans., 86 (1990) 11231130. (a) Jirong Xiao and J. Wei, Diffusion mechanism of hydrocarbons in zeolites. I. Theory. Chem. Eng. Sci., 47 (1992) 1123-1141. (b) Jirong Xiao and J. Wei, Diffusion mechanism of hydrocarbons in zeolites. II. Analysis of experimental observations. Chem. Eng. Sci., 47 (1992) 1143-1159. (a) R. Krishna, A unified approach to modelling of intraparticle diffusion in adsorption processes. Gas Sep. Purif., 7 (1993) 91-104. (b) R. Krishna and J.A. Wesseling, Mass Transfer. Ellis Horwood, New York, 1990. F. Kapteyn, W.J.W. Bakker, Guhong Zheng and J.A. Moulijn, Temperature and occupancy dependent diffusion of n-butane through a silicalite 1 membrane. Microporous Mater., 3 (1994) 227-234. F. Kapteyn, W.J.W. Bakker, Guhong Zheng, J. Poppe and J.A. Moulijn, Permeation and separation of light hydrocarbons through a silicalite-1 membrane. Application of the generalised Maxwell Stefan equations. Chem. Eng. J., 57 (1995) 145-153. H. Abdul Rahman, M.A. Hasanain and K.F. Loughlin, Quaternary, ternary, binary and pure component sorption on zeolites. Ind. Eng. Chem. Res., 29 (1990) 1525.
432
9 ~ TRANSPORTANDSEPARATIONPROPERTIESOFMEMBRANESWITHGASESANDVAPOURS
91. E.J. Magim, A.T. Bell and D.N. Theodoru, Transport diffusivity of methane in silicalite from equilibrium and non-equilibrium situation. J. Phys. Chem., 97 (1993) 4173. 92. A.B. Shelekhin, A.G. Dixon and Y.H. Ma, Theory of gas diffusion and permeation in inorganic molecular sieve membranes. AIChE J., 41 (1995) 58-67. 93. R.S.A. de Lange, J.H.A. Hekkink, K. Keizer, A.J. Burggraaf and Y.H. Ma, Gas transport in microporous sol-gel derived ceramic membranes. II. Sorpfion studies. J. Porous Mater., 2 (1995) 141-149. 94. (a) D.T. Hayhurst and A.R. Paravar, Diffusion of C1-C5 normal paraffins in silicalite. Zeolites, 8 (1988) 27. (b) D.B. Shah, S. Chokchai and D.T. Hayhurst, Measurements of transport rates of C4 hydrocarbons across a single crystal silicalite membrane. J. Chem. Soc. Faraday Trans., 89 (1993) 3161-3167. 95. M. Hassan, J.D. Way, P.M. Thoen and A.C. Dillon, Separation of gas mixtures using hollow fiber silica membranes, in: Y.H. Ma (Ed.), Proceedings of 3rd International Conference on Inorganic Membranes (ICIM3), July 10-14, 1994, Worcester, MA, USA. Distributed by Worcester Polytechnic Institute, 100 Institute Rd. Worcester, MA 01609, USA. pp. 325-335. 96. Yi Hua Ma and Shouhe Xia, Formation of zeolite membranes from sols. Patent application WP191-04 dated 3 Dec. 1992. 97. Meng Dong Jia, B. Chen, R.D. Noble and J.L. Falconer, Ceramic zeolite composite membranes and their application for separation of vapor/gas mixtures. J. Membr. Sci., 90 (1994) 1-10. 98. Meng Dong Jia, K.V. Peinemann and R.D. Behling, Ceramic zeolite composite membrane: preparation, characterisation and gas permeation. J. Membr. Sci., 82 (1993) 15-26. 99. F. Kapteyn, W.J.W. Bakker, G. Zheng, J.A. Moulijn and H. van Bekkum, Permeation and separation behaviour of a silicalite (MFI) membrane, in: 1st International Workshop on
Catalytic Membranes, 26-28 September, 1994, Lyon, France. 100. R. Krishna and L.P.J. van der Broeke, The Maxwell-Stefan description of Mass transport across zeolite membranes. Chem. Eng. J., 57 (1995) 155-162. 101. D.M. Ruthven, Lecture: Diffusion in zeolites. 4th International Symposium European Science and Engineering Program on Advanced Separation Science and Engineering, 4-6 October 1994, Leuven, Belgium. 102. P. Uchytil, Gas permeation in ceramic membranes. Part 1, Theory and testing of ceramic membranes. J. Membr. Sci., 97 (1994) 139-144. 103. Y.S. Lin and A.J. Burggraaf, CVD of solid oxides in porous substrates for ceramic membrane modification. AIChE J., 9 (1992) 251-263. 104. Y.S. Lin and A.J. Burggraaf, Experimental studies in pore size change of porous ceramic membranes after modification. J. Membr. Sci., 79 (1993) 65-82. 105. R. Govind, Future prospects and application for metal based membranes. Key Eng. Mater., 91/92 (1991) 319. 106. Huiming Deng, Minyan Zhou and B. Abeles, Diffusion reaction in mixed-electronic solid oxide membranes with porous electrodes. Solid State Ionics, 74 (1994) 75-84. 107. B.N. Nair, Internal Report, Lab. of Inorganic Materials, Faculty of Chem. Technology, University of Twente, Enschede, The Netherlands, 1994. 108. A.J. Burggraaf, Key Points in understanding and development of ceramic membranes" in: Y.H. Ma (Ed.), Proceedings of 3rd International Conference on Inorganic Membranes (ICIM3), July 10-14, 1994, Worcester, MA, USA. Distributed by Worcester Polytechnic Institute, 100 Institute Rd., Worcester, MA 01609, USA. pp. 1-17.
9 ~ TRANSPORTANDSEPARATIONPROPERTIESOFMEMBRANESWITHGASESANDVAPOURS
433
109. T.L. Ward, Assessment of inorganic membrane technology for petrochemical applications, in: Y.H. Ma (Ed.), Proceedings of 3rd International Conference on Inorganic Membranes (ICIM3), July 10-14, 1994, Worcester, MA, USA. Distributed by Worcester Polytechnic Institute, 100 Institute Rd. Worcester, MA 01609, USA. 110. Y. Teraoka, T. Nobunaga and N. Yamazoe, Effect of cation substitute on the oxygen semipermeability of perovskite type oxides. Chem. Lett., (1988) 503-506. 111. Y. Shen, M. Liu, D. Taylor, S. Balagopal and A. Joshi, Mixed ionic electronic conductors based on Bi-Y-O-Ag metal-ceramic system, in: T.A. Ramanarayan et al. (Eds.), Proceedings of the 2nd International Symposium on Ionic and Mixed Conductors. Proceedings Vol. 94.12, The Electrochemical Society Inc., pp. 574-583. 112. (a) Chu-sheng Chen, Fine grained zirconia-metal dual phase composites, oxygen permeation and electrical properties. Ph.D Thesis, University of Twente, Enschede, The Netherlands, 1994. (b) C.S. Chen, H. Kruidhof, H.J. Bouwmeester, H. Verweij and A.J. Burggraaf, Oxygen permeation through erbia stabilized bismuth oxide noble metal composites. Solid State Ionics, submitted. 113. S. Yan, H. Maeda, K. Kusakabe and S. Morooka, Thin palladium membrane formed in support pores by metal-organic chemical vapor deposition and application to hydrogen separation. L&E.C. Res., 33 (1994) 616-622. 114. A. Sengupta and K.K. Sirkar, Analysis and design of membrane permeators for gas separation", in: R.D. Noble and S.A. Stern (Eds.), Membrane Separations Technology, Principals and Applications. Elsevier, Amsterdam, 1995, pp. 449-550. 115. J.N. Armor, Challenges in membrane catalysis. Chemtech, 22 (1992) 557-563. 116. H. Nagamoto and H. Inoue, A hydrogen reactor with catalytic membrane. Chem. Eng. Commun., 34 (1985) 315-323. 117. U. Balachandral, J.T. Dusek, S.M. Sweeney, R.B. Poeppel, R.L. Mieville, P.S. Maiya, M.S. Kleefish, S. Pei, T.P. Kobylinski and A.C. Bose, Methane to syngas via ceramic membranes. Am. Ceram. Soc. Bull., 74 (1995) 71-75. 118. H.W. Brinkman, H. Kruidhof and A.J. Burggraaf, Mixed conducting yttrium-bariumcobaltoxide for high oxygen permeation. Solid State Ionics, 68 (1994) 173-176. 119. R. van Doorn, PhD Thesis, University of Twente, Enschede, The Netherlands, 1996.
This Page Intentionally Left Blank
Fundamentals of Inorganic Membrane Science and Technology Edited by A.J. Burggraaf and L. Cot 9 1996, Elsevier Science B.V. All rights reserved
Chapter 10
Dense ceramic membranes for oxygen separation H.J.M. Bouwmeester and A.J. Burggraaf Laboratory for Inorganic Materials Science, Faculty of Chemical Technology, University of Twente, P.O. Box 217, 7500 AE Enschede, The Netherlands
10.1 INTRODUCTION Dense ceramic membranes exhibiting high oxygen ionic and electronic conductivity have become of great interest as a potentially economical, clean and efficient means of producing oxygen by separation from air or other oxygencontaining gas mixtures. In addition to infinite permselectivity, notably high oxygen flux values are measured through selected mixed-conducting oxides with the perovskite structure. These may be in the range exhibited by microporous membranes, albeit that sufficiently high temperatures are required, typically above about 700~ It is generally accepted that, provided they can be developed with sufficient durability and reliability, mixed-conducting oxide membranes have great potential to meet the needs of many segments of the oxygen market. It is further expected that the Oxygen fluxes can be improved by thin film deposition on a porous substrate preferably of the same material to avoid compatibility problems. The applications envisioned range from small-scale oxygen pumps for medical applications to large-scale usage in combustion processes, e.g. coal gasification [1-4]. As oxygen, but also nitrogen, ranks among the top five in the production of commodity chemicals in the United States [5] successful development of the mixed-conducting oxide membranes could thus have clear economic benefits, at the expense of market share from more traditional supply options. Whilst the targeted membranes will be most competitive at small and intermediate scale level in which flexibility of operation is desired, they may
436
10 n DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
eventually challenge the present commercial status of cryogenics, pressureswing adsorption (PSA) and polymeric membranes [1-4]. Another application of mixed-conducting oxide membranes is to be found in the field of chemical processing, including the partial oxidation of light hydrocarbons, e.g. natural gas to value-added products such as ethane/ethene [6--11] and syngas [12-14], waste reduction and recovery [15]. The catalyst may be either the membrane surface itself or another material deposited in particulate form on top of the membrane. Besides the controlled supply or removal of oxygen to or from the side where the catalyst and the reactants are located, a promising feature is that the oxygen flux may alter the relative presence of different oxygen species (02,0-) on the catalyst surface, thereby providing species that may be more selective for partial oxidation reactions. This review addresses recent developments in the area of mixed ionic-electronic conducting membranes for oxygen separation, in which the membrane material is made dense, i.e. free of cracks and connected-through porosity, being susceptible only for oxygen ionic and electronic transport. Current work on different mixedconducting oxides is reviewed using concepts from electrochemistry and solidstate chemistry. Emphasis is on the defect chemistry, mass transport and the associated surface exchange kinetics, providing some basic background knowledge which aids further development of these materials into membranes for the aforementioned applications. There is no attempt to discuss inroads against competing technologies, or to speculate on new opportunities that may result from successful development. New developments in dense ceramic membrane research could offer very economical ways of separating hydrogen such as the proton-conducting ceramics or thin Pd-foils. These are not considered in this review. For a general discussion on the topical area of membrane technology and its impact in various applications the reader is referred to specific reviews, for example, see Refs. [16-20] and other chapters in this textbook. 10.2 GENERAL SURVEY
In this section, a brief overview is given of major membrane concepts and materials. Besides membranes made from a mixed ionic-electronic conductor (MIEC), other membranes incorporating an oxygen ion conductor are briefly discussed. Data from oxygen permeability measurements on selected membrane materials are presented.
10.2.1 Major Membrane Concepts In this chapter, a membrane is regarded as a barrier between two enclosures which preferentially allows one gas (i.e. oxygen) to permeate owing to the presence of a driving force such as a pressure or electric potential gradient.
10 m DENSE CERAMICMEMBRANESFOR OXYGENSEPARATION
437
.__ 02"_>
__ O2-.~ P 02,,
PO 2 '
*e
e
(b)
(a)
P02 "
P02 '
(c)
e
,#
02
(d)
Fig. 10.1. Different membrane concepts incorporating an oxygen ion conductor: (a) mixed conducting oxide, (b) solid electrolyte cell (oxygen pump), and (c) dual-phase membrane. Also shown is the schematics of an asymmetric porous membrane (d), consisting of a support, an intermediate
and a barrier layer havhlg a graded porosity across the membrane. The separation of oxygen using an MIEC membrane is schematically shown in Fig. 10.1a. The driving force for overall oxygen transport is the differential oxygen partial pressure applied across the membrane. As the MIEC membrane is dense and gas-tight, the direct passage of oxygen moleculesis blocked, yet oxygen ions migrate selectively through the membrane. Dissociation and ionization of oxygen occurs at the oxide surface at the high pressure side (feed side), where electrons are picked up from accessible (near-) surface electronic states. The flux of oxygen ions is charge compensated by a simultaneous flux of electronic charge carriers. Upon arrival at the low pressure side (permeate side), the individual oxygen ions part with their electrons and recombine again to form oxygen molecules, which are released in the permeate stream.
438
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
Mixed conduction also plays an important role in many other processes, e.g. in improving electrode kinetics and catalytic behaviour [21]. In fact, all oxides exhibit to some degree mixed ionic and electronic conduction, and selective oxygen permeation has been reported even for dense sintered alumina above 1500~ [22,23]. Although it is common to speak of mixed conduction when the total conductivity is provided by near equal fractions (transference numbers) of the partial ionic and electronic conductivity, respectively [24], from the point of view of oxygen permeation it is more useful to relate mixed conduction to their absolute values. Volume diffusion theories treating ambipolar transport in oxides clearly indicate that higher currents (fluxes) are obtained when either the electronic or the ionic conductivity increases, or both increase simultaneously. The flux at a given total conductivity is maximum when the ionic and electronic transference numbers are equal, i.e. 0.5. In this view, alumina is not a good mixed conductor. Materials showing predominant electronic conduction may thus prove to be excellent mixed conductors when their ionic conductivity is also substantial. The general objective for optimum membrane performance therefore is to maximize the product of mobility and concentration of both ionic and electronic charge carriers in appropriate ranges of temperature and oxygen partial pressure. Owing to the ability to conduct both oxygen ions and electrons, the MIEC membrane can operate without the need of attachment of electrodes to the oxide surface and external circuitry. The latter represents an inherent advantage over traditional oxygen pumps in which a solid oxide electrolyte is sandwiched between two gas-permeable electrically conductive electrodes (Fig. 10.1b). An advantage of electrically-driven oxygen separation may be its ability to deliver oxygen at elevated pressures, eliminating the need for compressors [25]. Figure 10.1c shows a dual-phase membrane, which can be visualized as being a dispersion of a metallic phase into an oxygen ion conducting host or matrix, e.g, Pd metal into stabilized zirconia. This challenging approach was first described by Mazanec et al. [26] and offers an alternative use of oxide electrolytes in the field of dense ceramic membranes. Industrially important solid oxide electrolytes to date are mainly based on oxygen-deficient fluorite-related structures such a s Z r O 2 and C e O 2 doped with CaO o r Y 2 0 3 . Unless operated with an internal or external circuitry, the oxygen flux through these materials in usual ranges of temperature and oxygen pressure is negligibly low, preventing their practical use as oxygen separation membrane. The existence of a non-vanishing electronic conduction in the ionic domain, and concomitant oxygen semipermeability, however, can be detrimental considering their use as solid electrolytes in fuel cells and oxygen sensors [27,28]. While past efforts were focused on expanding the electrolytic domain of oxygen ion conducting fluorite-type ceramics, more recently one has begun to introduce enhanced electronic conduction in fluorite matrices. Extrinsic elec-
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
439
tronic conduction in ionically conducting matrices can be obtained by dissolution of multivalent cations in the fluorite oxide lattice. Notable examples include yttria-stabilized zirconia doped with either titania [29,30] and ceria [31, 32]. Electronic conductivity in these solid solutions is reportedly found to increase with increasing dopant concentration, but may be limited by the solid solubility range of the multivalent oxide. As conduction occurs via a small polaron mechanism (electron hopping) between dopant ions of different valence charge, its magnitude will strongly vary with temperature and oxygen partial pressure. In general, the extent of mixed conductivity that can be induced in fluorite ceramics is limited, which restricts its possible use as ceramic membrane, unless very high temperatures of operation (> 1400~ and stability down to very low values of oxygen partial pressure are required as, e.g., in the production of gaseous fuels CO and H2 by direct thermal splitting of CO2 and H20, respectively, and extraction of the oxygen arising from dissociation [33]. Since the first report on high oxide ion conductivity in some of the rare earth aluminates in the mid sixties [34,35], materials with oxygen-deficient perovskite and perovskite-related structures receive much attention for the development of new solid electrolytes and mixed conductors for numerous applications [36]. Currently, extensive research is conducted on acceptor-doped perovskite oxides with the generic formula Lal_xAxCOl_yByO3_~(A = Sr, Ba, Ca and B = Fe, Cu, Ni). Teraoka et al. [37-39] were the first to report very high oxygen fluxes through the cobalt-rich compositions, which perovskites are known to become highly oxygen anion defective at elevated temperatures and reduced oxygen partial pressure. The oxygen-ion conductivity in the given series can be 1-2 orders of magnitude higher than those of the stabilizedzirconias, though in usual ranges of temperature and oxygen partial pressure electronic conduction in the perovskite remains predominant [39,40]. Besides potential use of these perovskite compositions as catalytically active electrodes in, e.g. fuel cells, oxygen pumps and sensors, the compounds have a bright future for use as oxygen separation membrane. The precise composition may be tailored for a specific application, but this has not yet been fully developed. Structural and chemical integrity of the cobaltites, however, is a serious problem and needs to be addressed before commercial exploitation becomes feasible. For the sake of completeness, a schematic representation of a porous ceramic membrane is given in Fig. 10.1d. The majority of porous ceramic membranes are composite or asymmetric in structure. They include materials like 0~-A1203, ~-A1203, TiO2 and SiO2, and generally consist of a thin layer of either a mesoporOUS (2 Z" ~x >~~~ ~~~ ~ ~
~
=o
o--~s
9
,
P
3
CERAMIC
z
o
~
,~~--~' ~ ~=~.
~ -:~:~:~ ~ ~
~
~~"
~
~~ >~.~ ~ . - ~ g~~
~~,~~ ~
~~zo~ >~ b xO.~ ~,
+
0
F
~9~
~
~'~
9
~
~
~
~
,~ ~
T
~
~ ~
!
6-~
m
"Q "~ t',.
!
.
"~ ~
.
.
I
.
~0
~-~ ~ -
72~ z ~
~
u.iu~ B-~O:~ 0~~
.-,,~,m SN
~
"~
~
"i
u~ ~
28. 29. 30. 31. 32.
_:,_:, ~ : ':-:u:.~ ~ ~~0 ~~ : u5 ~'~
24. 25. 26. 27.
H.W. Brinkman, H. Kruidhof and A.J. Burggraaf, Solid State lonics, 68 (1994) 173-176. H. Iwahara, T. Esaka and T. Mangahara, I. Appl. Electrochem., 18 (1988) 173-177. J. Fouletier, P. Fabry and M. Kleitz, I. Electrochem. Soc., 123(2) (1976) 204-213. H.J.M. Bouwmeester, H. Kruidhof, A.J. Burggraaf and P.J. Gellings, Solid State lonics, 53/56 (1992) 460-68. C.S. Chen, B.A. Boukamp, H.J.M. Bouwmeester, G.Z. Cao, H. Kruidhof, A.J.A. Winnubst and A.J. Burggraaf, Solid State loizics, 76 (1995) 23-28. C.S. Chen, PhD Thesis, University of Twente, The Netherlands, 1994. T.J. Mazanec, T.L. Cable and J.G. Frye, Solid State lonics, 53/56 (1992) 111-118. T.J. Mazanec, and J.G. Frye Jnr., Eirr. Patent Awl. 0399 833 A1 (1990). Y.S. Shen, M. Liu, D. Taylor, S. Bolagopal, A. Joshi and K. Krist, Mixed ionic-electronic conductors based on Bi-Y-0-Ag metal-ceramic system, in: T.A. Ramanarayanan, W.L. Worrell and H.L. Tuller (Eds.), Proceedings ofthe 2nd International Symposillm uti Ionic nizd Mixed C o n d i ~ t i q Ceramics, Vol. 94-12. The Electrochemical Society, Pennington, NJ, 1994, pp. 574-595. J.E. Ten Elshof, D.N.Q. Nguyen, H.J.M. Bouwmeester and H. Verweij, Solid State lonics, submitted. C.S. Chen, H.J.M. Bouwmeester, H. Venveij and A.J. Burggraaf, Solid State lonics, submitted. M. Dumelie, G. Nowogrocki and J.C. Boivin, J.C. Solid State lonics, 28/30 (1988) 524-528. I.C. Vinke, K. Seshan, B.A. Boukamp, K.J. d e Vries and A.J. Burggraaf, Solid State lonics, 34 (1989) 235-242. R.S.A. De Lange, J.H.A. Hekkink, K. Keizer and A.J. Burggraaf, Microporous Materials, 4 (1995) 169-186.
10 m D E N S E C E R A M I C M E M B R A N E S F O R O X Y G E N S E P A R A T I O N
3
19. 20. 21. 22. 23.
0
I
rn 0
5m
! n
B
E
s
xSrxCo1_yFeyO34, approach those exhibited by the porous membranes. It should be noted, however, that these types of membrane have different requirements. The high temperature needed for operation using membranes based on oxygen ion conductors may be restrictive in certain applications, but beneficial to others, e.g. coal gasification and partial oxidation of light paraffins [25].
10.2.3 Factors Controlling Oxygen Permeation The rate at which oxygen permeates through a non-porous ceramic membrane is essentially controlled by two factors, the rate of solid state diffusion within the membrane and that of interfacial oxygen exchange on either side of the membrane. The oxygen flux can be increased by reducing the thickness of the membrane, until its thickness becomes less than a characteristic value, Lc, at which point the flux of oxygen is under conditions of mixed control of the surface exchange kinetics and bulk diffusion [41]. Below Lc, the oxygen flux can only marginally be improved by making the membrane thinner. For predominant electronic conductors like, for example, the perovskites Lal_xSrxCOl_yFeyO3~, Lc is determined by the ratio of the oxygen self diffusivity and surface exchange coefficient. Both parameters can be measured simultaneously by using 180-160 isotopic exchange techniques. Calculations show that Lc may vary from the ~tm-range to the cm-range, depending on material and environmental parameters. Modelling studies, however, show that significant increase in the rate of interfacial oxygen transfer and, hence, in the oxygen flux can be achieved by deposition of a porous MIEC layer on top of the (thin) non-porous membrane [42--44]. Since a number of simplifying assumptions is made, such as neglect of changes in material parameters with variation in the chemical potential of oxygen, the models developed are valid only in the limit of small Po2-gradients across the MIEC membrane. For a more rigorous approach,
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
449
referring to actual operating conditions of oxygen separation membranes, much more work is needed to arrive at a better understanding of the transport processes under oxygen potential gradients. In particular, our present understanding of the factors that govern the surface exchange kinetics is rather poor. Effects related to microstructure, including grain boundary diffusion and (local) order-disorder phenomena, may also influence overall oxygen transport. Besides the processing into defect-free thin films and associated problems of compatibility between deposited membrane layer and the porous substrate material, chemical stability at high temperatures, effects induced by the presence of an oxygen potential gradient like segregation of impurities to the surface and to grain boundaries, kinetic demixing and kinetic decomposition could affect membrane performance or limit operational life. In many cases, these difficulties remain to be overcome before commercial exploitation becomes viable. All these factors are important and govern the selection of materials. In the following sections, the emphasis is on the basic elements of mixed ionic and electronic transport through dense ceramic membranes. Due to size considerations, we shall mainly focus this chapter to mixed-conducting acceptordoped perovskite and perovskite-related oxides. Other membrane concepts are also discussed, but only briefly. The examples chosen illustrate the fundamental factors determining the oxygen fluxes through dense ceramic membranes, which is the primary aim of this chapter. 10.3 F U N D A M E N T A L S
10.3.1 Bulk Transport The basic assumption of the theory presented in this section is that the lattice diffusion of oxygen or the transport of electronic charge carriers through the bulk oxide determines the rate of overall oxygen permeation. Moreover, oxygen is transported selectively through the membrane in the form of oxygen ions, rather than molecules, under the driving force of a gradient in oxygen chemical potential. The flux of oxygen ions is charge compensated by a simultaneous flux of electrons or electron holes, which is enabled without the use of external circuitry. We only briefly review the fundamentals of solid state diffusion through mixed conducting oxides and the reader is referred to Refs. [45-47] for a more complete discussion.
10.3.1.1 WagnerEquation Considered here is the case where the interaction of gaseous oxygen with the oxide lattice can be represented by a chemical reaction of the form 1 1 The notation adopted for defectsis from Kr6gerand Vink [48].
450
10 ~ DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
1 -~ 0 2 4- W6 4-
2 e'= O~)
(10.1)
a s s u m i n g that oxygen vacancies are the mobile ionic defects. These m a y be obtained, e.g., by d o p i n g of the oxide lattice with aliovalent cations. The intrinsic ionization across the b a n d g a p can be expressed by nil = e' + h.
(10.2)
The single particle flux of charge carriers, with neglect of cross terms b e t w e e n fluxes, is given by (3"k
jk =
z~k F2
Vnk
(10.3)
w h e r e Z k is the charge n u m b e r and C~kthe conductivity of charge carrier k, F the F a r a d a y constant and Vrlk the gradient of the electrochemical potential. The latter comprises a gradient in chemical potential V~tk a n d a gradient in electrical potential VO, for each individual charge carder k given by VT~k = V ~ k + Zk F
V~)
(10.4)
The charge carrier diffusing m o r e rapidly causes a gradient in the electrical potential V~, in which the transport of carriers with opposite charge is accelerated. At steady state, no charge accumulation occurs. The fluxes of ionic a n d electronic defects are therefore related to each other by the charge balance 2 j v 6 = je" - jh.
(10.5)
Equation 10.5 can be used together with Eqs. (10.3) a n d (10.4) to eliminate the electrostatic potential gradient. The flux of oxygen vacancies is then obtained in terms of the chemical potential gradients only. If it is further a s s u m e d that internal defect chemical reactions are locally not disturbed by the transport of matter, the chemical potential gradients of individual charge species can be converted into the virtual chemical potential of gaseous oxygen, ~to2. The following differential relations hold at equilibrium 2
2
It is tacitly assumed here that the chemical potential of lattice oxygen ~to~)is constant. The present formulation of the defect equilibrium for the formation and annihilation of oxygen vacancies and electrons by the reaction of the solid with environmental oxygen, however, is written in terms of the 'virtual' chemical potentials of the constituent structure elements. In so doing, one does not properly take into account the so-called site-exclusion effect, because the chemical potential of the oxygen vacancy 1/6 and that of lattice oxygen O~) cannot be defined independently from one another. In the present context, it suffices to say that the derived equations are in agreement with those obtained from a more rigorous thermodynamic treatment based upon the 'true' chemical potential for the building unit vacancy, i.e. (V6-O~). For further reading concerning the definition of chemical potentials, the reader may consult Refs. [45] and [46].
10 u DENSE CERAMICMEMBRANESFOR OXYGENSEPARATION 1
451
-~ V~to2 + V~tv6 + 2V~te, = 0
(10.6)
V~I,e, 4- V~h.---- 0
(10.7)
where ~tv6 denotes the chemical potential of the oxygen vacancy, ~te' and ~h. denoting the chemical potential of electrons and electron holes, respectively. The flux of oxygen through the membrane can be derived by combining Eqs. (10.3)-(10.7), using the relationship Jo2 = -1/2 jr6. One finds
J02
=-
1
~ ((~e'4- (~h.)(~V 6
J .........
42F 2 k ((~e' 4- (~h.) 4- O'V6 j
JV~t02
(10.8)
or in a more generalized form 1 (~elt~ion Jo2 = -- 42 F 2 V~to2 Gel 4- (~ion
(10.9)
where (~ion -- (~V 6 and (3"el ---- (~h. + (~e' are the partial ionic and electronic conductivity, respectively. The conductivity term in Eq. (10.9) is equivalent to teltiont~total = tiont~el = telt~ion, where tel and tion a r e the fractions (transference numbers) of the total conductivity (~totalprovided by electronic and ionic defects, respectively. Integration of Eq. (10.9) across the oxide membrane thickness, L, using the relationship V~to2 = ORT In Po2/Ox (x = distance coordinate) and assuming no divergence in the fluxes, yields the Wagner equation in the usual form In P"o2
RT Jo2 = - 2------42 f ~
(~elt~i~ d In Po2 ~
(10.10)
(~el + (~ion
lnP'02
The limits of integration are the oxygen partial pressures maintained at the gas phase boundaries. Equation (10.10) has general validity for mixed conductors. To carry the derivation further, one needs to consider the defect chemistry of a specific material system. When electronic conductivity prevails, Eqs. (10.9) and (10.10) can be recast through the use of the Nernst-Einstein equation in a form that includes the oxygen self-diffusion coefficient Ds, which is accessible from ionic conductivity measurements. This is further exemplified for perovskitetype oxides in Section 10.6.4, assuming a vacancy diffusion mechanism to hold in these materials.
10.3.1.2 Chemical diffusion coefficient The preceding theory was used by Wagner to describe oxide film growth on metals [49,50]. The driving force for diffusion is not a concentration gradient,
452
10 ~ DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
but rather a chemical potential gradient. An important and necessary assumption is that the internal defect reactions are fast enough to attain local chemical equilibrium so that the concentrations of involved ionic and electronic (electrons or holes) charge carriers at any distance coordinate in the oxide are fixed by the local value of the virtual chemical potential, ~to2. The effective transport is still that of neutral oxygen atoms by which the theory fits that of a chemical diffusion process in terms of Fick's first law 3co jo = - D 3x -
(10.11)
where the driving force for diffusion is the gradient in neutral oxygen, 3Co/3X. The coefficient of proportionality, denoted by D, is called the chemical diffusion coefficient. By virtue of Eqs. (10.9) and (10.11 ), one obtains ~_
1
(3"el(3"i~
3~t~
(10.12)
8 F 2 (~el + ($ion OCO
Here we note that Jo2 = 1/2jo. Because 3co/3X =-3Cv/3x, a similar expression is obtained when diffusions were dominated by neutral vacancies. The thermodynamic factor 3~o/3Co in Eq. (10.12) can be determined directly from experiment by measuring the oxygen stoichiometry as a function of oxygen partial pressure, either by gravimetric or coulometric measurements. In view of Eqs. (10.6) and (10.7), it comprises contributions from both ionic and electronic defects, which reflect their non-ideal behaviour. For materials with prevailing electronic conductivity Eq. (10.12) may be siml~lified to yield an exact relation between the chemical diffusion coefficient D and the oxygen tracer diffusion coefficient D*: D* D=~ HR
1,/23~to2
R T 3 In Co
(10.13)
Here Ha is the Haven ratio, defined as the ratio of the tracer diffusion coefficient D* to the quantity D ~ derived from dc ionic conductivity measurements, (~ion
Da = ~
RT
c o Z2 F 2
(10.14)
The Haven ratio may deviate from unity when correlation effects and possibly different jump distances and jump frequencies can not be neglected [51]. For a vacancy diffusion mechanism Ha equals the well-known tracer correlation factorf.
10 -- DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
453
10.3.1.3 Trapping of Electronic and Ionic Defects Equation (10.10) and those derived from it are valid as long as fully ionized oxygen defects contribute to transport. Different equations are obtained if valency changes of oxygen defects occur. Wagner [50] proposed to put the influence of reactions between ionic and electronic defect species in the cross terms of the Onsager equations. Maier [52-54] explicitly attributed individual diffusivities and conductivities to the new defect species, using the concept of a conservative ensemble accounting for free and trapped species. Following his approach, the reversible reaction between electrons and oxygen vacancies, Vo + e ' : Vo (10.15)
Vo + e': V ~ leads to the following expression for the oxygen flux,
Jo2 =
((~e' + (~h') ((~V6 + 4rYv6) + (~v6 r~v6 ] i [ I V,o2 (10.16) 42 F214s[vo~ + ((~e"+ (~h')+ ((~V6 + (~V6)
J
where we have adapted Eq. (33) in Ref. [53] (Part I) i n t o a form to be similar to Eq. (10.8), in which ionic transport is by doubly ionized oxygen vacancies only. The Onsager coefficient SWoaccounts for the contribution of neutral defects, enabling oxygen transport even when the electronic conductivity of the oxide is zero. We further note that the counter-diffusion of two Vo and a single Vo would result in a net neutral oxygen flux, as reflected by the last term in the numerator of Eq. (10.16). Maier [53] also examined the case in which electronic or ionic defects are associated (trapped) with immobile centres such as dopant ions. Trapping inevitably leads to a decrease in concentrations of the charge carriers available for transport. The impact of these phenomena is that the transport equations for evaluation of data obtained from electrochemical measurements like, for example, ionic conductivity, concentration cell, permeability and Hebb-Wagner polarization experiments should accordingly be modified. It is shown by Maier how these are influenced by trapping effects observed in perovskite SrTiO3, and by the transport properties of the high-temperature superconductor YBa2Cu306+x. Because of the large oxygen excess possible in the latter material it is assumed that transport occurs by differently ionized ionic defects, partly even by neutral oxygen species. For references, see the papers by Maier [52-54].
454
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
10.3.1.4 Empirical Equations Evaluation of j% from Eq. (10.10) requires that data exist for the partial conductivities r~io~and (3"elas a function of oxygen partial pressure between the limits of the integral. In what follows, some special relations for either prevailing electronic or ionic conduction are discussed. For the sake of approximation, in defect chemical studies often an empirical power law is used for the partial conductivity of the rate determining species, O'ir s u c h
asr
odP%) = r~~ P~2
(10.17)
where ~o is the conductivity at standard state. The value of n can be derived from experimental data of steady-state oxygen permeation. For proper evaluation it is necessary that the Po~-gradient across a specimen is varied within the assumed range of validity of the empirical power law. Inserting Eq. (10.17) in Eq. (10.10), one finds after integration, assuming ~i > P o 2 " .
In the range of temperatures (610-810~ and oxygen pressures (10-4 - 1 atm) covered by experiment, the concentration of minority charge carriers, i.e., electron-holes, in BE25 is proportional to P~2 with n = 1/4. However, the apparent value derived from experiment increases gradually from 1/4 to higher values upon decreasing specimen thickness from 0.285 cm to 200 gm, indicating permeation to be limited by two or more processes differing in order. The activation energy of the oxygen flux was found to increase too in the same direction. The observed behaviour can be attributed to the change-over from diffusion to
468
10 - - DENSE C E R A M I C M E M B R A N E S F O R O X Y G E N S E P A R A T I O N
surface control u p o n decreasing sample thickness. The experimental data can be fitted well by means of Eq. (10.37), though it is necessary to adapt the kinetic order of the surface reaction with respect to oxygen to a value of 5/8. The parameters c~ and 13obtained from numerical fitting appear to exhibit different activation energies; 136 + 4 kJ mole -] and 99 + 4 kJ mole -1, respectively, which indicates that the surface process is less limiting at higher temperatures. Isotopic exchange measurements on sintered dense discs of BE25 showed a P~2 dependence with m = 0.60 at 550~ and m = 0.54 at 700~ for the overall surface oxygen exchange rate [67,104]. Figure 10.6 shows that the value for the surface oxygen exchange rate j~x (= (~ Pc~2),normalized to air, obtained from the fit of the data agrees with that measured by isotopic exchange. The thickness, at which point the oxygen flux is half of that expected under conditions of pure diffusion-controlled kinetics, imposing opposite sides of discs to pure oxygen and helium gas, was calculated at 0.16 cm at 650~ and 0.09 cm at 800~ These values were found to be in good agreement with estimates of the parameter Lc as noted before in Section 10.3.2.2.
-6.50
=r~
opic exchange
-7.50
0
oxygen .
u.._.l
X,,,.~
o
-8.50 O
-9.50 0.90
1.00
1.10
1.20
1.30
IO00/T[K] Fig. 10.6. Data for the surface oxygen exchange rate, normalized to air, of 25 tool% erbia-stabilized bismuth oxide (BE25) from (a) isotopic exchange and (b) oxygen permeation measurements. Reprinted from Bouwmeester et al. [96].
10 m DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
469
10.4.3 Electrochemical Oxygen Separation
10.4.3.1 Oxygen Pump The open-cell emf generated across an oxygen concentration cell such as
O2(Po2' ), Pt I CSZ IPt, O2(Po2" )
(10.39)
with each side maintained at a different oxygen partial pressure Po2' and Po2" is given by, Eeq =
(1 -
_
RT
Po2'
tel ) - ~ - I n ~ Po2"
(10.40)
where tel is defined as a mean electronic transference number. In the absence of any electronic conduction, i.e. when tel = 0, Eq. (10.40) simplifies to the Nernst equation. When the cell arrangement delivers a current I under load conditions, the cell voltage drops below the value Eeq, due to ohmic losses IRi (Ri = electrolyte resistance) and polarization losses at both Pt electrodes. As an approximation,
E=
Eeq - I R i - n
(10.41)
where 1] represents the total cathodic and anodic polarization loss. Upon shortcircuiting both Pt-electrodes, the emf of the cell drops to zero while oxygen is transported from the high pressure side PO2' t o the!ow pressure side P O 2 " . By applying an external power source, the applied dc voltage can be used to enhance the magnitude of the current but also to reverse its sign. That is, oxygen may be pumped in both directions; the rate of transport equals I/4F according tO Faraday's law. This is the basic principle of electrochemical oxygen separation. An important phase during device development is optimization of the pumping rate, i.e. ohmic and polarization losses must be kept as low as possible. Much efforts have been concentrated on development, fabrication and testing of zirconia-based separators. For example, Clark et al. [105] has described the performance of a multi-stack yttria-stabilized zirconia (YSZ) based separator. Each cell contained a 125 ~tm thick YSZ layer of diameter 6.35 cm, whereas porous strontium-doped lanthanum manganite electrodes were used to eliminate the need for costly Pt. The largest of these separators, built with 20 cells, was found to be capable of an oxygen flux up to 1 1 min -1 at an operating temperature of 1000~ Factors influencing the efficiency of the oxygen separation process and systems analysis of conceptual oxygen production plants are also addressed. A major drawback of ZrO2-based materials is the high temperature required for operation, typically 900-1000~ expressing the need for development of oxide electrolytes which exhibit significant levels of ionic conduction at modest temperatures. Several alternative materials may be considered. To provide a
470
10 m DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
reference point for discussion, the ionic conductivity of YSZ is about 0.1 S c m -1 at 950~ This value is found in bismuth oxide stabilized with dopants such as E r 2 0 3 and Y203 and in cerium oxide doped with Gd203, Sm20 3 or Y20 3 already in the range 650-700~ [62,63] which electrolytes are less useful in, for example, fuel cells or sensor applications due to the presence of rather reducible ions Bi3+ and Ce 4+ and, hence, a non-negligible contribution of electronic conduction. The suitability of Bi0.571Pb0.42801.285 as electrolyte membrane has been proposed for temperatures as low as 600~ [106]. This material suffices however from structural instabilities. Having its mechanical properties enhanced by incorporating Z r O 2 into the starting material, the optimized membrane is able to operate continuously up to 300 mA c m 2 at 600~ Fast ionic conduction at modest temperatures has also been reported in Bi4V2_yCUyOll (BICUVOX) 3 [107-109], which phases possess an intergrowth structure consisting of Bi2 O2+ blocks alternating with perovskite blocks. The material Bi2V0.gCu0.105.35was found to exhibit an ionic conductivity of I x 10 -3 S c m -1 already at 240~ which is about two orders of magnitude higher than that of stabilized bismuth oxide [108]. In most cases the ability of these electrolytes for electrochemical oxygen separation has not yet been fully explored. Thus, it can not be excluded that relevant properties like, for example, oxygen ion conductivity, phase stability, gas tightness, mechanical strength and compatibility with electrode materials will not be affected during prolonged operation. Of course, the current-voltage characteristics and operational life are influenced not only by the quality of the solid electrolyte but also by the properties of the electrodes. For a recent review on oxygen electrode kinetics, see Ref. [64].
10.4.3.2 Dual-phase Composites As seen from Table 10.1 impressive oxygen fluxes have been reported through 25 mol% yttria-stabilized bismuth oxide (BY25) [110] and 25 mol% erbia-stabilized bismuth oxide (BE25) [111,112], which oxide electrolytes were rendered electronically conductive by dispersion with silver metal. A prerequisite is that both constituent phases in the composite membranes do form a continuous path for both ionic and electronic conduction, having their concentrations above the critical (percolation threshold) volume fraction ~)c.The latter quantity determines the minimum volume fraction in which conduction is possible and is a function of, for example, the relative dimensions and shape of the particles of both constituent phases [113]. In actual composite materials, 3
It may be noted that BICUVOXrepresents only one member of a family of Bi203-based solid electrolyte phases, whichmay be derived from Bi4V2Ollbysubstitution of copper for vanadium. Many cations may be substituted for vanadium and the general acronym BIMEVOXwas given to these materials, which have been claimed for electrochemical oxygen separation at temperatures as low as 500 K [109].Besides copper, high oxide ion conductivity is reported for substituents titanium and niobium [212].
10 m D E N S E C E R A M I C M E M B R A N E S FOR OXYGEN S E P A R A T I O N
471
however, the interconnectivity between particles will not be ideal. These may be linked up to form so-called dead-ends or isolated clusters, which do not contribute at all to the conductance of the percolative system. Accordingly, conduction is expected to proceed through a significantly smaller fraction of consolidated particles or grains, which implies that the actual volume fraction of each phase should always be somewhat in excess of ~c. The optimum volume ratio is just above ~c of the high conducting phase, i.e. the metal phase, in order to have the highest effective ionic conductivity of the composite. Dual-phase membranes made of BY25-Ag [110] and YSZ-Pd [114] behave quite similar in having their conductivity threshold at about 33-35 vol% of the metal phase. These membranes were made by conventional ceramic processing techniques. The value of ~c obtained for these composite materials agrees well with the high concentration limit predicted by simple effective medium theory in which the composite is described as a three-dimensional resistor network [115]. The effective ionic conductivity is reduced relative to that what is expected purely on the basis of the volume fraction of the ionic conducting phase, which originates, at least partly, from the enhanced tortuosity of the migrating path for the oxygen anion due to partial blocking by the metal phase. It is therefore expected that a further gain in the oxygen flux can be realized through proper design of the microstructure [112,114]. The optimum situation would correspond with the one in which the particles of each phase line up in strings (or slabs) parallel to the applied gradient in oxygen partial pressure. Even though, theoretically, the critical volume fraction of the metal phase could be reduced in this way to a value practically equal to zero, such an approach is bounded by the additional requirement for practical membranes of fast surface exchange kinetics, especially for very thin membranes. The exchange reaction at the composite surface is confined to the three-phase boundary (tpb) between the gas, metal and electrolyte formed by particle grains being connected to the percolative network. Fast oxygen transfer can be sustained only if the corresponding length or area available to oxygen exchange is large enough, where it should be noted that the exchange reaction can only take place at a point remote from the tpb line which is shorter than the spill-over distance of electro-active species across the surface. The electrical field necessary to guide the current becomes distorted in the vicinity of the surface of a coarse-grained composite, where the separation between adjacent tpb lines is too large and, hence, only part of the surface is effective towards oxygen exchange. This contribution is stressed in the SOFC literature and is known as the constriction effect [116]. Often, it is the synergism between electrode and electrolyte material that leads to fast exchange characteristics. The oxygen flux through disc membranes made of BE25-Au (40 vol%) was found to increase almost one order of magnitude by substituting gold for silver in the composite [112]. This observation can be related to the higher activity of silver in the
472
10 m DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
oxygen exchange reaction on BE25, compared with gold, imposing less limitations on overall oxygen transport. Materials like, for example, Bi2CuO44 [117], TiN [112], MgLaCrOg,s [26] have been proposed to replace the inert metals. Even though, in the examples chosen, ionic and electronic transport are confined to separate phases, mixed ionic-electronic conductors could be useful. A systematic evaluation of dual phase membranes, however, is too new so far to come to definite conclusions. Besides simple modelling in terms of a short-circuited oxygen concentration cell, to our knowledge no one has yet described oxygen permeation through dual-phase membranes, taking into account the distinct three-dimensional aspects of the microstructure that may arise in practical composite materials. Besides high values for the oxygen flux (and permselectivity), commercial use of membrane systems will demand chemical, mechanical and structural integrity of applied materials in appropriate ranges of temperature and oxygen partial pressure. Dual-phase membranes have the obvious potential to distribute specific requirements among the system components. INTRODUCING ELECTRONIC CONDUCTION IN FLUORITE-TYPE OXYGEN ION CONDUCTORS
10.5
10.5.1 Introduction Stimulated by the search for candidate materials for electrodes in solid oxide fuel cells (SOFC) and oxygen separation membranes, researchers have explored the possibility of introducing electronic conductivity in oxygen-ion conducting fluorite-type matrices by doping with multi-valent dopants. The major factors which establish electronic conduction in the mixed-conducting oxide solid solutions obtained are, at a given temperature, (i) the multi-valent dopant fraction, (ii) its redox characteristics and (iii) oxygen partial pressure. The suggested mechanism for electronic conduction is the hopping of electrons between adjacent dopant ions of different valence charge. However, experimental data of oxygen permeation is still scarce. In the following sections, we briefly focus on the defect chemistry, which includes some fundamentals of mass transport, and shall summarize relevant work on selected oxides. 10.5.2 Defect Chemistry
The topic of 'mixed conduction in nonstoichiometric oxides" was reviewed by Tuller [24], and his comprehensive paper is recommended to the reader interested in more detail concerning the role of multivalent dopants on the defect chemistry of fluorite and fluorite-related oxides, and corresponding transport properties. Equations which express the oxygen flux in solid solutions of, e.g.,
10 -- DENSECERAMICMEMBRANESFOR OXYGENSEPARATION
473
ceria in stabilized zirconia, as a function of temperature, oxygen partial pressure and dopant concentration have been developed recently by Ling et al. [118] and Marques et al. [119]. In addition to the defect reactions given in Section 10.4.2.1, one extra reaction needs to be considered, i.e. the ionization of the multivalent cation. On using the general notation N for the multivalent cation one may write, N M' ~--- N~ + e
(10.42)
with equilibrium constant, [N~] n
KN=
[NM']
(10.43)
Mass conservation requires that [NM' ] + [N~4 ] -[NM]tota 1
(10.44)
Electroneutrality relation Eq. (10.31) must be rewritten to include the charged species NM': 2[Vo] + p + [D'] = 2[00" ] + n + [A'] + [NM']
(10.45)
With the aid of experimentally derived equilibrium constants, Eqs. (10.42)(10.45) may be used to construct the Kr6ger-Vink defect diagram, from which expressions for the partial conductivities of the mobile ionic and electronic defects can be derived [24]. The defect diagram obtained by Marques et al. [119] for lightly ceria-doped ZrO2-Y203, ignoring the possibility of defect association, is shown schematically in Fig. 10.7a. The corresponding conductivity diagram (Fig. 10.7b) can be obtained by multiplying each of the mobile species by their respective charge and mobility, which leads to the following expression for the total electronic conductivity, Gel = (~p 4- (~n 4- (~h
(10.46)
where (~n and o v represent the intrinsic n-type and p-type conductivities, respectively, and Oh is the extrinsic electronic conductivity owing to the multivalent cations. As distinct from the n- and p-type contributions, for which a band-like mechanism is assumed, the extrinsic contribution to electronic conduction is assumed to proceed via a small-polaron mechanism, involving the activated hopping of electrons between adjacent dopant cations of different valence charge. As the small polaron mobility includes the fraction of sites not already occupied by electrons [24], the extrinsic electronic conductivity Oh depends on both [Ce~r] and [Cezr'] and is given by (~h = F[Cezr'] [Ce~r] u~ e x p ( - E H / k T )
(10.47)
474
10 ~ DENSECERAMICMEMBRANESFOR OXYGENSEPARATION
(a)
23
.....
,//
---
Vo
. FM
-N-.-'-"-" =--" --a::~-"~--.----
.....
21 A |
E
0 ,,0
"10
19
0
17
-
i
/,
",,
9 I
..... I
I
I
-30
-20
-10
0
o ,
I 10
Iog(Po=/Pa)
(b) - ~ ~ "
.-;-,
E
0 6O
b 0
" "
~ ~
" - "
- , r ~ b
--
,
,,
,,,
,,
,
-4
-8
,.~~
~
-12 /./'/
P1/2 !.,
I,
',
-30
-20
-10
.
I
0
Iog(Po=/Pa) Fig. 10.7. (a) Defect and (b) conductivity diagram for ceria-doped YSZ at 1000~ The relevant parameters to construct the diagrams are given in Ref. [119]. The theoretical dependence of ionic transference number tionand oxygen permeability jo~ are given in (c). Dashed lines in (c) refer to YSZ. FM' (Yzr') represents the aliovalent dopant used. Reproduced (slightly adapted) from Marques et al. [119].
10
-
-
475
DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
(c)
,,
,,=
,
i
,,,,
,,,i
,,
,
,,
,,
,
,,
,,,i,,
II
0.25
";',
I
o
[I
I III II
0.8
0.20 I I I I
|
E
I
0.15
0.6
i,.0
0
E
0.10
0 -., 9
0.05
0.4
-
0.2
I
i
J i ....
-20
-10
,,
|
,
.
0
Iog(Po2/Pa) Fig. 10.7c. C a p t i o n o p p o s i t e .
where u ~ is the pre-exponential of the mobility and EH the hopping energy. Accordingly, r~h displays a maximum at a critical oxygen pressure, P~/~, characterized by equal concentrations of [Ce~.r] and [Cezr']. On assuming that the oxygen vacancy concentration remains fixed by the aliovalent dopant concentration, one may derive from Eqs. (10.42)-(10.45) that the Po2-dependence of the small polaron conductivity at given temperature takes the form, K ~ D1/4 J-O 2
D1/4 + 1)2 u~ exp(-EH/kT) (10.48) (K, ~o~ where K~ -" K~[ C e z r ] t o2t a l and K~ = 21/2Ke/(KceKg1/4[Yzr ,]1/2). The partial ionic and electronic conductivities may be substituted into the Wagner equation (Eq. (10.10)) to derive an expression for the oxygen flux. Typical results of such calculations are given in Fig. 10.7c, in which the oxygen flux at given temperature is plotted against Po~", assuming air to be present at the feed side of the membrane. It can be seen that the oxygen flux saturates upon lowering Po2". An inflection point occurs at Po2" = P1/2. At the lowest values of Po2", the curve bends upwards again due to the onset of the intrinsic electronic conduction. For extended discussion we refer the reader to the original papers [118,119]. c~h - F
10.5.3 Examples (a) Ceria-doped ZrO2-Y203 Electrical properties of solid solutions ZrO2-CeO2-Y203 have been investi-
476
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
gated thoroughly by Cal6s and Baumard [31,32]. The main features have been confirmed by others [120,121]. The amount of 10 mol% yttria used by Cal6s and Baumard ensured a minimum concentration of oxygen vacancies in a wide range of experimental conditions. For all ceria dopant levels and temperatures (1000-1400~ ionic conduction is found to predominate at high Po2 values. Doping with ceria decreases the ionic conductivity up to (ZrO2)o.45--(CeO2)o.45(Y203)0.1, beyond which it increases again up to the composition as high as (CeO2)0.9--(Y203)0.1.
For not too low ceria contents the total electrical conductivity displays a maximum at reduced P02 values, shifting to lower P02 values as the temperature is decreased. It thereby follows the predictions of the preceding section, which is generally taken as evidence that the electrons in the ceria-based solid solutions move by a hopping mechanism. The maximum can be correlated with the presence of nearly equal concentrations of Ce 4+ and Ce 3+ if one takes into account the concomitant change in ionic conductivity with decreasing P02" Cal6s and Baumard deduced that, for (ZrO2)o.8r(CeO2)o.09-(Y203)0.1 at an oxygen pressure of about 10-13-10-14atm and temperature 1200~ (~elis of the same order of magnitude as Oionand approximates 0.06 S cm d. The contribution of the electronic to the total conductivity, at a given P02 and temperature, increases with increasing ceria content, albeit at the expense of the ionic conductivity (up to the composition (ZrO2)0.45--(CeO2)0.45--(Y203)0.1). For the most reducing conditions the conductivity becomes predominantly ionic again, albeit that the corresponding value is significantly less than that observed at high Po~. This may cause surprise knowing that the major fraction of the cerium ions under the reduced conditions adopts the trivalent state and, hence, the concentration of oxygen vacancies will be enhanced. The reduced ionic conductivity at low Po2 is attributed to enhanced defect interactions and/or lattice distortions. This type of behaviour is reminiscent to that of zirconia and ceria based electrolytes, for which it is observed that the ionic conductivity increases with the extent of aliovalent doping up to a certain limit beyond which defect ordering or formation of defect associates lowers the ionic conductivity [73]. In solid solutions with a high ceria content, for example, (ZrO2)0.45--(CeO2)0.45-(Y203)0.1, the ionic conductivity at 1100~ decreases rapidly at an oxygen partial pressure below about 10-11 atm due to the formation of ordered pyrochlore-type domains, as confirmed by XRD measurements. The more recent work on 10 mol% ceria-doped YSZ (5.8 mol% yttria) by Ramanarayanan et al. [122] showed that the ionic transference number tion decreases with reduction in grain size. This observation suggests that the preferred path for electronic conduction is via the grain boundary. TEM imaging confirmed a strong tendency for cerium cations to segregate to the grain boundary, showing enrichments up to about 20 mol% compared with the value of 11 mol% observed in the lattice.
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
477
To the best of our knowledge literature reports from oxygen permeation measurements on solid solutions ZrO2--CeO2-Y203 are not available. Recent data from measurements on the related system ZrO2--CeO2-CaO are included in Table 10.1.
(b) Titania-doped ZrO2--Y203 Pure titania has the rutile structure and therefore has limited solubility in YSZ. The observed linear decrease in lattice parameter with increasing titania concentration in these solid solutions suggests that titanium cations enter the lattice substitutionally for zirconium. Concordant with the data from XRD measurements [29,30,123] the cubic fluorite structure is retained upon addition of 12-20 mol% titania, above which a second phase appears, claimed to be ZrTiO4 [123]. The spread in data of the solubility limit produced by different authors may be due to slight differences in, e.g., yttria concentration, sample processing, sintering temperature and impurity content in the cited studies. Microstructural investigations based on SEM and TEM indicated that precipitates of the second phase actually may appear already at lower titania contents [123,125]. Contrary to the earlier observations [29,30], recent studies on electrical conductivity by Marques et al. [124] and Lindegaard et al. [125] indicate that the lattice ionic conductivity decreases with the extent of incorporation of titania into YSZ. Results confirm that the ionic conductivity of 10 mol% titania-doped YSZ in air, at a typical temperature of 1000~ is about ten times less than that of undoped YSZ. The hopping electronic conductivity at this temperature is estimated to be ---10-7 S cm -1 [121]. For similar dopant levels addition of titania appears to be more effective in enhancing the electronic conductivity of YSZ than ceria, which is not expected considering the redox behaviour of pure ceria and titania. Using thermogravimetric measurements on YSZ with ceria and titania additions up to 10 mol%, Marques et al. [124] confirmed that Ce 4+ cations in these solid solutions are more easily reduced than Ti4+. The observed increase in grain boundary conductivity with increasing titania concentration [29,30, 126] and decreasing Po~ [125] would indicate that electronic conductivity occurs at the grain boundaries. Liou and Worell [29,30] presumed segregation of Tizr to the grain boundary region, but within experimental uncertainty of EDS (Energy Dispersive Spectroscopy of X-rays) no evidence was found for titania-rich grain boundaries in the already cited study by Marques et al. [124]. The higher electronic conductivity of titania-doped YSZ was interpreted to reflect the formation of highly mobile electronic defects (large polarons) in the bulk, by comparison with the low mobility of small polarons formed in ceria-doped specimens. At 1000~ significant levels of electronic conduction in titania-doped YSZ, as in ceria-doped specimens, are found only under strongly reducing atmospheres. Data of oxygen permeability have been presented for the ZrO2-Y203TiO2 system by Arashi and Naito [127] (see also Table 10.1). By virtue of its
478
10 m DENSE C E R A M I C M E M B R A N E S F O R O X Y G E N S E P A R A T I O N
stability up to 2000 ~ it is proposed to be used as a m e m b r a n e for direct hydrolysis of water to produce hydrogen [128].
(c) Miscellaneous materials Terbia has been dissolved in pure ZrO2 to form mixed-conducting solid solutions with Tb203. 5 concentrations as high as 50 mol% by Iwahara et al. [129]. Briefly, the electrons in these mixed oxides move by hopping between Tb 3+ and Tb 4+ ions, the coexistence of which ions has been confirmed using the XANES (X-ray absorption near edge structure) technique [130]. The relative contribution of the electronic to the total conductivity measured by Iwahara et al. increases with increasing terbia concentration, in spite of the fact that the relationship between the latter quantity and (Itotalr at a given temperature and oxygen pressure, turns out to be very complex. At 900~ and atmospheric pressure, (~totalfor (ZrO2)0.7-(Tb203.5)0.3 is 1.8 x 10 -2 S cm -2, and tio n is 0.30. Oxygen from permeation, at 900~ was found at the argon side of a 2-3 m m thick disc m e m b r a n e of this composition, at a rate of about 5 x 10 -9 mol c m -2 s -1, the value of which was measured with ambient air maintained at the feed side. W h e n terbia is dissolved in YSZ, this ensures a m i n i m u m value for the oxygen vacancy concentration, which is then fixed by that of yttria. Cao et al. [131,132] examined the electrical conductivity and oxygen permeation of selected compositions, including (ZrO2)07-(Tb2035)03-y--(Y2OB)y with y = 0, 0.025, 0.05 and y = 0.072. At 900~ (~totaldecreases f r o m l . 2 x 1 0 - 2 S c m -2 for y = 0 to 0.86 • 10-2 S cm -2 for y = 0.072, where tioniS 0.37 and 1, respectively. The oxygen flux, at 900~ passing from the air to the helium side of 2 m m thick disc-shaped membranes varied in the range 2.6-3.7 x 10-11 mol cm -2 s-1. Hardly any effect of the yttria-content on oxygen fluxes was measured. Based upon these results, amongst some additional experimental facts, e.g., the Po2-dependence of the oxygen flux, it is concluded that the surface exchange reaction is the rate limiting step for oxygen permeation. Regrettably, no account is given as to w h y the data of oxygen permeation are almost two orders of magnitude lower than the one claimed for y = 0 by Iwahara et al. [129]. Dense thin films of several microns could be grown successfully on different porous ceramic substrates by electrochemical vapour deposition (ECVD) [133,134]. An oxygen permeation flux of 7 x 10-1~mol c m -2 s -1 at 953~ was measured for a film (ZrO2)0.86--(Tb2OB.5)0.10--(Y203)0.04 of thickness 8 ~tm deposited on a coarse a-alumina substrate, which value increased to 3 x 10-8 mol c m -2 s -1 if the helium line was switched to CO/CO2 having Po2-5 x 10-16 atm. In these experiments, air was supplied to feed side of the membrane. Finally, we briefly describe the observations recently m a d e in the present authors' laboratory in an attempt to increase the oxygen permeation flux through stabilized bismuth oxide by substitution of the 8-Bi203 host with 40 mol% terbium on the bismuth sites (BT40). Measurements using the concentration cell method and ac impedance confirmed that BT40 exhibits good p-type
10 n DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
479
conductivity and is an excellent mixed conductor with ionic transference numbers, tio n = 0.74 at 650~ and tion = 0.85 at 800~ in air [135]. Using ambient air as the source of oxygen and helium as the sweep gas on the other side of dense BT40 disc membranes, in the range of thickness 0.07-0.17 cm and temperature 600-800~ did not yield the expected increase in the oxygen flux, over BE25 [136]. Isotopic exchange measurements in the relevant range of oxygen partial pressure and temperature showed that both oxides exhibit an almost equal activity in oxygen exchange [104], which is in support of the conclusion made from oxygen permeation measurements that the oxygen fluxes through BT40, at the conditions covered by the experiments, are limited by the surface exchange kinetics. Additional attempts have been presented to render hosts with the fluorite and the related pyrochlore structure electronically conductive by doping with mixed-valence a n d / o r shallow dopants. The list of dopant materials examined includes oxides of elements of, for example, Ti, Cr, Mn, Fe, Zn, Fe, Sn, Ce, Pr, Gd, Tb and U. In general, however, the extent of mixed conductivity that can be obtained in fluorite-type ceramics is rather limited, by comparison with the corresponding values found in some of the perovskite and perovskite-related oxides considered in the next section. 10.6 A C C C E P T O R - D O P E D PEROVSKITE A N D PEROVSKITE-RELATED OXIDES
10.6.1 Introduction
The general trend observed from the pioneering studies on oxygen permeation through perovskites of the type L n l _ x a x C O l _ y B y O 3 _ a (Ln = La, Pr, Nd, Sm, Gd; A = Sr, Ca, Ba; B= Mn, Cr, Fe, Co, Ni, Cu) by Teraoka et al. [37-39] is that higher oxygen fluxes are facilitated by increased A-site substitution, and a lower thermodynamic stability of the particular perovskite. Clearly, not all these perovskite compositions are useful for oxygen delivery applications. For example, ceramics based on Lal_xAxCrOB-a (x = Sr, Ba, Ca), Cal_xSrxCrl_yMnyO3_a and Cal_xCaxCrl_yCoyO3_a have been proposed for use as interconnection material (separator) in solid oxide fuel cells (SOFC), and therefore should be dense and impermeable in order to prevent burning off of the fuel without generating electricity [137,138]. Selected perovskite compositions are also targeted in basic SOFC research for use as potential electrode material for the cathodic reduction of oxygen. The most promising cathode materials to date are the manganites Lal_xSrxMnO34 [137,138]. The composition with x = 0.15 scarcely permeates oxygen up to 900~ as was measured by feeding air and helium to opposite sides of a dense sintered membrane of I mm thickness [136]. The observed behaviour is consis-
480
1 0 - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
tent with the low value of the oxygen self-diffusivity in La0.sSr0.5MnO3~, determined by 180-160 isotopic exchange, and can be attributed to the small negative departure from oxygen stoichiometry exhibited in the range of temperature and oxygen pressure covered by experiment [139]. On the other hand, oxygen transport is actually predicted to be quite fast under conditions of high oxygendeficiency, i.e. low oxygen partial pressures, as the oxygen vacancy diffusion coefficient of La1_xSrxMnO34 was found to be comparable in magnitude with that of Fe- and Co-based perovskites [140]. Emerging from the first of these studies by Teraoka et al. [37] is that in the series La1_xSrxCo1_yFeyO34the oxygen fluxes increase with Co and Sr content, the highest flux being found for SrCo0.sFe0.2034. Data were obtained with air on one side of a 1 mm thick disc specimen, using helium as sweeping gas on the other side, up to a maximum temperature of 1150 K. The observed oxygen fluxes were found to be roughly proportional to the ionic conductivity of the perovskites, which is in agreement with the fact that the electronic conductivity of compositions in this series can be extremely high, typically in the range 102103 S cm -1 [40]. Four-probe dc measurements using electron blocking electrodes showed that the ionic conductivity at 800~ in air can be 1-2 orders of magnitude higher than that of stabilized zirconia [40]. These findings have been confirmed by others, apart from scatter in the published data, which partly reflects the experimental difficulties in measuring the ionic conductivity in these predominantly electronic conductors [141-144]. In a subsequent study, Teraoka et al. [38] investigated the influence of A and B site substitution on oxygen permeation through La0.6A0.4Co0.sFe0.2034(A = La, Na, Ca, Sr, Ba) and La0.6Sr0.4Co0.8B0.203,s(B = Cr, Mn, Fe, Co, Ni, Cu). As seen from Figs. 10.8 and 10.9, the oxygen permeability in the two series increases in the respective orders La < Na < Sr < Ca < Ba and Mn < Cr < Fe < Co < Ni < Cu, which differ from trends in the periodical system, as far as comparison is meaningful. Results from ionic and electronic conductivity measurements of La0.6A0.4Co0.8Fe0.203_8 (A = La, Ca, Sr) and La0.6Sr0.4Co0.8B0.203.8(B = Fe, Co, Ni, Cu) suggest that oxygen permeation is governed by the ionic conductivity [39]. In the homologous series Ln0.6Sr0.4CoO3_a,the oxygen flux was found to increase in the order La 3+ < Pr 3+ < Nd 3+ < Sm 3+ < Gd 3+ which corresponds with a decrease in radius of the lanthanide-ion [38]. Since the initial observations by Teraoka et al., a considerable number of studies have appeared. Selected perovskite compositions have been re-examined, while a few others have been adapted in an attempt to optimize the oxygen fluxes. The list of materials for which oxygen permeation data are presently available has been extended to include, LaCoO34 [145], La1_xSrxCoO34 [146-149], La1_xSrxFeO34 [150,151], Lal_xAxCol_yFeyO34 (A = Sr, Ca) [12,139,144, 152], SrCo0.8Fe0.2034 [13,41,153,154], SrCoo.sBo.203_8 (B = Cr, Co, Cu) [154], SrCol_• (B = Cr, Mn, Fe, Ni, Cu, x = 0...0.5) [155], and Y1_xBaxCoO34 [156]. In general, fair agreement
10 -- DENSE CERAMICMEMBRANESFOR OXYGENSEPARATION
481
1.50
9
Ba
9
Ca
9
Na
9
La
Sr A ,-;.,
1.00
E o
o
E ,,, O
0.50
0.00 300
500
700
900
temperature (~ Fig. 10.8. Temperature variation of the oxygen permeation rate from the air to the helium (30 cm 3 min -1) side of disc membranes La0.6A0.4Co0.8Fe0.2034(A = Na, Ba, Ca, Sr), 20 m m in diameter and 1.5 m m thick, after Teraoka et al. [38]. Reproduced (data re-scaled) from Teraoka et al. [38].
is obtained with data produced by Teraoka et al., albeit that in a number of studies the observed oxygen fluxes are reportedly found to be significantly lower [12,153,154]. The pioneering studies by Teraoka et al. [37-39] have opened a very challenging research area as the perovskites, e.g. Lal_~SrxCo1_yFeyO34, have a bright future for use as oxygen separation membrane. The precise composition may be tailored for a specific application, but this has not yet been fully developed. One of the important issues is considered to be the low structural and chemical stability of the perovskites, especially in reducing environments, which remains to be solved before industrial applications become feasible. In order to meet this challenge, it is necessary first to understand the factors that limit and control the quality criteria for any given application. The perovskite and related oxides exhibit a great diversity of properties, like electrical, optical, magnetic, catalytic properties, which have been studied extensively. In the following sections, we mainly focus on those properties affecting the magnitude of the oxygen fluxes through these materials.
482
10 n DENSECERAMICMEMBRANESFOR OXYGENSEPARATION 1.50
~"
1.00
+
9
Cu
9
Ni
9
9
Co
9 9
Fe Cr Mn
.'~~ 0.50
0.00
500
~
~
,
600
I
700
temperature
,
I
800
,
900
( ~ C)
Fig. 10.9. Temperature variation of the oxygen permeation rate of La0.6Sr0.4Co0.8B0.203-8(B = Cr, Mn, Fe, Co, Ni, Cu) after Teraoka et al. Experimental conditions are specified in the legend of Fig. 10.8. Reproduced (data re-scaled) from Teraoka et al. [38].
10.6.2 Structure and Defect Chemistry 10.6.2.1 Perovskite Structure The ideal perovskite structure ABO3 consists of a cubic array of corner-sharing BO 6 octahedra, where B is a transition metal cation (Fig. 10.10). The A-site ion, interstitial between the BO 6 octahedra, may be occupied by either an alkali, an alkaline earth or a rare earth ion. In m a n y cases the BO 6 octahedra are distorted, or tilted, due to the presence of the A cation, which is generally larger in size than the B cation. Alternatively, the perovskite structure m a y be regarded as a cubic close-packing of layers AO 3 with B cations placed in the interlayer octahedral interstices [157]. The latter turns out to be more useful in distinguishing different structural arrangements (stacking sequences) of perovskite blocks. The tolerance limits of the cationic radii in the A and B sites are defined by the Goldschmidt factor, which is based on geometric considerations: t = (ra + ro) / (~-(rB + ro)), where rA, rB and ro are the radii of the respective ions [158]. When the distortion becomes too large, other crystal symmetries such as
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
483
~ A @0 QB ::::iiii:i::..
Fig. 10.10.Ideal perovskite structure. orthorhombic and rhombohedral appear. Nominally, the perovskite structure should be stable between 1.0 < t < 0.75. The ideal perovskite lattice exists only for tolerance factors t very close to one. Clearly, it is the stability of the perovskite structure that allows for large departures from ideal stoichiometry, resulting either from the substitution with aliovalent cations on the A or B-site or from redox processes associated with the presence of transition metal atoms which can adopt different formal oxidation states. Oxygen vacancies are free to move among energetically equivalent crystallographic sites as long as the perovskite structure exhibits ideal cubic symmetry. The degeneracy between sites disappears upondistortion of the lattice towards lower symmetries. The onset of electronic conductivity mainly depends on the nature of the B-site cation. The total electrical conductivity can be either predominantly ionic as in the acceptor-doped rare earth aluminates or predominantly electronic as in the late transition metal containing perovskites considered below.
10.6.2.2 Nonstoichiometry Important contributions to the area of defect chemistry of the acceptor-doped Lnl_xAxBO3, perovskites, where B is selected from Cr, Mn, Fe or Co, have been made by a number of investigators. Particular reference is made to reviews provided by Anderson [159,160] and Mizusaki [161]. The substitution of divalent alkaline-earth ions on the A-site increases the concentration of oxygen vacancies. Temperature and oxygen partial pressure determine whether charge compensation occurs by an increased valency of the transition metal ion at the B-site or by the formation of ionized oxygen vacancies. Thermogravimetric studies have indicated that in, for example, LaCrO3, YCrO3 and LaMnO3 the
484
10-- DENSECERAMICMEMBRANESFOROXYGENSEPARATION
native nonstoichiometric ionic defects are cation vacancies, leading to oxygenexcess stoichiometries [160]. For simplicity, it is assumed here that extrinsic ionic defects generated by A-site substitution prevail, i.e. only oxygen-deficient stoichiometries are considered. Furthermore, crystallographic sites available for oxygen are taken to be energetically equivalent. For the purpose of our discussion, LaFeO3 is considered to be the host for substitution. The dissolution of SrFeO 3 into this material can be represented by, SrFeO3
LaFeO 3 )
SrLa' + FeFe + 30~)
(10.49)
The incorporation of Sr 2§ thus leads to charge compensation by the formation of Fe 4+ ions, which is in accord with the Verwey principle of controlled ionic valency [162]. The extent of oxygen non-stoichiometry is established by the following defect chemical reactions, 2FeFe + O~) ~ 2Fete ~
2Fete + V'o + 1/2 0 2
FeFe' + FeFe
(10.50) (10.51)
with the corresponding equilibrium constants, [Fete] 2 [V'o] Po1/2 2 Kg = [FeFe]2 [O~)]
Kd =
[FeFe'] [FeFe] [Fete] 2
(10.52)
(10.53)
The oxygen vacancies formed at elevated temperatures and low oxygen partial pressure are assumed to be doubly ionized. The thermally activated charge disproportionation reaction given by Eq. (10.51) reflects the localized nature of electronic species and may be treated as equivalent to the genera.tion of electrons and electron holes by ionization across a pseudo band gap (cf. Eq. (10.27)). The associated free enthalpy of reaction may be taken equal to the effective band gap energy. At fixed A / B site ratio the following condition must be fulfilled [FeFe'] + [Fete] + [FeFe] = 1
(10.54)
and the condition of charge neutrality is, [SrLa'] + [FeFe'] = 211/'O] + [FeFe]
(10.55)
In the absence of extended defects, i.e. no interaction between point defects, Eqs. (10.52)-(10.55) may be used with the aid of experimentally determined equilibrium constants to construct the Kr6ger-Vink defect diagram, from which ex-
10 - - DENSE CERAMICMEMBRANESFOR OXYGENSEPARATION
485
pressions for the partial conductivities of the mobile ionic and electronic defects can be derived. Oxygen nonstoichiometry of the perovskites Lal_xSrxBO3.s(B = Cr, Mn, Co, Fe) and its relationship with electrical properties and oxygen diffusion has been studied extensively [159-161]. Typical nonstoichiometry data for La1_xSrxFeO34 and for some other perovskites as obtained from gravimetric analysis and coulometric titration are given in Fig. 10.11. At small oxygen deficiency, acceptor dopants are the majority defects. The charge neutrality condition then becomes, [SrLa'] = [FeFe]
(10.56)
In this region, one finds for the oxygen non-stoichiometry 5, oc p o~/2 2
(10.57)
noting that 8 = [V6], by definition. A plateau is observed around the point of electronic stoichiometry, 8 = x/2, where the charge neutrality condition reads, [Srca'] = 2[V6]
(10.58)
3.05
& &
A
.
.
.
.
.
-
/
2.95 o3
2.9
Lao.9Sro.lCo03.6 9 Lao.oSro.lFeO3.s
0
o Lao.3Sro.7Cr03-8 zx Lao.2Sro.sMn03-8
2.85
.8
lu
-20
~
I
I
I
-16
I
I
,
I,
I
I
-12
I,
.
I
I
-8
I
I
I
I
-4
I
I
'
0
Iog(P 0 z/atm) Fig. 10.11. Data of oxygen nonstoichiometry of Lao.75Sro.25CrO3-~, Lao.gSro.lFeO3-~, Lao.gSro.lCoO3-~, and Lao.sSro.2MnO3-~ at 1000~ as a function of oxygen partial pressure. Solid lines are results from a fit of the random point defect model to the experimental data. Reproduced (slightly adapted) from Van Hassel et al. [185].
486
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
corresponding with a minimum in the electronic conductivity of La1_xSrxFeO34 [163,164]. In this region, the oxygen non-stoichiometry is virtually constant. As the oxygen activity decreases further, oxygen vacancies are again generated, down to the oxygen activity at which decomposition of the perovskite structure occurs. The onset of the different regions depends on the nature of the transition metal B-cation. The incentive of B-site substitution can therefore be to optimize oxygen transport in appropriate ranges of oxygen partial pressure and temperature. As discussed below, doping may also increase stability or suppress cooperative ordering of oxygen vacancies. 10.6.2.3 Localized versus Delocalized Electrons Given the relative success of the above point defect scheme to model the experimental data of oxygen nonstoichiometry and electrical conductivity for Lal_vSrxFeO34 [165,166] and Lal_xSrxCrO3.s [167], its use is less satisfactory for Lal_xSrxCoO34 and Lal_xSrxMnO3~, which compounds show notably high values for the electronic conductivity. Nonstoichiometry of the compounds La1_xSrxCoO34 (x = 0, 0.1, 0.2, 0.3, 0.5 and 0.7) in the range 10-5 < Po2 < 1 atm and 300< T < 1000~ was investigated by Mizusaki et al. [168] using thermogravimetric methods. At 800~ 5 in Lal_xSrxCoO3_s varies almost proportional to Po2n with n = - 1 / 2 for x = 0 to n = - 1 / 1 6 for x = 0.7 (see Fig. 10.12). No plateau is observed around 8 = x/2. Fitting the 5-Po2 relationship in accord with the random point defect model leads to very large concentrations of disproportionation reaction products Coco" and Coco. A corollary is that the pseudo bandgap must be very small. The model fit, however, is less satisfactory for high Sr substitutions [169]. A similar e x p l a n a t i o n holds for La1_xSrxMnO34, d i s r e g a r d i n g the oxygen-excess stoichiometries seen in this system at high oxygen partial pressures. At high oxygen deficiency of the perovskite, the validity of the ideal mass action equations (based upon dilute solution thermodynamics) cannot be assumed a priori. In addition, interaction and association between defects are expected at high defect concentrations. A further limitation concerns the nature of electronic defects. The general assumption, that in the first row transition metal perovskites changes in the oxygen content leads to changes in the 3d electronic configuration, may be too naive. It is based implicitly on the idea that oxygen is strongly electronegative and, by comparison, the 3d electrons can be easily ionized. There is substantial evidence from soft-X-ray absorption spectroscopy (XAS) based studies that the electron holes introduced by doping with divalent earth-alkaline ions go to states with significant O 2p character [170]. This has also been reported for the perovskite-related oxide YBa2Cu306§ [171]. In a localized description, i.e. assuming a narrow bandwidth of the hole band derived from the O 2p band, this would imply that 0 2- is effectively converted into O-.
10 -- DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
I
i''
I
-"
487
~
' I
"'
-i
....
-1
.3
-2
a
0
0 ,-4
m
-3
m
....
i
I
_
1
-4
1
log
1,,
-2
(P
02
I
/atm)
l
0
Fig. 10.12. O x y g e n pressure dependence of 6 in Lal-xSrxCoO3-6 for different strontium contents at 800~ Reprinted from Mizusaki et al. [168].
A proper description of electronic defects in terms of simple point defect chemistry is even more complicated as the d electrons of the transition metals and their compounds are intermediate between localized and delocalized behaviour. Recent analysis of the redox thermodynamics of La0.sSr0.2CoO34based upon data from coulometric titration measurements supports itinerant behaviour of the electronic charge carriers in this compound [172]. The analysis was based on the partial molar enthalpy and entropy of the oxygen incorporation reaction, which can be evaluated from changes in emf with temperature at different oxygen (non-)stoichiometries. The experimental value of the partial molar entropy (free formation entropy) of oxygen incorporation, Aso2, could be
488
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
fitted by assuming a statistical distribution among sites on the oxygen sublatfice, Aso2 = s~ - 2k In
(3 - 8 ) 8
(10.59)
where s o is a constant. That is, no entropy change associated with electron annihilation can be identified. The partial molar enthalpy (free enthalpy of formation of vacancies) associated with oxygen incorporation was found to decrease almost linearly with 8. A first inclination might be to assume that the mutual repulsion between oxygen vacancies increases with increasing oxygen deficiency. But this interpretation immediately raises the question why such a behaviour is not found in the case of La1_xSrxFeO34 [166,167]. Instead, the experimental data are interpreted to reflect the energetic costs of band filling. With increasing oxygen nonstoichiometry in La0.sSr0.2CoO34 the two electrons, which are needed for charge compensation of a single oxygen vacancy, are donated to an electron band broad enough to induce Fermi condensation characteristic of a metallic compound. The average density of electron states at the Fermi-level is determined to be 1.9_+0.1 eV-1 per unit cell. The physical significance of the work is that the defect chemistry of La0.8Sr0.2CoO3-4 cannot be modeled using simple mass action type of equations. An empirical model for the oxygen non-stoichiometry of La0.sSr0.2CoO3-8 is proposed, which demonstrates that the density of states is related to the slope of the log-log plots of 8 versus Po~. In support of these interpretations, it is noted that XAS has not been successful in detecting charge disproportionation in LaCoO34, due to localization of electrons, in the temperature range 80-630 K [173]. The nonstoichiometry data obtained for La0.8Sr0.2CoO3-4are found to be in good agreement with earlier results from gravimetric analysis in the series La1.~rxCoO34 obtained by Mizusaki et al. [168], which authors arrived at more or less similar conclusions regarding the role of electronic states in the energetics of oxygen incorporation into these compounds.
10.6.3 Oxygen Desorption and Perovskite Stability As seen from Fig. 10.11, the value of (3-8) in Lal_xSrxCoO34 falls off with decreasing oxygen activity much more rapidly than for the other compounds shown. The general trend at which the perovskites become nonstoichiometric follows that of the relative redox stability of the late transition metal ions occupying the B-site, i.e. C r 3+ > Fe 3+ > Mn 3+ > C o 3+. The reductive nonstoichiometry of the cobaltites increases further by partial B-site substitution with copper and nickel. The reductive (and oxidative) nonstoichiometry and the stability in reducing oxygen atmospheres of perovskite-type oxides was reviewed by Tejuca et al. [174]. Data from temperature programmed reduction (TPR) measurements indicate that
10 E D E N S E C E R A M I C M E M B R A N E S FOR OXYGEN S E P A R A T I O N
489
the stability (or reducibility) of the perovskite oxides increases (decreases) with increasing size of the A ion, which would be consistent with the preferred occupancy of the larger Ln 3+ ion in a 12-fold coordination. The trend is just the reverse of that of the stability of the corresponding binary oxides. The ease of reduction increases by partial substitution of the A ion, e.g., La 3+ by Sr 2+. Trends in the thermodynamic stabilities of perovskite oxides have been systematized in terms of the stabilization energy from their constituent binary oxides and the valence stability of the transition metal ions by Yokokawa et al. [175]. The stability of the undoped perovskites LaBO3_a, at 1000~ expressed in terms of Po2 decreases in the order LaCrO34 (10 -20 atm) > LaFeO34 (10 -17 atm) > LaMnO3_~ (10-15 atm) > LaCoO34 (10 -7 atm), noting that the cited value for LaCrO34 corresponds with the lowest limit in a thermogravimetric study by Nakamura et al. [176]. The same trend was found by means of TPR [174]. Tabata et al. [177] and Seyama [178] both described significant differences in the chemical composition of the surface, due to Sr segregation, compared with the bulk composition in a series of powders Lal_~SrxCoO3_~. This indicates a behaviour of the surface different from that of the bulk in these compounds. Not only can this account for a number of observations made in the total oxidation of CO and C H 4, as discussed by the authors, but it is also considered to be an important factor when one tries to correlate the composition of a perovskite with its activity in surface oxygen exchange. The sorpfion kinetics of oxides is certainly influenced by their corresponding defect structure. A number of interesting observations were made by Yamazoe and co-workers [179,180], showing that for perovskites LaMO3_~ (M = Cr, Mn, Fe, Co, Ni), Lal_vqrxCoO3_~(x = 0, 0.2, 0.4 and 1) andLa0.8Ao.2CoO3_8 (A = Na, Ca, Sr and Ba), two distinct types of oxygen are desorbed upon heating in a helium stream after a pre-treatment step in which the oxide was saturated in an oxygen-rich atmosphere at high temperature, followed by slow cooling to room temperature. The oxygen desorbed in a wide range at moderate temperatures, referred to as c~-oxygen, was found to be correlated with the amount of partial substitution of the A ion. The onset temperature of the so-called ~-desorpfion peak observed at high temperature was correlated with the thermal decomposition temperature of the corresponding transition metal oxides. Accordingly, the ~-peak corresponds with the reduction of the transition metal ion from B 3+ to B2+. The partial substitution of Co by Fe in the series Lal_xSrxCOl_yFeyO3_~ stabilizes the Co B+oxidation state (no -peak observed), while shifting the o~-type of desorpfion to lower temperatures [181,182].
10.6.4 Equationsfor Oxygen Transport Equations for oxygen transport can be derived from the point defect equilibria discussed in Section 10.6.2.2. This provides us with some general insight
490
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
into the transport behaviour of oxygen-deficient perovskites. Strictly speaking, the equations presented below are valid at low defect concentrations only, i.e. assuming oxygen defects to be randomly distributed. Oxygen transport in the perovskites is generally considered to occur via a vacancy transport mechanism. On the assumption that the oxygen vacancies are fully ionized and all contribute to transport, i.e., oxygen defects are not associated, the Nernst-Einstein equation reads, Gi~
4 F 2 [ V o ] Dv ----- RTVm
(10.60)
where Dv is the vacancy diffusion coefficient and Vm is the perovskite molar volume. Since electronic conduction in the perovskites predominates, i.e. Gel > Glow,the integral in the Wagner equation (Eq. (10.10)) involves only Gionover the applied oxygen partial pressure gradient. Using Eq. (10.60), we may rewrite the Wagner equation, to give In P"o2 jo 2 =
Dv 4VmL ~ ~)d In Po2
(10.61)
In P'o2
by virtue of 6 = [Vo]. Evaluation can be performed numerically provided that Dv and the 5-1n(Po2) relationship are known. The ability of Eq. (10.61) to quantitively fit experimental data of oxygen permeation is illustrated for La0.9Sr0.1FeO3~ in Fig. 10.13. Similar results have been presented for, e.g., La0.75Sr0.25CrO3~ [183] and La0.70Ca0.30CrO3~ [184]. The analytical solution of the integral given by Eq. (10.61) incorporating random point defect chemistry has been given by Van Hassel et al. [185]. When data of oxygen nonstoichiometry follows a simple power law 8 ~ P"02 , integration of Eq. (10.61) yields an expression similar to that of Eq. (10.18) having ~ = DvS~ n. Examination of the data from oxygen permeability measurements on disc specimens of thickness 2 mm in a series Lal_xSrxCoO3_8 (0_ x _< 0.8) in a study by Van Doom et al. [148] indicate that the results, at 1000~ can be fitted well by this equation, the validity of which is usually restricted to a small range in oxygen partial pressure. For compositions x _0.1 atm) and high temperature the perovskite phase is thermodynamically stable. At relatively low oxygen partial pressure and low temperature a perovskite-brownmillerite two-phase region is found. The brownmillerite phase has only a small homogeneity region around 3 - 5 = 2.5. Below Tt, the situation during flux measurements therefore becomes very complicated, considering the fact that the Po2gradient across the membrane also may cross the two-phase region provided, of course, that such a gradient is imposed during experiment. The studies report slow kinetics of transformation between the brownmillerite and perovskite phases in view of the long times for the oxygen flux to reach steady-state conditions at these modest temperatures. Kruidhof et al. [154] attributed these to a progressive growth of microdomains of the ordered structure in a disordered perovskite matrix. Based on experiments, in which the membrane thickness was varied in the range 5.5-1.0 mm, Qiu et al. [153] arrived at the conclusion that the surface oxygen exchange process is the rate limiting step in the overall oxygen permeation mechanism. Further experimental evidence that the oxygen fluxes through SrCo0.8Fe0.2034 are limited by the surface exchange kinetics was given by the present authors [41]. Fitting the oxygen permeation fluxes obtained from measurements at 750~ under various oxygen partial pressure gradients to Eq. (10.18) yielded a positive slope of n = +0.5, where a value between 0 and -0.5 is expected from the experimentally observed In 5--ln PO2 relationship" However, these results merit further investigation as the flux data were taken at a temperature just below the order--disorder transition in this material. It is already known for some time that SrCoO3_ 6 transforms reversibly from a brownmillerite-like structure to defective perovskite at about Tt = 900~ in air. Kruidhof et al. [154] observed that the transition temperature is not, or only slightly, affected if SrCoO3_ a is substituted with either 20 mol% Cr or Cu at the Co-sites. Interesting to note is that the oxygen flux for the undoped and doped specimens is very small below T t, as expected for an ordered arrangement of oxygen vacancies, but is found to increase sharply (between 5-6 orders of magnitude) at the onset of the phase transition to defective perovskite, up to values between 0.3-3 x 10 -7 mol c m -2 s -1. In view of these results, the perovskite phase in SrCoO3_ a s e e m s to be stabilized by the partial substitution of Co with Fe, but not with Cu or Cr, thereby suppressing the brownmillerite-perovskite two phase region to lower oxygen partial pressures.
10.6.7.2 Experimental Difficulties In a number of studies, the oxygen fluxes through, e.g., SrCo0.sFe0.203_a have been reported to be significantly lower than claimed by Teraoka et al. [37]. Such
504
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
conflicting results reflect the difficulties in measuring the oxygen fluxes at high temperatures and may, at least partly, be due to specific conditions, including (1) edge-effects associated with the required sealing of sample discs to avoid gas bypassing, giving rise to non-axial contributions to the oxygen flux, (2) possible interfacial reactions when a glass is used for sealing, (3) undesired spreading of the glass seal (when its softening temperature is too low) over the oxide disc surface, and (4) the precise value of the Po2-gradient across the membrane. With regard to the first point, it is frequently the cross-sectional area of the disc that is used in the calculation of the oxygen flux. In the usual experimental arrangements however an appreciable portion of the membrane is 'clamped' between impermeable annular plates or glass rings. This edge effect means that the usual assumption of one-dimensional diffusion is not strictly correct. Another contribution to non-axial transport is that of flow of oxygen through the side walls of disc specimens, if left uncovered. Appreciable errors creep in if these edge effects are neglected as shown, for example, on the basis of a solution of Fick's second diffusion equation (with a constant diffusion coefficient) by Barrer et al. [243]. This is further demonstrated in Fig. 10.17, showing the effect of sealing edges on the departure from one-dimensional diffusion. These results were obtained from a numerical procedure to solve the steady-state diffusion equation in cylindrical coordinates [112]. Neglecting edge effects corrupts analysis of experiments in which the membrane thickness is varied, and may lead to erroneous conclusions when one tries to infer from the acquired data the influence of the surface exchange kinetics on overall oxygen permeation. Finally, it cannot be excluded that the observed oxygen fluxes are specific for the particular sample under investigation and may be affected, for instance, by microstructural effects, a point to which we return in Section 10.6.7.5. The gas flow rate of, in particular, the inert gas used to sweep the oxygen-lean side of the membrane affects the Po~ -gradient across the membrane. Under ideal gas mixing conditions, the Po2 at the oxygen-lean side of the membrane is determined by the amount of oxygen permeating through the membrane. If the flow rate is not adjusted to obtain a constant Po2 at this side of the membrane, but a constant gas flow rate is used, the Po2-gradient gets smaller with increasing oxygen flux. This may give rise to an apparent activation energy for overall permeation, which may depart significantly from the one derived if a constant Po2 were maintained at this side of the membrane [148,149,153]. The adjustable range of the sweeping gas flow rate (to a constant Po2 at the outlet of the reactor) may be limited during experiment, being determined by the requirement that the reactor behaviour remains close to that of a CSTR (continuous stirred tank reactor). Using a constant value of Po2 at the oxygen-lean side of 2 m m thick disc membranes of Lal_~SrxCoO3,s (x = 0.2, 0.3, 0.4, 0.5 and 0.6), Van Doorn et al. [148,149] showed that the activation energy Eact for oxygen permeation in the
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
505
2b
(a)
membrane r
seal
2a
(b)
1.40
1.30~
3.75
C9 t_
0
dk.J
o r o
. I t_.
1.20 7.5
(1)
E 0
ID 1.10
1.00
, 0.70
, 0.80
,
, 0.90
, 1.00
a/b Fig. 10.17. (a) Schematic cross-section of a disk membrm~e. Dashed parts indicate insulating boundaries. (b) Influence of sealing edge-effects oil the departure from one-dimensional diffusion. A geometric factor G is used for correction of the flux (normalized to surface area with diameter 2a). Relevant parameters are defined in Fig. 10.17a.
r a n g e 900-1100~ decreases from 164 kJ mole -1 for x = 0.2 to 81 kJ mole -1 for x = 0.6. O p p o s e d to these results, Eact decreased from 121 kJ mole -1 to 58 kJ mole -1 w h e n a constant gas flow rate o f the h e l i u m w a s used. Besides an i m p r o v e d fit to the A r r h e n i u s equation in the former case, Eact can be correlated w i t h the s u m of the enthalpies for m i g r a t i o n a n d that for the formation of oxide ion vacancies for each of the investigated compositions. Such a correlation is expected if o x y g e n t r a n s p o r t is d r i v e n b y the g r a d i e n t in o x y g e n nonstoichiom e t r y across the m e m b r a n e d u e to the i m p o s e d P o d g r a d i e n t . It suggests that o x y g e n vacancies are free and non-interactive in Lat_xSrxCoO34 u n d e r the con-
506
10 ~ DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
ditions covered by experiment. Oxygen permeation fluxes for strontium-doping levels above x = 0.6 were found to be partially controlled by the surface exchange kinetics, as already mentioned in Section 10.6.4. In contradiction to the observed behaviour at high temperatures, results from thermal analysis and oxygen permeation measurements indicated that a phase transition, with a small first order component, probably related with order-disorder of oxygen vacancies, occurs in selected compositions Lal_xSrxCoO3,s in the range 750-775~ [148,149]. Long times extending to over 30 h were needed for equilibration towards steady-state oxygen pemeation at these modest temperatures. Such a behaviour is reminiscent of that observed for SrCo0.8Fe0.203@ where this can be attributed to the slow kinetics of the transformation between the brownmillerite and perovskite phases at modest temperatures. In the case of La1_vSrxCoO34 (x = 0.50 and 0.70), microdomains were observed in electron diffraction and HRTEM, corresponding to ordered arrangements of oxygen vacancies in these compounds at room temperature, as mentioned in the previous section. Another factor that is considered to be responsible for a reduced oxygen flux is the surface modification of the perovskite oxide membrane by reaction with impurities in the gas phase, as emphasized by Qiu et al. [153]. Referring to the surface degradation by reaction with minor amounts of CO2 and corresponding deterioration of the properties observed for YBa2CuO6+ x superconducting thin films [244], a similar modification effect could occur when, e.g. ambient air is used as the source of oxygen at the membrane feed side. With the help of N2 and 02 admixed to feed side pressure Po2' = 0.21 atm, Qiu et al. found the oxygen fluxes through SrCo0.8Fe0.2034 in the range 620-920~ to be larger by a factor of about 6 than when ambient air was used as feed gas, but still a factor of about 5 smaller than measured by Teraoka et al. Similar experiments were conducted in our study on SrCo0.8Fe0.203,s [41,154], where this effect was not noted in the temperature range 700-950~ so that we are inclined to believe that other factors must account for the disagreements in oxygen fluxes. This interpretation is supported by experimental evidence disclosed in a number of patents: that the oxygen fluxes through perovskite membranes remain stable as long as these are operated above certain critical temperatures, the precise value depending on the type of alkaline-earth dopant applied. Below these temperatures, a loss in oxygen flux may be observed over a period of about 100 h by as much as 30-40% when a membrane is exposed to CO2 and H20 impurities in the feed gas. This is further exemplified in Section 10.7.
10.6.7.3 Surface Exchange Kinetics Attention has already been drawn to the importance of the surface exchange kinetics in determining the rate of oxygen permeation through mixed-conduct-
10 m D E N S E C E R A M I C M E M B R A N E S F O R O X Y G E N S E P A R A T I O N
507
ing oxides in Section 10.3.2.2. Though for the perovskites a value of 100 ~tm is often quoted for the characteristic membrane thickness Lc, at which the change over from bulk to surface control occurs, in a number of cases much higher values are found, up to about 3000 ~tm (Table 10.2). As was emphasized earlier, the parameter Lc is not an intrinsic material property and, hence, may be specific to the sample under investigation and experimental conditions. The basic assumptions made in the derivation, notably that of small Po -gradients across the membrane, may restrict its use in practical situations, where these gradients can be substantial. Experimental evidence that the oxygen fluxes are limited by the surface exchange kinetics has been found in a number of cases, as discussed elsewhere in this text.
10.6.7.4 Behaviour in Large Po2-Gradients The mixed-conducting perovskite oxides have attracted particular interest for use as dense ceramic membrane to control partial oxidation of methane to C2-products or syngas. Such a process bypasses the use of costly oxygen since air can be used as oxidant on the oxygen-rich of the membrane. Using SrCo0.8Fe0.2034 tubular membranes fabricated by an extrusion method, Pei et al. [13] observed two types of fracture of the tubes during the process for generating syngas. The first fracture, occurring short (within 1 h) after initiation of the reaction at 800~ resulted from the Po -gradient across the membrane and the accompanying strain due to lattice mismatch and the brownmillerite-perovskite phase transition. The second type offracture, occurring after prolonged exposure to the reducing environment, resulted from chemical decomposition towards SrCO3, and elemental Co and Fe. Similar observations have been reported for tubes made of La0.2Sr0.8Co0.4Fe0.603-~ [14], and in that study an optimized composition was also claimed, but not given, showing stable performance for up to 500 h. Using a rhodium-based reforming catalyst inside the tubes, methane conversions over 99% were achievable. Ten Elshof et al. [10] studied the oxidative coupling of methane using a disc reactor with La0.6Sr0.4Co0.sFe0.2034 as the catalyst membrane for the supply of oxygen to the methane feed stream. Examination of the oxygen fluxes measured under various Po2-gradients in the range of thickness 0.55-0.98 mm suggested that the surface exchange reaction limits the rate of oxygen permeation. The oxygen flux was found to increase only slightly when methane was admixed with the helium used as the carrier gas. The methane was converted to ethane and ethene with selectivities up to 70%, albeit with a low conversion, typically in the range 1-3% at operating temperatures 1073-1173 K. The selectivity observed at a given oxygen flux and temperature was about twice as low if the same amount of molecular oxygen was co-fed with the methane feed stream in a single chamber reactor design, suggesting that the membrane-mode of operation
508
1 0 - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
is conceptually more attractive for generating C2-products. Decomposition of the oxide surface did not occur as long as molecular oxygen could be traced at the reactor outlet, which emphasizes the importance of surface-controlled oxygen flux for membrane-driven methane coupling. That is, for a bulk diffusion-controlled oxygen flux the surface would become reduced by the methane, until the depth of reduction has progressed up to a point where the oxygen flux counterbalances the consumption of oxygen by methane. On the one hand the slow surface exchange kinetics observed on La0.6Sr0.4Co0.aFe0.2034limits the magnitude of the oxygen fluxes, on the other hand its existence prevents the oxide surface from reduction, i.e. as long as the rate of oxygen supply across the membrane exceeds the rate of (partial) oxidation of methane. Noteworthy is that segregation of strontium occurred on both sides of the membrane, as confirmed by depth-profiling Auger analysis. The extent of segregation appeared to be influenced by the imposed Podgradient across the membrane, and was also found if a pure helium stream was passed along the oxygen-lean side of the membrane. Van Hassel et al. [150] studied oxygen permeation through Lal.~SrxFeO3,s (x = 0.1, 0.2) membranes in a disc reactor using CO-CO2 based gas mixtures to control the Po2 at the oxygen-lean side. Ambient air was used as the oxygen source at the opposite side of the membrane. At 800-1100~ the oxygen flux was found to increase linearly with the partial pressure of CO. Deposition of a 50 nm thin porous Pt layer on this side of the membrane increased the oxidation rate and likewise the oxygen flux, by a factor of about 1.8. In a separate study [245], the oxygen flux was found to be invariant with the thickness of the membrane in the range 0.5-2.0 ram, while no effect was observed upon varying the Po~ at the oxygen-rich side. It was concluded that the oxygen flux is fully limited by the carbon monoxide oxidation rate. The experimentally determined rate constants scale with Sr-content in the extended range of composition 0.1 < x < 0.4. The latter can be accounted for, in view of the fact that the oxygen deficiency of the ferrites is fixed by the dopant concentration in a wide range of oxygen partial pressure, by assuming that oxygen vacancies act as active sites in the oxidation reaction of CO on the perovskite surface following either an Eley-Rideal or a Langmuir-Hinselwood type of mechanism.
10.6.7.5 Grain Boundary Diffusivity Besides the possibility of surface exchange limitations, oxygen transport through dense ceramics is necessarily influenced by the presence of high diffusivity paths along internal surfaces such as grain boundaries. A systematic study investigating to which extent these preferred diffusion paths contribute to the diffusivity in the perovskite oxides is however still lacking. Both impurity and solute segregation take place at grain boundaries (and the external surface) or in their close proximities (less than 3 or 4 atomic distances) during sintering
10 m DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
509
and subsequent heat treatments. An obvious consideration is that, in general, these significantly alter the magnitude of ionic transport along and across the grain boundaries. In many cases the ceramics invariably contain impurities present in the starting powder or added as a sintering aid to lower the sintering temperature and/or to achieve high density. It therefore can not be excluded that disagreements in the literature regarding the magnitude of the oxygen fluxes can be explained on the basis of different ceramic processing techniques used by various authors. In general, the presence of high diffusivity paths is important in ceramics where lattice diffusion is slow. Analyzing 180 depth profiles using secondary ion mass spectroscopy, Yasuda et al. [246] noted a significant contribution of the grain boundary diffusion to the diffusivity in the interconnect material La0.?Ca0.35CrO34, where the tracer diffusivity is of the order of ~10 -13 cm 2 s -1 at 900~ Erroneous results were obtained when isotopic exchange was performed by gas phase analysis, which resulted in apparent tracer diffusion coefficients that were almost 2 orders in magnitude higher. More recently, Kawada et al. [184] confirmed the existence of high diffusivity paths along grain boundaries in La0.~Ca0.3CrO34 using depth profiling and imaging SIMS of lso-160 exchanged specimens. But, oxygen permeation measurements suggested negligible contribution of grain boundary diffusion to the steady-state oxygen flux. These data were obtained at 1000~ for a sample of thickness 0.75 mm. An oxygen pump and sensor were used to control the permeate side Po2. The results are well-described by the Wagner equation assuming a random point defect scheme for La0.~a0.3CrO34, as discussed in Section 10.6.2.2. For fast ionic conductors grain boundary diffusion will have little influence, or indeed may become blocking to the diffusion from one grain to the next as is recognized in the interpretation of impedance spectra from ionic conductivity of zirconia and ceria-based solid electrolytes. In these ceramics silicon is the most common impurity detected along with enhanced yttrium segregation. Various models to account for the effects of segregation at grain boundaries and how these affect the electrical properties have been discussed by Badwal et al. [247]. Although there is no unique model describing the ceramic microstructure, the most widely adopted model for doped zirconia and doped ceria is the brick-layer model. In this model bricks present the grains and mortar the grain boundary region, i.e. assuming the grain boundary phase to completely wet the grains [248,249]. The grain boundaries in series with the grains, along the direction of charge flow, mainly contribute to the grain boundary resistivity. For doped zirconia and ceria the grain boundary resistivity can be of similar order of magnitude or higher than the bulk resistivity. Pores at the grain boundaries can have a positive effect on the oxygen transport. It is evident that more detailed studies are needed to aid in the interpretation of oxygen transport through the mixed-conducting perovskite oxides, where similar blocking effects can be expected.
510
10.7
10 - - D E N S E C E R A M I C M E M B R A N E S F O R O X Y G E N S E P A R A T I O N
FINAL
R E M A R K S
The considerations in this chapter were mainly prompted by the potential application of mixed-conducting perovskite-type oxides to be used as dense ceramic membranes for oxygen delivery applications, and lead to the following general criteria for the selection of materials - high electronic and ionic conductivity, - high catalytic activity towards oxygen reduction and reoxidafion, - ability to be formed into dense thin films, free of micro-cracks and connected-through porosity, chemical and structural integrity (i.e. no destructive phase transition) within appropriate ranges of temperature and oxygen partial pressure, - low volatility at operating temperatures, thermal and chemical compatibility with other cell components, low cost of material and fabrication. The precise perovskite composition may be tailored for a specific application. To obtain a high performance membrane, however, many technical and material problems remain to be solved. This final section will focus on several issues, which are not yet well understood, but are thought to be of importance for further development of the membrane devices. In the first place our understanding of factors that control and limit the interfacial kinetics is still rudimentary, and therefore should be a fruitful area for further investigation. The apparent correlation between the surface oxygen exchange coefficient ks and the tracer diffusion coefficient D* for different classes of oxides, the fluorite-related and the perovskite-related oxides, as noted by Kilner et al. [73], clearly indicate the potential of isotopic 180--160 exchange. However, a problem remains how to relate the observations (at equilibrium) from isotopic exchange to the conditions met during membrane operation. In chemical relaxation experiments, the oxide is studied after perturbation of the equilibrium state. These methods are thus complementary and probably their combined application, whenever possible together with spectroscopic techniques, such as FT-IR, UV and EPR, has a great capacity to elucidate the kinetics of surface oxygen exchange. Though, at first glance, the limited exchange capability of the perovskites, relative to diffusion, puts limits on attempts to improve the oxygen fluxes or to lower the operating temperatures by making thinner membranes, it is expected that the surface exchange kinetics can be significantly improved by surface modification. One approach is coating with a porous surface layer which will effectively enlarge the surface area available to exchange, as discussed in Section 10.3.2.3. Improvements can also be expected by finely dispersing precious metals or other exchange active second phases on the oxide surface. It is clear that further investigations are required to evaluate these innovative approaches.
10 - - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
511
As yet, more work is also required to gain insight in the role of the ceramic microstructure in the performance values of membranes, and to evaluate different processing routes for the fabrication of perovskite thin films. Besides the technological challenge of fabrication of dense and crack-free thin perovskite films, which need to be supported if its thickness is less than about 150 ~tm, a number of other problems relate to the long-term stability of perovskite membranes, including segregation, a low volatility of lattice components, etc. Some of these problems are linked to the imposed oxygen pressure gradient across the membrane. Aside from the lattice expansion mismatch of opposite sides of the membrane, attention is drawn to the potential problem of demixing, which arises in almost all situations where a multicomponent oxide is brought into a gradient of oxygen chemical potential. The available theories predict that, if the mobilities of the cations are different and non-negligible at high temperatures, concentration gradients appear in the oxide in such a way that the high oxygen pressure side of the membrane tends to be enriched with the faster moving cation species. Depending on the phase diagram, the spatially inhomogeneous oxide may eventually decompose. The latter may cause surprise, if the (homogeneous) oxide is stable in the range of oxygen partial pressures covered by experiment. This is why these processes have been termed kinetic demixing and kinetic decomposition by Schmalzried et al. [250,251] who were the first to study them. Degradation phenomena have been shown to occur in, for example, Col_xMgxO, Fe2SiO4 and NiTiO3. Internal oxidation or reduction processes sometimes lead to precipitation of a second phase in the matrix of the parent phase. Another possible consequence of the demixing process is the morphological instability of the (moving) low pressure interface due to formation of pores, which may eventually penetrate throughout the ceramic. The above phenomena have been the subject of a number of theoretical and experimental studies in the last decade [252-256], to give only a brief number. A review up to 1986 has been written by Schmalzried [257]. To our knowledge, no report has been made up to now of demixing phenomena in mixed oxide ion-electronic conductors. Since they cannot be excluded to occur on the basis of theoretical arguments, this is also why the phenomena deserve (more) attention in order to be able to control deterioration of membrane materials. Intergrowth structures in which perovskite-type blocks or layers are held apart by non-perovskite ones could offer a new strategy for identifying new materials, as was suggested earlier by Goodenough et al. [212]. In such structures, vacancy transport is confined to two-dimensional layers or to sites which link up to form channels extending throughout the crystal. An interesting variation to the BIMEVOX compounds, already discussed in Section 10.4.3.1, is found in derivatives of Sr4Fe6013. Its orthorhombic structure can be described as built of perovskite layers alternating with sesquioxide Fe203 layers perpendicular to the b-axis. The discovery of high levels of oxygen permeation through mixed
512
10 m DENSE C E R A M I C M E M B R A N E S F O R O X Y G E N S E P A R A T I O N
metal oxide compositions obtained by partial substitution of iron for cobalt, for instance, SrCo0.sFeOx recently translated into a patent for this class of materials [258]. Tubes made from the given composition showed oxygen fluxes similar to those through known state-of-the-art materials having a perovskite structure, but did not fracture in the process for preparing syngas as was found for some of the perovskite materials. As noted before, the membrane performance could be affected by the presence of H20, CO2 or other volatile hydrocarbons in the gas phase of both compartments. As laid down in patent literature [1-3], the oxygen fluxes through Mg-, Ca-, Sr-, and Ba-doped perovskites deteriorated over time, roughly 30-50% over a time period of about 100 h, if the air used as feed gas contained several percent of H20 and amounts of CO2 on a hundreds of ppm level. It was claimed, that either no deterioration is found or the fluxes can be restored to their initial values if the temperature is raised above certain critical values, 500~ for magnesium, 600~ for calcium, 700~ for strontium and 810~ for barium. Though no explanation was given, it is possible that carbonate formation took place. One may further note that the tendency for carbonate formation increases at lower temperatures. A surprising observation was recently made in the author's laboratory in a study of oxygen permeation through Lal_xSrxFeO34 (0.1 < x < 0.4) [151]. Long times to reach steady-state oxygen permeation at 1000~ extending over hundreds of hours were observed, yet could be avoided by exposing the permeate side surface of the membrane for a 1-2 h to 1:1 CO/CO2 gas mixture. A clear explanation cannot yet be given for this observation, which is still under investigation, though a reconstruction of the surface by the reducing ambient cannot be excluded. The oxygen permeability measured if helium was used again as the sweeping gas on this side of the membrane, was found to be limited by diffusional transport of oxygen across the membrane [151]. A similar type of observation was made by Miura et al. [152], who noticed the oxygen flux through slib-casted membranes of La0.6Sr0.4Co0.sFe0.203,stobe greatly improved if these were freed from surface impurities, like SrO, following an acid treatment. One final point to note is the ability of acceptor-doped perovskite oxides to incorporate water, and some contribution of proton conduction therefore cannot be excluded. If water insertion occurs at low temperature, this might lead to residual stresses in the ceramics. Besides water may play an active role in the surface oxygen exchange. For example, on Bi2MoO6, which has an intergrowth structure consisting of Bi2O2+ blocks alternating with MO42- layers of cornershared MO 6 octahedra, exchange with 1802-enriched oxygen could not be observed experimentally [259]. On the other hand, Novokova and Jiru [260] demonstrated that exchange of water with lattice oxygen on an industrial bismuth molybdate catalyst proceeds rapidly at 200~ and is even measurable at room temperature.
1 0 - DENSE CERAMIC MEMBRANES FOR OXYGEN SEPARATION
513
Acknowledgements The authors are indebted to colleagues H. Kruidhof, R.H.E. van Doorn, J.E. ten Elshof, M.H.R. Lankhorst and B.A. van Hassel for many useful discussions and for providing experimental data. Paul Gellings and Henk Verweij are gratefully acknowledged for valuable comments and careful reading of the manuscript. The Commission of the European Communities and the Netherlands Foundation for Chemical Research (SON) are thanked for financial support.
List of Abbreviations and Symbols
Abbreviations: BE25 25 mole% erbia-stabilized bismuth oxide BICUVOX Bi4V2-yCUyOll BIMEVOX general acronym for materials derived from Bi4V2O11, like BICUVOX BT40 40 mole% terbia-stabilized bismuth oxide BY25 25 mole% yttria-stabilized bismuth oxide CSZ calcia-stabilized zirconia ECVD electrochemical vapour deposition EDS energy dispersive spectroscopy (of X-rays) EPR electron proton resonance electromotive force emf fourier transform infrared spectroscopy FT-IR high resolution transmission electron microscopy HRTEM mixed ionic-electronic conductor MIEC nuclear magnetic resonance NMR scanning electron microscopy SEM secondary ion mass spectroscopy SIMS solid oxide fuel cell SOFC transmission electron microscopy TEM three phase boundary tpb temperature programmed reduction TPR ultra-violet spectroscopy UV X-ray absorption near edge structure XANES X-ray absorption spectroscopy XAS X-ray diffraction XRD yttria-stabilized zirconia YSZ
Symbols: Ci N
D D*
mole fraction or concentration of species i chemical diffusion coefficient tracer diffusion coefficient
514
Ds Dv dp e
E Eeq
F G
I
ji .0
Jex
ks k K L Lc Ld Lp n
P Po2 PO2' Po2"
R Si o si
S t tel
ti tion T Tt ui o bli
Vm zi
10- DENSECERAMICMEMBRANESFOROXYGENSEPARATION
self-diffusion coefficient vacancy diffusion coefficient pore diameter elementary charge
emf emf at equilibrium Faraday constant geometric factor used to account for non-axial contributions to the oxygen flux Haven ratio electrical current flux of species i balanced surface exchange rate atequilibrium, mol 02 c m -2 s -1 surface exchange coefficient, cm s reaction rate constant equilibrium constant for a reaction membrane thickness characteristic thickness of membrane Debye-Hiickel screening length characteristic thickness (active width) of porous coating layer frequently used to designate the mole fraction of electrons, yet its use is multipurpose mole fraction of electron holes oxygen partial pressure oxygen partial pressure at feed side of the membrane oxygen partial pressure at permeate side of the membrane radius of species i gas constant entropy of species i entropy of species i at standard state surface area Goldschmidt factor electronic transference number transference number of species i ionic transference number temperature transition temperature electrical mobility of species i electrical mobility of species i in standard state molar volume charge number of species i (positive for cations and negative for anions)
10 m D E N S E C E R A M I C M E M B R A N E S F O R O X Y G E N S E P A R A T I O N
Greek: (x
8
1~i 0 o
(~el (3"h o
(~ion (~n (3"p (~total Xs
515
surface exchange coefficient bulk diffusion coefficient reduction factor deviation from ideal oxygen stoichiometry e n h a n c e m e n t factor overpotential electrochemical potential of species i porosity chemical potential of species i s t a n d a r d chemical potential of species i electronic conductivity polaron h o p p i n g conductivity electrical conductivity of species i conductivity of species i at s t a n d a r d state ionic conductivity n-type electronic conductivity p-type electronic conductivity total conductivity tortuosity electric potential of phase (Galvani potential) critical (percolation threshold) v o l u m e fraction
REFERENCES
1.
2.
3.
4. 5. 6. 7. 8.
M.F.Carolan, P.N. Dyer, J.M. LaBar Sr. and R.M. Thorogood, Process for recovering oxygen from gaseous mixtures containing water or carbon dioxide which process employs ion transport membranes. US Patent 5,261,932, 1993. M.F. Carolan, P.N. Dyer, S.M. Fine, J.M. LaBar Sr. and R.M. Thorogood, Process for recovering oxygen from gaseous mixtures containing water or carbon dioxide which process employs barium-containing ion transport membranes. US Patent 5,269,822, 1993. M.F. Carolan, P.N. Dyer, J.M. LaBar Sr. and R.M. Thorogood, Process for restoring permeance of an oxygen-permeable ion transport membrane utilized to recover oxygen from oxygen-containing gaseous mixtures. US Patent 5,240,473, 1993. M. Liu, A.V. Joshi, Y. Shen and K. Krist, Mixed ionic-electronic conductors for oxygen separation and electrocatalysis. US Patent 5,273,628,1993. R.M.Thorogood, Developments in air separation. Gas Sep. Purif., 5 (1991) 83-94. E.A.Hazbun, Ceramic membrane for hydrocarbon conversion. U.S Patent 4,791,079,1988. E.A. Hazbun, Ceramic membrane and use thereof for hydrocarbon conversion. U.S. Patent 4,827,071, 1989. W. Wang and Y.S. Lin, Analysis of oxidative coupling in dense oxide membrane reactors, J. Membr. Sci., 103 (1995) 219-233.
516
10-- DENSECERAMICMEMBRANESFOROXYGENSEPARATION
9. J.E. ten Elshof, B.A. Van Hassel and H.J.M. Bouwmeester, Activation of methane using solid oxide membranes. Catal. Today, 25 (1995) 397--402. 10. J.E. ten Elshof, H.J.M. Bouwmeester and H. Verweij, Oxidative coupling of methane in a mixed-conducting perovskite membrane reactor. Appl. Catal. A: General, 130 (1995) 195-212. 11. T. Nozaki and K. Fujimoto, Oxide ion transport for selective oxidative coupling of methane with new membrane reactor. AIChE J., 40 (1994) 870-877. 12. C.-Y. Tsai, Y.H. Ma, W.R. Moser and A.G. Dixon, Simulation of nonisothermal catalytic membrane reactor for methane partial oxidation to syngas, in: Y.H. Ma (Ed.), Proceedings of the 3rd International Conference on Inorganic Membranes, Worcester, 1994, pp. 271-280. 13. S. Pei, M.S. Kleefisch, T.P. Kobylinski, K. Faber, C.A. Udovich, V. Zhang-McCoy, B. Dabrowski, U. Balachandran, R.L. Mieville and R.B. Poeppel, Failure mechanisms of ceramic membrane reactors in partial oxidation of methane to synthesis gas. Catal. Lett., 30 (1995) 210-212. 14. U. Balachandran, J.T. Dusek, S.M. Sweeney, R.B. Poeppel, R.L. Mieville, P.S. Maiya, M.S. Kleefisch, S. Pei, T.P. Kobylinski, C.A. Udovich and A.C. Bose, Methane to synthesis gas via ceramic membranes. Am. Ceram. Soc. Bull., 74(1) (1995) 71-75. 15. A.G. Dixon, W.R. Moser and Y.H. Ma, Waste reduction and recovery using O2-permeable membrane reactors. Ind. Eng. Chem. Res., 33 (1994) 3015-3024. 16. H.P. Hsieh, Inorganic membrane reactors. Cat. Rev.-Sci. Eng., 33(1/2) (1992) 1-70. 17. W.R. Gurr, An operators view on gas membranes, in: M.K. Turner (Ed.), Effective Industrial Membrane Processes m BeneJits and Opportunities. Elsevier, London, 1991, p. 329. 18. V.T. Zaspalis and A.J. Burggraaf, Inorganic membrane reactors to enhance the productivity of chemical processes, in: R.R. Bhave (Ed.), Inorganic Membranes, Synthesis, Characteristics and Applications. Van Nostrand Reinhold, New York, 1991, pp. 177-207. 19. T.J. Mazanec, Prospects for ceramic electrochemical reactors in industry. Solid State Ionics, 70/71 (1994) 11-19. 20. G. Saracco and V. Specchia, Catalytic inorganic membrane reactors: Present experience and future opportunities. Catal. Rev.-Sci. Eng., 36(2) (1994) 303-384. 21. P.J. Gellings and H.J.M. Bouwmeester, Ion and mixed-conducting oxides as catalysts. Catal. Today, 1 (1992) 1-101. 22. D. Hayes, D.W. Budworth and J.P. Roberts, Selective permeation of gases through dense sintered alumina. Trans. Br. Ceram. Soc., 60 (1961) 494-504. 23. D. Hayes, D.W. Budworth and J.P Roberts, Permeability of sintered alumina materials to gases at high temperatures. Trans. Br. Ceram. Soc., 62 (1963) 507-523. 24. H.L. Tuller, Mixed conduction in nonstoichiometric oxides, in: O. Sorenson (Ed.), Nonstoichiometric Oxides. Academic Press, New York 1981, pp. 271-337. 25. J.W. Suitor, D.J. Clark and R.W. Losey, Development of alternative oxygen production source using a zirconia solid electrolyte membrane, in: Technical progress report for fiscal years 1987, 1988 and 1990. Jet Propulsion Laboratory Internal Document D7790, 1990. 26. T.J. Mazanec, T.L. Cable and J.G. Frye, Jr., Electrocatalytic cells for chemical reaction. Solid State Ionics, 53-56 (1992) 111-118. 27. M. Kleitz and M. Siebert, Electrode reactions in potentiometric gas sensors, in: T. Seiyama (Ed.), Chemical Sensor Technology, Vol. 2. Elsevier, Amsterdam, 1989, pp. 151-71. 28. W.C. Maskell and B.C.H. Steele, Solid state potentiometric sensors. J. Appl. Electrochem., 16 (1986) 475-489.
10 -- DENSECERAMICMEMBRANESFOROXYGENSEPARATION 29. 30.
31. 32. 33. 34. 35. 36. 37. 38. 39.
40. 41.
42. 43. 44. 45. 46. 47.
48.
49. 50. 51. 52.
517
S.S. Liou and W.L. Worrell, Electrical properties of novel mixed-conducting oxides. Appl. Phys. A., 49 (1989) 25-31. S.S. Liou and W.L. Worrell, Mixed conducting electrodes for solid oxide fuel cells, in: S.C. Singhal (Ed.), Proceedings of the 1st International Symposium on Solid Oxide Fuel Cells. The Electrochem. Society, Pennington, NJ, 1989, pp. 81-89. B. Cal6s and J.F. Baumard, Electrical properties of the ternary solid solutions (ZrO21-xCeO2 x)0.9-Y203 0.1 (0 < x _, x 0
Fig. 11.9. Methane conversion, CO mid H2 selectivities and 02 permeation in a solid oxide m e m b r a n e reactor. R e p r o d u c e d from Balachandran et al. [113] with permission.
11 - - CURRENT DEVELOPMENTS AND FUTURE RESEARCH IN CATALYTIC MEMBRANE REACTORS
549
of the membrane. Figure 11.9 shows the CH4 conversion and CO/H2 selectivities during a 21 day run. They all remain in the 90+% range throughout the whole run. The use of solid oxide membranes in partial oxidation reactions aims to avoid the complete oxidation of the desired partial oxidation products. When compared to similar efforts using microporous membranes solid oxide membrane reactors, are at a disadvantage (except for reactions that take place at high temperatures) because oxygen transport through the oxide lattice is generally low when compared with the permeability of porous materials. The synthesis of non-porous ceramics with good oxygen permeability and selectivity at lower temperatures is not a simple task. The application of membrane reactors to partial oxidation is often complicated by the fact that the desired product is often more reactive with oxygen than the reactant itself as was observed by Julbe et al. [114,115] for the methane oxidative coupling using lanthanum oxychloride membranes.
11.5 THEORETICAL C O N S I D E R A T I O N S
The modelling and simulation of catalytic membrane reactors has attracted the interest of many investigators over the last ten years. Most studies have focused on particular membrane reactor systems aiming to simulate their performance in terms of attainable yield and selectivities. The considerable body of modelling work in this area has been reviewed by Tsotsis et al. [13]. The authors of this paper presenteda discussion of the pre-1993 modelling literature in terms of a general mathematical model of a PBCMR, which is shown schematically in Fig. 11.10, with catalytic beds present both in the inner and outer membrane regions. The model takes into account mass and energy balances in the tubeside
~"
OUTER TUBE
z IS
~MBRANE
TUBESm~
CATALYST BED
SHELLSIDE CATALYST BED
Fig. 11.10. Schematic of m e m b r a n e reactor for PBCMR model. R e p r o d u c e d from Tsotsis et al. [13] with permission.
550
11 ~ CURRENT DEVELOPMENTS A N D FUTURE RESEARCH IN CATALYTIC MEMBRANE REACTORS
and shellside and in the membrane itself and accounts for the existence of pressure drops in the shellside and tubeside. The membrane is considered to consist of a single permselective layer either dense or mesoporous following a Knudsen type diffusion mechanism, an assumption utilized by most pre-1993 modelling investigations. There are a number of modelling efforts, however, which cannot be discussed in the context of the generalized model of Tsotsis et al. [13]. Van Swaaij and coworkers, for example, have modeled the behaviour of CNMR reactors using the Dusty Gas Model description of transport [5,6]. They have shown that when bulk diffusion and convective flows must be taken into account the Dusty Gas Model provides a more accurate description of transport through the membrane. The earlier studies of the group modelling the application of CNMR to reactions requiring strict stoichiometric ratios have been reviewed in detail by Tsotsis et al. [13]. More recent efforts by the same group deal with the application of the CNMR Dusty Gas Model to the combustion of CO and hydrocarbons. Membrane reactors utilizing multilayered membranes have been modelled in recent studies by Becker et al. [116] and Tayakout et al. [117,118]. In contrast to prior efforts these models account for mass transport both through the mesoporous permselective layer and the underlying macroporous support layer(s). Both are isothermal models. Becker et al. [116], however, in their analysis utilized the experimentally measured temperature gradients along the reactor length in the calculation of reaction constants and transport coefficients. Both models assume dilute reactant mixtures and, therefore, neglect complications resulting from changes in the number of moles due to the reaction. The reaction studied was ethylbenzene dehydrogenation in the model of Becker et al. [116] and cyclohexane dehydrogenation in the model of Tayakout et al. [117,118]. A schematic of the reactor analyzed by both groups is shown in Fig. 11.11 (in the model of Becker et al. [116] there is no catalyst bed in region 4). At steady-state Becker et al. [116] write the following set of equations.
On the tubeside (region 1)
3C~ 1 3 ! 3Ct] rl U T --~- = D t r -~r r ---~-r ) p b O~i O cyo > CYl~
layer
.
~
v
~d I "
super ,er-equivalent :=> I~01> )ecific adsorption soeciJ
I ,1
Fig. 12.11. Representation of potential evolution in a perpendicular direction to the oxide surface when counter ions are specifically adsorbed on the surface.
to the pore wall. This potential, called zeta potential, (~d, c a n be related both to the characteristics of metal oxide membrane and to the feed ionic strength. According to the space charge model (SC), when a solution is flowing in the porous structure under a pressure gradient, pore wall is reduced to the Sternsurface between the static and the mobile portions of solution. The pore radius equivalent to the Stern-surface is called hydrodynamic radius rh with rh = rp- l
(12.24)
where rp is the original pore radius and I usually taken as a counter-ion diameter. So that when a pressure gradient, Ap, acts through the membrane, the solution close to the pore wall stays immobile while the rest of it moves along the pore. This movement leads to the appearance of an electrical potential drop from one side to the other side of the membrane, A~g.This electrical potential results from an electrical field which develops because the flux of the counter-ions is greater than that of the coions into the membrane pores. This electrical field generates an electrokinematic flow of the counter-ions that is opposed to the previous one thereby satisfying the constraint that there is no net current flow through the membrane. The combination of ~)(r) and A~geffects corresponds to a dynamical contribution to the total electric potential profile, according to the space charge model which was originally proposed by Osterle et al. [26-28].
588
12 m TRANSPORT A N D FOULING P H E N O M E N A IN LIQUID PHASE SEPARATION
O(r,z) = (ziF/RT) ~(r) + ~ ( z )
(12.25)
The z-dependent potential ~(z), in zero current conditions, is related to streaming potential Vp. In the case of ceramic membranes ~)d and Vp are the two quantifies which influence both retention of charged solutes and volume flux. ~)d is related to ionic strength and pH of the feed solution; Vp is related to ionic strength but also to the transmembrane pressuredriven flow resulting from the pressure gradient Ap applied to the membrane. As a whole some general rules can be pointed out concerning the effect of ~)d and Vp on membrane behaviour. The spatial extent of the double layer in the radial direction of the pores is characterized by the Debye-length so as a high ionic strength leads to a short Debye-length and a weak electric effect on transport. The ionic strength is related to ions concentration in the feed solution but also to the valence of co- and counter-ions. Multivalent counter-ions which adsorb in the double layer have a more marked effect than monovalent counterions in diminishing the spatial extent of the double layer and the resulting zeta potential. On the contrary, due to their higher electric charge, multivalent coions are more rejected than monovalent coions. Taking into account the distribution of charges in the radial direction of pores, volume flux in the axial direction can be described by the addition of two opposite contributions: the convective-diffusive flow and the back electrokinematic flow due to streaming potential. The electrokinematic flow for a porous ceramic membrane can be related to the following parameters: - -
-
4?s - (ziF/R T)(~d
(12.26)
a dimensionless zeta potential dependent upon ceramic surface characteristics and pH of feed solution; rE = r p / ~ D
(12.27)
a dimensionless length also called electrokinetic radius accounting for the ionic strength of ionic feed solution; Ls = Lp/Lo
(12.28)
a dimensionless hydraulic permeability resulting from the variation of rE and ~)s. L0 is the pure water permeability for the membrane with the same pore radius and a zero surface charge density while Lpis the membrane permeability in presence of a counter-electrokinematic flow. Due to Donnan exclusion principle [29] charged membranes can reject inorganic salts even though they have pores much larger than the salts and this ion rejection is known to decrease with increasing feed ionic strength. The example of 1.1 electrolyte filtration through different pore sizes at a pH far from the
121 TRANSPORTANDFOULINGPHENOMENAIN LIQUIDPHASESEPARATION Dolman b--x~ effect
Streaming
bound~ ~ layer
membrane
boundary layer
,-.......
1
589
rp~l nanofiltration
n+@
@ C )~~, .' -.@ . . . - @ : [ :@ " . .~ _ @ ~ @ @
o
Stem-layer
.........| ..........~ . ~ / / / / / / / / ~ ~ rp>l ultrafiltration
............. .| ........ | .i__~ |174 @.@.@. o.:..,~-...,.,,--@.@__| @ !]..........@ G | =.=..-:::,.,7.=.i-.-. | |
e -~ ~ ~ ~ ~ 1 | .......G . . . . . e . . . . . |
|
|174g4:::~---"-:::G-~ | |174
G _ N ~ ...@ ~ @ ...
..... |
G
| |
|
o
rp>> 1 microfiltration
e
convective flow ~ -.-.:r,~--- electrokinematic flow Fig. 12.12. Influence of zeta-potential (Stem-layer thiclaless l) and Streaming-potential (electrokinematic flow) on ion rejection and volume flux for porous ceramic membranes exhibiting negatively charged pore walls. Cases of micropores (nanofiltration), mesopores (ultrafiltration) and macropores (microfiltratio11).
isoelectric point (high zeta-potential for ceramic m e m b r a n e materials) is given in Fig. 12.12. According to the above described dimensionless p a r a m e t e r s the occurrence of electrokinematic flow (Ls < 1) is expected for rE < 1. Usually electrokinematic effect is likely to occur for nanofiltration and ultrafiltration
590
12 - - TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
membranes when deci- to centimolar electrolyte solution are used. Nevertheless this effect can be expected for microfiltration membranes in the case of very low ionic strength resulting in a Debye-length which were calculated to reach several tens of nm [30-32].
12.3 R E C E N T D E V E L O P M E N T S ULTRAFILTRATION
IN MICROFILTRATION
WITH CERAMIC
AND
MEMBRANES
Ceramic membranes were first applied to microfiltration processes. Several authors published a comprehensive description of basic transport phenomena involved in ceramic macroporous structures [1,33]. Lately improvements in ceramic membrane processing led to commercial ultrafiltration membranes exhibiting a mesoporous structure with transport phenomena close to those encountered in microfiltration. As described in Section 12.2, the major limitation in membrane performances for micro- and ultrafiltration processes is caused by concentration polarization and fouling. Methods that help to reduce concentration polarization and fouling can be classified into three categories: (i) chemical cleaning methods including strong acid and basic solutions or oxidizing agents due to high chemical resistance of ceramic membranes; (ii) physical methods such as backflushing and the use of turbulence promoters; (iii) hydrodynamic methods related to module design. In fact two aspects have been more specifically investigated in recent years concerning cross-flow filtration systems based on ceramic membranes: - the hydrodynamic of microfiltration and ultrafiltration systems and its influence on membrane performance in terms of fouling reduction and permeability enhancement; - the influence of membrane material (metal oxides in most cases) on selectivity and permeability.
12.3.1 Hydrodynamic of Micro- and Ultrafiltration systems In a review on cross-flow microfiltration Belfort et al. [34] outlined the importance of module design and hydrodynamic operating conditions in order to improve performances of cross-flow filtration using macroporous membranes. The authors suggest that unsteady flow conditions can be even more effective in disturbing the flux-limiting effects of concentration polarization and fouling [35]. Various approaches to inducing instabilities in bulk flow across a membrane surface include designing membrane surfaces with organized roughness, pulsation of axial and lateral flow, and the use of curvilinear flow under conditions that promote instabilities or vortices. A number of these devices shown in Figs. 12.14 and 12.15 can be adapted to ceramic membranes.
12 ~ TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
591
mlalltltJm~mmltl~,.-..~ r
porous wall
parabolic flow
with pulsation
Fig. 12.13. Effect of pulsations on flow profile in a s m o o t h - w a l l e d duct.
Permeate flux enhancement by pressure and flow pulsations has been investigated by many authors [36-38]. The effect of flow oscillations in a smoothwalled duct is shown in Fig. 12.13. P u l s a t e flows were applied to mineral microfiltrations membranes during apple juice filtration [36] illustrating the advantage of this method to enhance permeability compared to steady flow regime. With carefully chosen pulsations permeate flux increased up to 45% at I Hz pulsation frequency. Moreover well defined pulsations decreased the hydraulic power dissipated in the retentate per unit volume by up to 30%. In an other work on cross-flow filtration of plasma from blood [37] permeate flux increase was also observed when pressure and flow pulsations at I Hz are superimposed on the retentate. Pulsate flow can also be used to good advantage in rough walled ducts and those with inserts. Simulation of cross-flow filtration for baffled tubular channels and pulsate flow were reported by Wang et al. [38]. Wall and central baffles, in a similar way as in Fig. 12.14, with and without pulsations have been considered. Reynolds numbers up to 200 have been used in simulation that is notably lower than values used to obtain turbulent flow in smooth porouswalled channels. Concentration polarization effects have been included in calculation. The addition of pulsations improved the fluxes, the relative improvements being greater for the wall baffles. However the absolute values of the predicted fluxes were found greater for the central baffles. It has been suggested in the literature [34] that filtration devices producing Taylor or Dean vortices can help depolarization of the solute build up on membranes. This seems to be an attractive way because of excellent bulk fluid mixing, high wall shear rates and weakly decoupled cross-flow with transmembrane flux. Unfortunately there are some severe limitations on a technical and economical point of view with such devices. Build up and scale up of these modules are expensive with difficulties in repairing and changing membranes. A good compromise between economic and technical constraints has been described by Charpin et al. [39]. It consists in the preparation of mineral (metal
592
12 ~ TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
(a)
~
"
K
,
~
protuberance
~.x~-:-_--_--_--_--:__-_~_, ~ , ~
.....................
~ - ,.:...~,,:.?!:.-.:.2.;;::~;~;~"~~:7:~!:':".:;:~:?.: : "'.i~!~":.~":r>; I porous wail i
.....~-.~......................... . ] porous wall [,. . . . . . ~
~ (b)
------
| , s ,,. /.~,;:.
x
~
~
J
.
~'''--'--''~
I inserts
~
i ;.,;~ ;,:,,4.s. ;~,.~:,f.~.~,,:..,,,:.~.
:L',,:,:i:
~::., ',','.::',.
:;~',.:
..~,;i'.4.is.~i.:s:~.::Ss:.~:./:;:'.,'z.:
~ . . - . 5;~.: 9
t!,.!~.-.'~.-;_.:.;C::'_,:!.:_-_.,;..,7.~.:.:_:~.,::,."._~ :-_".~i..................... ;; ~::"F.":,'.'~:':":~"~:"Y~"~)~/......?;';"~:1
}porous wa. !
. . . .
Fig. 12.14. M e t h o d s for haducing flow instabilities: (a) plachag objects ( p r o t u b e r a n c e s , baffles) o n t o the m e m b r a n e surface to f o r m a r o u g h surface, (b) plachag objects hato the flow charmel a w a y f r o m the m e m b r a n e surface. flow x
,'."
.~'~'.'/:'/'
:,,..;.~,.~~.,~.~z , .: ". .",'..:..
"~ ~. " '.'," ' / .~,~,',
, .
/---
tubular helicoidal shape
,j
~.~
,
,~?,~..~'
Fig. 12.15. Schematic representation of a tube with ma il~ler surface exhibiting an helicoidal profile.
or ceramic) m e m b r a n e s exhibiting an inner helicoidal structure as s h o w n in Fig. 12.15. These helicoidal shaped tubes can be sealed in m o d u l e s in the same w a y than with classical tubes or monoliths.
593
12 m T R A N S P O R T A N D F O U L I N G P H E N O M E N A IN LIQUID P H A S E S E P A R A T I O N
Rotating disc systems have also been described as efficient devices to overcome flux limitation due to matter deposit on membrane surface during crossflow filtration [40]. The problem of erosion of a macroscopic particle solid deposit on a rotating disc membrane has been quantified by Aubert et al. [41]. The influence of the transmembrane pressure, the thickness of the initial deposit and the pore size on the critical shear stress have been investigated and described by empirical fits. It results from this study that fouling is more efficiently eliminated at high Ap and large pore size.
12.3.2 Influence of Membrane Material on Permeability and Solute Rejection The influence of metal oxide derived membrane material with regard to permeability and solute rejection was first reported by Vernon Ballou et al. [42,43] in the early 70s concerning mesoporous glass membranes. Filtration of sodium chloride and urea was studied with porous glass membranes in closeend capillary form, to determine the effect of pressure, temperature and concentration variations on lifetime rejection and flux characteristics. In this work experiments were considered as hyperfiltration (reverse osmosis) due to the high pressure applied to the membranes, 40 to 120 atm. In fact, results reproduced in Table 12.3 show that these membranes do not behave as hyperfiltration membranes but as membranes with intermediate performances between ultra- and nanofiltration in which surface charge effect of metal oxide material plays an important role in solute rejection. Rejection data for NaC1 were explained according to a low-capacity ion exchange mechanism. The ion exchange mechanism put forward in this work is not consistent with the porous structure of the membranes and the high transmembrane pressure used in the filtration experiments. Ion exchange TABLE 12.3 Rejection of NaC1 (58.5 g mole-1) and Urea (56 g mole-1) ushlg mesoporous glass membranes over a range of solute concentration, from Ref. [42] Solute
Feed concentration g/1
R tool/1
NaC1 NaC1
0.47 1.30
0.008 0.022
0.86 0.68
NaC1 Urea Urea Urea
9.11 1.74 3.80 11.32
0.156 0.029 0.063 0.189
0.42 0.41 0.38 0.37
594
12 - - TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
mechanism is better related to the working conditions described by Singh and Singh [44] for zirconium phosphate membranes. Regarding the work of Vernon Ballou [43] pore sizes were calculated from nitrogen isotherm data. Narrow pore volume distributions of unused glass membranes were found between 1.9 and 2.2 nm pore radius. One can see that NaC1 rejection cannot be explained by an hyperfiltration mechanism based on selective diffusion of water through the membrane. Moreover the decrease of NaC1 rejection when ionic strength of the feed solution increases is in favour of mass transport through an array of metal oxide microcapillaries filled with electrolytes. Unfortunately these glass membranes were not stable over a long period of time due to solubility of silica. Interpretation of the results were altered by a loss of rejection and an increase of permeability mainly caused by broadening of pore size distribution with time. Since then, the evidence of pH effect on cross-flow micro- or ultrafiltration using ceramic membranes has been given in the literature [45-48]. Hoogland et al. [45] showed that permeability of a Ceraflo (Norton) (x-alumina membrane towards pure water and mineral slurries was dependent of pH. For pure water the maximum of permeability was found in a pH range near the isoelectric point of the membrane (low pH) while the membrane resistance was highest far from the isoelectric point (high pH) when the charge of the membrane is strongly negative. At high pH, this effect can be explained by the flow through a charged porous barrier which leads to electro-osmosis phenomenon and an effective loss of permeability. Fluxes measured with mineral slurries (silica particles) were also dependent on pH. Higher permeabilities for the membrane were found at low and high pH while flux decline was maximum at intermediates pH. In this case two phenomena due to pH act simultaneously. One is related to the alumina membrane, the other to silica particles. At low pH near the isoelectric point of the particles, there is formation of large-size flocs generated by aggregation of weakly charged particles. These flocs prevent penetration of the individual particles inside the porosity and are easily removed from the membrane surface by the effect of cross-flow. At high pH both the membrane and the particles exhibits negative charges which lead to repulsive forces at the membrane-solution interface and depolarization of the membrane. At intermediate pH polarization and membrane resistance are maximum. One important parameter, the ionic strength of filtered solutions, was not investigated in this work. The effects of pH and ionic strength on the performance of an (x-alumina microfiltration membrane from U.S. Filter was evaluated by Nazzal and Wiesner [46]. Concerning pH effect on flux, results obtained in this work perceptibly differ from the previous one. Here the membrane operated at a significantly higher permeation rate at a pH well below the isoelectric point of the membrane. This variance can be explained considering the isoelectric point of the membrane was found at pH = 8.3 in this case while it was at pH = 3.5 in the
12 - - TRANSPORT A N D FOULING P H E N O M E N A IN LIQUID P H A S E S E P A R A T I O N
595
previous work. It can be observed that in both cases m a x i m u m of flux for pure water was measured at p H 3 to 4. Results concerning filtration studies with 0.2 ~tm titanium dioxide membranes supported on stainless steel or ceramic porous tubes were recently reported by Porter et al. [47,48]. Solutions containing sodium nitrate alone and in the presence of anionic, direct and acid dyes were filtered with adjusted solution pH. Electrolyte rejections and colour rejections were measured at p H values from 4 to 10. They showed that the charged membrane was responsible for ion rejection at low ionic concentration while rejection decreased to near 0% as the salt concentration was raised to 5000 ppm..These results are consistent with long range forces associated to Debye-length which can reach several hundred Angstroms in the solution for very low ionic concentrations. 12.4 NANOFILTRATION
WITH CERAMIC
MEMBRANES
In the early 1970s, several authors described separation membrane processes with intermediate performances between reverse osmosis (RO) and ultrafiltration (UF). Typically retention for these membranes was in the range of 50-70% for sodium chloride while it was in the 90% for organics. In the 1980s a suggested definition for these membranes was based on a molecular weight cutoff of 1000. Then "nanofiltration" was considered a suitable name for such a process which rejects molecules in the nanometer range. Presently basic properties of nanofiltration membranes can clearly be defined compared with ultrafiltration or reverse osmosis membranes: - a molecular weight cutoff of less than 1000 (membranes with MW cutoff of 1000 and above are considered UF membranes), - a lower transmembrane pressure and a higher flux than for RO, - a mixed mass transport mechanism involving convective and diffusive fluxes for both solutes and solvent, - in most cases membrane charged either positively or negatively due to their materials, - a marked influence of Donnan mechanism in the case of an aqueous feed solution containing mixed electrolytes. It results from these basic properties that nanofiltration offers unique performances for the separation of salts and organics. A negative salt rejection has been evidenced in these membranes which can be explained with reference to the above-mentioned capillary model in which the structure of nanofiltration membranes is represented by a bundle of charged capillaries with a pore radius in the nanoscale. In practice, this negative salt rejection effect can be usefully exploited in industrial desalting-concentration processes of molecules exhibiting molecular weight of less than 1000. In fact nanofiltration membranes are finding increased applicability in various fields but their transport mechanism
596
12 E TRANSPORT A N D FOULING P H E N O M E N A IN LIQUID PHASE SEPARATION
is not yet well understood. Up to now a number of published papers deal with the description of transport properties of organic nanofiltration membranes [49-51]. On the other hand, few data are available in the literature concerning ceramic nanofilters. In the following, recent results concerning separation properties of ceramic nanofilters are presented showing that some of these basic properties are relevant to describe mass transport and solute rejection observed with microporous ceramic membranes. Ceramic nanofilters are a new class of ceramic membranes which obey the basic properties of nanofiltration membranes. Some similarities can be noted between organic and inorganic NF membranes behaviour; however specificities exist with ceramic membranes due to the amphoteric properties of metal oxides in water media. Basically the structure of ceramic nanofilters can be described according to concepts developed for nanophase materials. The active layer is made of a supported microporous layer with a thickness in the micron range. This microporous structure which results from sintering of ceramic grains of less than 10 nm in size leads to membrane materials with a high surface area. Metal oxides already used for the preparation of micro- and ultrafiltration membranes can also be used for nanofilters. Microporous y-alumina, titania, zirconia and silica supported layers have been described by Julbe et al. [20] with suitable characteristics for nanofiltration. However, regarding industrial applications of these membranes for aqueous filtration, zirconia and titania are preferred to silica or y-alumina because of their stability in large pH and temperature ranges. The main characteristics of nanofiltration membranes made of oxide ceramics is that they exhibit a microporous structure with charged pore walls depending on pH and ionic strength of feed solutions. Three main cases are distinguished in the discussion of mechanisms involved in permeation and separation processes using microporous ceramic nanofilters: separation of neutral solutes in absence of electrolyte; - separation of pure electrolyte mixtures, - separation of solutions containing both organics (ionisable or not) and electrolytes; -
12.4.1 Separation of Neutral Solutes in Absence of Electrolytes When Donnan contribution can be neglected (case of neutral solutes), membrane cut-off can be determined based on respective sizes of model solutes and membrane pores. Mass transport can be described using both basic concepts of ultrafiltration and specific aspects of transport in micropores. Pure solvent flow can be described as a convective flow with a linear dependence to transmembrane pressure as shown in Fig. 12.16. With nanofiltration membranes a minimum value of pressure gradient has to be applied before to observe
12 - - TRANSPORT A N D FOULING P H E N O M E N A IN LIQUID PHASE SEPARATION
597
Jv
l
increasing pore diameter
Ap Fig. 12.16.Schematicrepresentationof hydraulicpermeabilityversus transmembranepressure for a microporous membrane. a solvent flux through the membrane. This is due to the occurrence of important capillary forces in the case of micropores of less than 2 nm in diameter. In the presence of solutes with small molecular weights, concentration polarization is likely to occur but with much less effect than in the case of ultrafiltration as explained in Section 12.2.1. A theoretical model concerning separation of sucrose and raffinose by ultrafiltration membranes has been proposed by Baker et al. [53] which assumes transport of solvent and solute exclusively through pores. This model can apply to ceramic nanofilters as they exhibit a porous structure with a pore size distribution. The retention characteristics of a given membrane for a given solute is basically determined by its pore-size distribution. The partial volume flux jv through the pores which show no rejection to the solute can be expressed as a fraction of the total volume flux Jv. (12.29)
jv= f . Jv
The solute rejection is then given as a function of the total water flux, of the solute diffusion coefficient Ds and of the pore fraction el permeable to the solute:
fexp Z R -
100 1 --
~al"/ds ) h
(12.30)
f - 1+ e x p / l i D s ) It follows from Eq. (12.30) that as Jv goes to zero, the exponential term goes to unity and the rejection coefficient reduce to zero. On the contrary as Jv tends to become very large, the exponential term tends towards infinity, and the rejection coefficient approaches a specific limiting value for a given solute. The same evolution of the rejection coefficient with volume flow and indirectly with transmembrane pressure was predicted by Tremblay [54] using the finely porous model proposed by Merten [55] and modified by Mehdizadeh
598
12 u TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
Separation factor
100
r ~,"" ~ ' ~
(%)
//
I . /
;
, /~/
80
9
60
-
.
I/.1'1"/
0
; :
e 0 0
r't "
40
Increasing Pressure Gradient
20
0
i 100
1000
I
10000
100000
Solute molecular weight Fig. 12.17. Evolution of separation factor versus solute molecular weight for different pressure gradients. Results obtained from mass transfer across micropores using radially averaged friction factors [54].
and Dickson [56] in which a radially friction factor b-1 has been included. This friction factor is directly related to the ratio ;~ of the solute radius to the pore diameter and represents the friction between a solute molecule translating along the centre line of a cylindrical pore and the pore wall.
1 P dCi 1
Ji- - ~ ~ d---~+ b Ci Jv = (1 -
~)2
b
(12.31) (12.32)
with b given by the Faxen equation b = 1 - 2.1044;~ + 2.089~ 3 - 0.948~5
(12.33)
It results from this approach that separation factors will depend on the ratio ;~ and on the operating pressure. As shown in Fig. 12.17 the influence of friction factor on separation factor is predominant at high pressure gradient.
12.4.2 Salt Rejection of Electrolyte Solutions Salt rejection of a single electrolyte by a nanofiltration membrane in the absence of Donnan contribution can be described by Eqs. (12.9) and (12.10) according to the work of Spiegler and Kedem [57]. With ceramic nanofilters the Donnan contribution has to be taken into account due to the amphoteric
12 - - TRANSPORT A N D FOULING P H E N O M E N A IN LIQUID PHASE SEPARATION
599
behaviour of metal oxide surface resulting in membrane materials with charged pore wall. The extended Nernst-Planck equation (12.19) has been applied by Tsuru et al. [58-60] to predict ion rejection in the case of charged membranes for single and mixed electrolytes. This approach showed good agreement with mass transport description obtained from irreversible thermodynamics. The general tendencies for ion rejection are as follow" for a single electrolyte solution, rejection dependency on volume flux is the same as that of neutral solutes. Increasing the charge density in the membrane make rejection higher. Rejection of divalent coion electrolyte is expected to be higher than that of monovalent coion electrolyte, while divalent counter-ion rejection seems lower than that of monovalent counter-ion electrolyte; - for a mixed electrolyte solution, rejections are shown as strongly dependent upon the volume flux, mole fraction, and the ratio of the feed concentration to the membrane charge density. Mono- and divalent coions are suggested to be separated effectively, and the monovalent coion to show negative rejection under a certain condition. However, mono- and divalent counter-ions are not so effectively separated as coions under ordinary conditions. Recently Wang et al. [61] proposed the comparison of different models from the literature to describe electrolyte transport through nanofiltration membranes. The space charge (SC) model described in Section 12.2.3 was compared with the Toerell-Meyer-Sievers (TMS) model. The SC model assumes that ion concentration and electric potential have a distribution in the radial direction in the membrane, while the TMS model supposes that both of them held uniform. The evolution of ion rejections versus Peclet number (Pe) for a 1-1 electrolyte (KC1) were compared for the two models respectively with increasing charge density at constant pore radius (5 nm) and for decreasing pore radius at a constant charge density (3.336 C.m-2). In agreement with general expressions derived from linear, non-equilibrium thermodynamic theory [7] the rejection was found to increase with Pe. This is consistent with the fact that at small Pe number there is a dominant contribution of diffusion to electrolyte transport while contribution of convection is dominant at large Pe number. With decreasing pore radius, the rejections calculated from the two models tend to coincide and shows almost the same value for pore radius of I nm. This can be explained because an overlap of double layers into the pores due to a Debye-length equal or larger than the pore size. This overlap of double layer renders the distribution of concentration and electric potential uniform in agreement with the TMS model. According to definitions of electrokinetic radius rE and dimensionless hydraulic permeability Ls given by Eqs. (12.27) and (12.28) the authors calculated evolution of Ls versus rE. In Fig. 12.18, curves Ls =firE) drawn at different potential gradients show that a maximum effect of the electrical force is expected for rE --- 1 and high potential gradient. -
600
12
Ls
-
-
TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
0.95
0.9 " 1%(;) I 9I .047 [ b 1.094 [ c 1.234 I ~ 1.46e l e 1.936 f | 2.34 g 14.68
0.85 0.8 0.75 . . . . . . . . . . 0.1
J!
9
10
1 i"E
Fig. 12.18. Dimensionless water permeability as functions of electrokinetic radius at different potential gradient (q0) [61].
In the above-considered works the behaviour of electrically charged nanofilters towards electrolyte solution has been mainly regarded with respect to Donnan analysis to explain the coion rejection. Bardot et al. [62] looked at the effect of transmembrane transport kinetics on counter ion rejection through an alumina/polysulfone composite membrane in the case of electrolyte mixtures. The rejection phenomenon is based on a "decompensation" of the convective and electric flows of a given counter-ion as a consequence of the addition of counter-ions with a different mobility. It has been shown both theoretically and experimentally (for negative charged membranes) that the same physics accounts for not only the improvement of the retention of more mobile counterions upon addition of less mobile but also for a significant deterioration of the retention (down to the negative one) of less mobile counter ions upon addition of more mobile. Experimental correlations of the phenomenon with the ratio of mobilities of counterions, the concentration of starting electrolyte and transmembrane pressure difference (Pe number) have been in complete agreement with theoretical predictions. However the influence of the ceramic support versus pH of the feed solution, which can be of great influence on ion rejection, is not discussed. The evidence of electrokinetic salt rejection by a microporous inorganic material was given by Jacazio et al. [63] based on the model of Osterle [26-28]. Experiments were carried out on the salt rejecting properties of compacted clay through which saline solutions were forced under high pressures. In accordance with the model the performance of the porous material was shown to depend on three main parameters: the ratio of the Debye length to effective pore
12 -- TRANSPORTAND FOULING PHENOMENA IN LIQUIDPHASE SEPARATION
601
radius; a dimensionless wall potential related to the ~ potential; and a Peclet number based on the filtration velocity through the pore. Comparison between the experimentally determined and theoretically predicted rejections of potassium chloride in the case of effective pore radius in the range 1-2 nm were shown to be excellent. Regarding the rejection of salt mixtures with inorganic membranes AlamiYounssi et al. [64] investigated the performance of a y-alumina membrane. In aqueous media containing indifferent electrolytes such as NaC1, the point of zero charge (zpc) of the y-alumina is near 8.5; in the presence of divalent anions or cations which are able to form surface complexes respectively with the surface groups A1OH ~ or A 1 0 , the zpc of the material can be shifted, respectively, towards higher or lower pH values, pH values for feed electrolyte solutions were measured to be in the range 5-5.9, which means that the membrane is positively charged. Results are discussed only in terms of effective charge of the membrane and valence of the co- and counter-ions present in the feed solution. In this case the membrane is positively charged and the rejection obeys to the prediction of Tsuru concerning mixed electrolyte solutions. Measured rejections are reported in Table 12.4. Rejections were shown to depend on the charge of the ions and decrease in the order: (divalent cation, monoanion) > (monocation, monoanion) or (dication, dianion) > (monocation, dianion). Another work from Rios et al. [65] also deals with performance of a positively charged y-alumina membrane fed with single NaC1, MgC12, Na2SO4 and MgSO4 solutions at various concentrations (10-4 to 10-1 mole-l-I), or even with electroTABLE 12.4 Rejection with a y-alumina membrane of mixed electrolyte water solutions [64] Sodium and calcium nitrates [Ca 2+] feed (M)
0
10-3
10-2
10-2
10-2
[Na+] feed (M)
10-2
10-2
10-2
10-3
0
Rejection N O 3-- (%)
68
Rejection Ca 2+ (%)
75
75
93
96
90
95
95
96
63
38
47
Rejection Na + (%)
68
Potassium and sodium nitrates [Ca 2+] feed (M)
0
10-3
10-2
10-2
10-2
[K +] feed (M)
10-2
10-2
10-2
10-3
0
Rejection N O 3- (%)
55
Rejection Ca 2+ (%) Rejection N a + (%)
55
18
45
50
68
32
60
56
68
15
18
25
602
12 - - TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
lyte mixtures (NaC1 + MgC12; Na2SO 4 + MgSO4) of constant counter ion concentration ( 1 0 -3 mole.l-I). Results that confirm the previous trends are explained using a new simplified model based on Eq. (12.19) that makes the assumption of Donnan effect at pore entrance. This model accounts for electrokinetic phenomena inside the pores and also considers differences in ion mobility. Zirconia nanofilters (partially stabilized or not) have been investigated by Guizard et al. [66,67] with respect to rejection performance towards model solutes. These membranes were synthesized by the sol-gel process using zirconium and magnesium alkoxide precursors, the later being used as stabilizer agent for the cubic zirconia phase. Pore diameter for these membranes is in the range I to 2 nm depending on preparation conditions. In agreement with data published in the literature a zpc near 7 has been evidenced. In this work special attention has been paid to the influence at one hand of pore size and specific surface area of the membranes, on the other hand of transmembrane pressure and ionic force of feed solutions. It has been shown that these parameters clearly relate to the dimensionless zeta-potential %, the electrokinetic radius rE and the dimensionless hydraulic permeability Ls resulting from the variation rE and %. The rejection versus pH of chloride and sulfate ions using a Na2SO4/NaC1 mixture (200 ppm) is shown in Fig. 12.19. At a pH < zpc chloride and sulfate must be regarded as counterions for the membrane while at pH > zpc they behave as coions. One can see that results are in good agreement with the prediction; sulfate are better rejected than chloride when they are coions of the
Rejection (%)
I00
',
,
'
!
'
'
'
I
,
,
-- Su,fate j
--~
80
60
Chl~
,
!
,
,
,
!
i
i
i
i
8
10
,
,
,
I
'--'~. iP."C~............i
40
................
. . . . . . . . . .
-
20
2
4
6
12
pH Fig. 12.19. Rejection of an electrolyte mixture Na2SO4/NaC1 (200 p p m ) b y a zirconia nanofilter. Effect of p H [66,67].
603
12 - - T R A N S P O R T A N D F O U L I N G P H E N O M E N A IN LIQUID P H A S E S E P A R A T I O N
Permeability (l/h.mZ.bar) 30
l..''
E zo
'
'
"
I
,
. . . .
i
I
'
"
"L.i
I
"~
'"
'"'
'
"
-,!
:!
.................... [............................ q """ - qm :
F - t.... ...................... - .............................. . ~ - ' , i i I . . . . I. . . . . . . . . .
'
. . :, . & . . . . . . . . . . . . . . . . . . . . . . . . . . ~ ............................... ! . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
-10
9
, i
iiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiii
0
, 0
$
10
15
20
P r e s s u r e (bar) Fig. 12.20.Permeability versus transmembrane pressure for two zirconia nanofilters with different microporous volumes: (a) 8x10 -2 cm3/g, (b) 6.6-10-2 cm3/g [66,67].
membrane. On the contrary both sulfates and chlorides are not rejected w h e n they are counterions of the membrane. Charge density at the pore wall is a key parameter for the description of electrolyte rejection by charged nanofiltration membranes. In ceramic nanofilters charge density can be related to the specific surface area measured on ceramic m e m b r a n e material. In the case of a negative charged m e m b r a n e material (pH > zpc), Figs. 12.20 and 12.21 show respectively the influence of transmembrane pressure on m e m b r a n e permeability and the sulfate rejection versus flux for two membranes exhibiting different specific surface area. Calculated hydraulic radius was almost the same for the two membranes (rh ----0.43 nm) so that permeability can be discussed in term of the Donnan effect and related streaming potential for electrically charged porous membranes assuming that m e m b r a n e thickness is the same in both cases. The effect of electrokinetic flow on m e m b r a n e permeability is shown in Fig. 12.20. When transmembrane pressure increases permeability increases at first and then decreases due to the opposite contribution of electrokinematic flow to convective flow. This can be explained by the occurrence of a non-negligible streaming potential for a transmembrane pressure higher than 3 bar. Moreover the electrokinematic flow effect was more marked for the m e m b r a n e with the higher surface area which is consistent with a higher charge density. If we consider now sulfate rejection by the two membranes, Fig. 12.21, a better rejection was obtained with the m e m b r a n e exhibiting the higher surface area and consistently the higher
604
12 - - TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
Rejection (%) 100
.....o ........ .....................i.....................i............... 6O
40
2O
0
0
$0
100
150
200
250
300
Permeate flux (l/h.m z) Fig. 12.21. Sulfate rejection versus volume flux through two zirconia nanofilters with different microporous volumes: (a) 8-10-2 cm3/g, (b) 6.6.10-2cm3/g [66,67]. charge density. However the evolution of sulfate rejection with flux is not totally explained by Tsuru calculation [59] on sulfate rejection through negatively charged polymer membranes. With these zirconia nanofilters a m a x i m u m in rejection versus flux was evidenced while an increase of rejection followed by a plateau has been described with polymer nanofiltration membranes.
12.4.3 Separation of Aqueous Ionized Molecule-Salt Solutions The case which consists of a mixture of a mono-monovalent salt like NaC1 and a multifunctional organic anion A n- containing n negatively charged groups per molecule in a sodium salt form has been described by Perry and Linder [50]. It has been assumed that the monovalent anion C1- is permeable through the m e m b r a n e and the organic anion A ~- is fully rejected. Accordingly a new expression for salt rejection was proposed: TR = 1 - [(1 - r~)~/(1 - r~F)]
(12.34)
F is defined as in Eq. (12.12) and [3 as [3- (1
+
n CAb/Csb) 1/2
(12.35)
with CAB and Csb respectively the concentration of the organic anion and the concentration of monovalent anion in the feed solution. When only pure salt is p r e s e n t C A B - - 0, [~ = 1, Eq. (12.34) becomes identical to Eq. (12.11).
12 -- TRANSPORTAND FOULINGPHENOMENA IN LIQUIDPHASE SEPARATION
605
Schirg and Widmer [52] published mathematical models for the calculation of retention and selectivity for nanofiltration of aqueous dye-salt solutions. A modification of Eqs. (12.11) and (12.12) has been proposed in which the integral salt permeability r could be described by the introduction of an exponential function c0 = (x C~
(12.36)
with (x a constant and 7 a coefficient for salt permeability dependence on concentration. Both calculations by Perry and Schirg have been performed to describe and to predict the rejection characteristics of organic nanofiltration membranes when ionic and chargedmolecular solute mixtures are used in the feed solution. Recently experiments were carried out with ceramic nanofilters [67] which showed that similar properties can be obtained. As an example, results concerning the rejection of a dye/electrolyte mixture at pH = 9 through a zirconia nanofilter are reported in Table 12.5. -As a general conclusion to this part dedicated to nanofiltration with ceramic membranes one can assume that the general behaviour of these membranes can be assimilated to the behaviour of electrically charged organic nanofiltration membranes. However some specificities exist with ceramic nanofilters due to a sintered metal oxide grains derived porous structure and an amphoteric character TABLE 12.5 Rejection of a mixture of an organic anion (bromocresol green) and salt anions (SO4, C1-) through a negatively charged zirconia nanofilter [67] Anion
Concentration (ppm)
Mw
Pressure ( A p ) (bar)
Rejection (%)
200
698
10 20
63 71
NaC1/Na2SO 4 mixture C1-
2000
58
SO~
2000
142
10 20 10 20
3 6 39 40
10 10 10
70 0 48
Bromocresol green A-
Bromocresol green/NaC1/Na2SO4 mixture A200 698 C12000 58 SO4
2000
142
606
12 - - T R A N S P O R T A N D F O U L I N G P H E N O M E N A IN LIQUID P H A S E S E P A R A T I O N
in water media. At this time few experimental data are available in view of an assessment of existing or new mathematical models well adapted to ceramic nanofilters. Further experiments with different categories of ceramic membrane material are needed to establish general principles of nanofiltration with ceramic membranes.
12.5 PROSPECTIVE ASPECTS
12.5.1 Organic-Inorganic Hybrid Membranes and Related Processes At the present time, organic-inorganic hybrid membranes do not exist at the commercial stage. However, recent results have shown the interest of these membranes in a non-limited list of applications such as gas separation, pervaporation, chemical and biological sensors, facilitated transport, ultra- and nanofiltration. The main interest of organic-inorganic membranes is that they can combine basic properties of both organic and inorganic membrane materials. Accordingly improved properties are expected from this new category of membrane. A short overview of recent works dedicated to these membranes is given hereafter which illustrates their potentiality in liquid phase separation. A first way to obtain an organic-inorganic hybrid membrane is to have a polymer material either deposited or grafted at the surface or embedded in the top-layer porosity of a ceramic support. For example Castro et al. [68] investigated the permeability behaviour of polyvinylpyrrolidone-modified porous silica membranes. The surface of 0.4 ~tm-pore-size silica membranes was modified with a covalently bonded polyvinylpyrrolidone brush layer. Hydraulic permeability measurements performed with six different solvents and both unmodified and modified membranes suggest that the permeability of the modified membrane is determined by the configuration of the terminally anchored polymer chains. In the modified ceramic-supported polymer membrane, the swelling of the polymer brush layer increased as the solvent power increased, resulting in a decrease in the pore radius and subsequently the permeability. In a previously mentioned study Bardot et al. [62] used nanofiltration membranes made by internal coating of porous tubular supports of R-alumina with sulfonated collodion followed by coagulation in an appropriate bath. More recently Sarrade et al. [69] have developed a hybrid nanofiltration membrane highly effective for separating non-charged solutes of molecular weight as low as 500-1000 Dalton in supercritical carbon dioxide medium. This is a combined organic-inorganic membrane that comprises a macroporous ~-alumina substrate (tubular or multichannel), an intermediate mesoporous inorganic titanium oxide layer (thickness: 1 ~tm) and a microporous Nation polymer top-layer (thickness: less than 0.1 ~tm). The overall performance and
12 - - TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
607
TABLE 12.6 Transport parameter value of a Nation/titania hybrid membrane [69,70]
(~ co(m.s-1)
EG
PEG 200
PEG 400
PEG 600
PEG 1500
6.010-2 4.410-5
4.410-1 2.1 10-6
5.710-] 8.810-7
8.1 10-1 1.710-7
9.610-1 3.1 10-8
t r a n s p o r t m e c h a n i s m t h r o u g h this m e m b r a n e have been studied using ethylene glycol (EG) and various polyethylene glycols (PEG) as m o d e l solutes [70]. Starting from Eqs. (12.9) to (12.12), the m e m b r a n e permeabilities to water, Lp, and solutes, co, as well as the reflection coefficients, r~, were d e t e r m i n e d at first. These values are reported in Table 12.6 and in Fig. 12.22. Using the theory p r o p o s e d by Verniory et al. [6] to account for h i n d e r e d transport in pores, the m e a n pore radius was estimated from these parameters. It is w o r t h noting that the m e a n value of 0.6 n m calculated for the m e m b r a n e is consistent w i t h the pore d i m e n s i o n (0.8 nm) directly m e a s u r e d using the biliquid p e r m p o r o m e t r y [71]. It has been s h o w n that, regardless of the size of the solute molecule, convective transport is always more i m p o r t a n t than diffusive transport. In 1,0
0,8 x
Membrane TN
0,6
0,4 m Membrane A 0,2
0,0
0,0
0,2
0,4
rs
0,6
0,8
1,0
(nm)
Fig. 12.22. Variation of reflection coefficient (~versus equivalent radius of model solutes rs for an alumina nanofilter (A) [65] and a Nation/titania hybrid membrane (TN) [70].
608
12 - - TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
accordance with nanofiltration behaviour this membrane also exhibited effective separation of ionic species. Organic-inorganic polymer at the molecular level are also of interest as shown in the following examples. A new concept in nanofiltration has been proposed by Guizard et al. [72] based on a hybrid polymer (cyclic polyphosphazene) supported on a zirconia ultrafiltration membrane. Excellent chemical and temperature resistances were obtained for these membranes due to intrinsic properties of polyphosphazenes as well as a high rejection of small organics and a good selectivity concerning multivalent versus monovalent ions. The reflection coefficient r~ was markedly related to the transmembrane pressure leading to adjustable working conditions. Another example is in an alternative way to selective transport of metal ions through liquid membranes, such as transport mediated by crown-ether and other macrocyclic ligands which has been extensively investigated during the last twenty years. No practical separation processes arose during this period mainly because liquid membranes suffer poor stability and thus short lifetimes: the membrane degradation is essentially due to the loss of carrier by dissolution in the aqueous phase and by emulsion formation at the membrane interfaces. Consequently the recent developments in facilitated transport membrane processes are focused on new membrane systems with improved lifetimes. One of these systems is based on the carrier grafting onto a solid membrane matrix. Grafting of benzo-15-crown-5 in a heteropolysiloxane membrane was investigated by Lacan et al. [73] in view of facilitated transport of alkaline ions. Very stable membranes over several months were obtained without loss of carrier during transport experiments. It has been demonstrated that covalently bound carriers allow facilitated transport of K § ions versus Li + ions to take place with high diffusion rates, high facilitation factor and good selectivity. These membranes open a new way in the application of facilitated transport to practical separation processes.
12.5.2 Coupled Membrane Processes Inorganic membranes, and to a less extent hybrid membranes, possess a high degree of resistance to chemical and abrasion degradation as well as tolerate a wide range of pH and temperature values. All these properties make them very useful for coupling with other processes and open up new fields of applications. In what follows, some examples of such integrated processes involving at least one membrane separation stage are presented.
Membrane bioreactor The idea of coupling membrane separation with bioreaction is not a new one. A lot of works published in the literature bear witness to this fact. But most of
609
12 - - TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
the time organic membranes and aqueous phase reactions are considered [74]. For areas of biotechnology or food engineering, a main advantage of inorganic materials is that they may be repeatedly autoclavable and are very stable against microbiological attack [32]. With them, reactions in pure organic solvent may be also successfully faced. As an example, the enantiomeric resolution of menthol (+) into methyl (-) laurate, through a biological catalysis method involving a lipase from Candida rugosa and lauric acid as substrate, was recently investigated using n-heptane as the solvent medium [75]. A zirconia membrane with a pore diameter of about 4 nm was chosen to'retain the biocatalyst. This lets the substrate and product molecules pass. The transmembrane pressure was selected so as to get a space time leading to an optimum reaction yield. At the reactor outside, menthol (+) and methyl (-) laurate were separated from permeate, and then menthol (-) was regenerated from ester.
Nanofiltration plus supercritical fluid extraction Supercritical fluid extraction is used to recover small organic solutes with molecular weight below 1500 daltons. In a state of continuity between vapour and liquid, supercritical fluids exhibit intermediate transport properties with lower viscosities than liquid and higher diffusivities than gases. Because of its
(~
A I CO 2
I I I I
I I
I
P > 74 bars T > 31 "C
Valve
P < 60 bars
exc6ange r I ,l
' ~ ' " ~ - " " " ~
I I
!
_II
I
I
I
T Extract
.4 -
/
Mixt
extracts
Fig. 12.23. Nanofiltration/supercritical fluid extraction coupled processes.
"
610
12 m TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
attractive critical temperature/pressure conditions, stability, low cost and nontoxicity, CO2 is today the most widely used SCF. As previously indicated, nanofiltration now provides new ceramic and hybrid membranes with cut-off in the range of 300 to 1000 daltons. On principle, nanofiltration plus supercritical extraction aims to both enhance the selectivity of extraction and lower energy consumption. A schematic view of the process is given in Fig. 12.23. Regarding selectivity, it may be thought that membrane sieving effect will induce a separation of supercritical fluid mixture into fractions respectively containing high (retentate) and low (permeate) molecular weight solutes. From an economic viewpoint, a substantial energy saving may be expected due to the fact that only permeate flow that just represents a small part of total CO2 will be submitted to a strong pressure reduction from extract recovery, while low soluble heavy compounds will be continuously deposited from retentate by means of a small pressure and/or temperature effect. Experiments have proved that silica [76], or titane-nafion membranes [77] were able to endure supercritical fluid conditions with no alteration. With y-alumina membrane, fouling strongly develops probably due to chemisorption. Working with model molecules such as ethylene glycol and polyethylene glycols (PEG 200-400-600), the process capability to extract and separate various size solutes has also been checked [78].
Ultrafiltration plus electrophoresis It is worth recalling that the flux and the selectivity of ultrafiltration may be improved when treating electrically charged solutes ~ as an example, alkaline gelatin molecules (pI = 4.7) processed at pH = 6.0 present a negative charge by superposing upon the driven pressure an electric field which acts on the retained solute to control concentration polarization. This is the so-called "electro-ultrafiltration process". In the past various works have underlined the influence on performance of such parameters as pressure, fluid velocity, electric field strength or starting conditions particularly with ceramic membranes [79]. With membranes cylindrically shaped, and for instance when processing a negatively charged solutes, a classical setting diagram consists both in installing a stainless steel wire as anode through the centre of the membrane and in closely surrounding the outside of the supporting tube by a cathode made of a large mesh stainless wire lattice (Fig. 12.24). Because the supporting tube is placed along with the membrane itself in the electric field acting area, disadvantages may result from the use of this traditional set-up: excessive energy consumption, parasite and uncontrolled effects (such as electro-osmosis fluxes). A new concept has been recently proposed with inorganic membranes to overcome some of these difficulties. It consists in designing electronic conductive membranes in which the original feature is the possibility of using the
12 -- TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
Membrane support
/
611
Central electrode
,/
~
ctive membrane External electrode for non-conductive membrane
Retentate
t' b Electric,potential
I
Fig. 12.24. Schematic view of electro-ultrafiltration using conductive inorganic membranes.
active layer both as a filter and as an electrode. RuO2-TiO2 membranes coated on an alumina support belong to this class. With them, performance may be enhanced by applying the electric field only inside the filtration module, specially when the membrane is used as the anode, a classical way to work RuO2-TiO 2 electrodes [80,81].
Cross-flow filtration with mobile turbulence promoters It is well known that pumping of the fluid has a major effect on flux in the mass transfer controlled region for UF/MF process. Indeed agitation and mixing of the fluid near the membrane surface sweep away the accumulated solutes, thus reducing the thickness of boundary layers. This is the simplest and most effective method of controlling the effect of concentration polarization.
12 - - TRANSPORT AND FOULING PHENOMENA IN LIQUID PHASE SEPARATION
612 J (I/h.m2)
Fluidized bed (Stainless steel beads 3mm in diameter and 7.9 in density)
50-
40
30 20
E m p t y robe a
i --
10
s = 0,42
i
,_..._._....p..--I
/k
2O
E = 0,6
,
I
~
30
~ = 0,7
~
,'k
I 4O
,
I 50
u(~)
s = 0,8
Minimum fluidization Fig. 12.25. Permeate flux versus tangential fluid velocity with a gelatine solution (10 g 1-1, Ap = 1.5 bar) ushlg a tubular a l u m i n a m e m b r a n e filled with a fluidized bed.
The magnitude of the effect of flow rate on the mass transfer coefficient will depend on whether the flow is turbulent or laminar, as well as on rheological properties of the fluid, the key factor being the shear stress at wall. Another less c o m m o n method to effect permeate flux increases is through the introduction of turbulence promoters in the flow conduit. Up to now more attention has been given to fixed promoters due to damage that ordinarily results from the movements of free agents at the very fragile surface of traditional organic membranes [82]. On the contrary, ceramic membranes (alumina as an example) are resistant enough to endure the continuous bombardment of fltfidized particles [83] or even the friction of transported solids [84]. With such devices, high permeate fluxes may be obtained (Fig. 12.25) with no sharp decrease in solute rejection (Fig. 12.26), even at tangential fluid velocity as low as a few ten centimetres per second. The analysis of mass transfer coefficients and hydraulic resistances showed that moving particles insure a significant reduction of the mass transfer boundary layer, as well as a continuous mechanical erosion of the deposit at wall. Polarization is strongly modified. From a practical viewpoint, low retentate velocities may offer some interesting developments in those cases where fragile molecules are to be treated, or long enough residence times are needed. Solid particles could also be used as catalyst (enzymatic supports as an example) for heterogeneous reactions, adsorbent for coupled MF/adsorption processes. As shown in Figs. 12.25 and 12.26 even the existence of optimum working conditions for fluidized bed devices at an intermediate bed porosity could be turned to advantage to elaborate new permeate flux control strategies.
12 -- TRANSPORT AND FOULING PHENOMENA IN LIQUIDPHASE SEPARATION
613
Ti
100
Fl/dJze, d bed
80
70
Empty tube
0
,
I
10
,
t
20
,,,
t
30
,
t
40
,
,
,
J U(cm/s)
50
Fig. 12.26. Gelatin rejection versus tangential fluid velocity usin~ a tubular alumina m e m b r a n e filled with a fluidized bed (feed solution 10 g 1-, Ap = 1.5 bar).
12.6 CONCLUSION Different aspects of liquid phase separation using inorganic membranes should be emphasized compared with organic membrane behaviour. The first characteristic of inorganic membranes designed for liquid filtration is that they exhibit a non-deformable porous structure with pore size adapted to three main processes: macropores for microfiltration, mesopores for ultrafiltration and micropores for nanofiltration. Modelling of mass transfer across these membranes is related to basic phenomena involved in liquid flow through porous media. The Darcy law applies to convective volume flux: through macro- and mesoporous membranes while a convection-diffusion mechanism better explains solvent flux in the case of microporous membranes. Due to pore shape resulting from packing and sintering of mineral particles the Carman-Kozeny model which includes specific surface area and tortuosity provides a better description of the permeability coefficient than the Hagen-Poiseuille law. The second characteristic of inorganic membranes used in liquid phase separation is that most are made of ceramic oxides. If solute rejection basically originates in size effects related to pore dimension, specific properties are attached to ceramic membrane material. The amphoteric behaviour of metal oxide surfaces is certainly the most important one as membranes can exhibit negative or positive charge density depending on the pH of feeding solutions. Two parameters, zeta-potential and streaming potential, greatly influence rejection and permeability of electrolyte solutions all the more as membranes exhibit small pore size and large specific surface area.
614
12 m TRANSPORT A N D FOULING P H E N O M E N A IN LIQUID PHASE SEPARATION
Fouling, responsible for flux decline, is also an i m p o r t a n t p a r a m e t e r to deal w i t h in the description of transport m e c h a n i s m s w i t h inorganic m e m b r a n e s . Three m a i n causes have been identified as i m p o r t a n t contributions to fouling of inorganic m e m b r a n e s . It has been suggested that the formation of ceramic m e m b r a n e s can be a first cause of flux decline as far as association of adjacent g r a n u l a r layers results in highly resistant b o u n d a r y layers. A second p h e n o m e n o n responsible for flux decline is the on-line m e m b r a n e fouling w h i c h is a function of the h y d r o d y n a m i c conditions and is i n d e p e n d e n t of the physical properties of the m e m b r a n e . Finally interaction b e t w e e n m e m b r a n e material a n d molecules or macromolecules can result in the formation of a d y n a m i c layer on the original filtering element. This layer can be r e g a r d e d as a formed-inplace m e m b r a n e w i t h specific separation properties and it is responsible for an additional resistance to the v o l u m e flux.
REFERENCES
1. R. Bhave, Inorganic Membranes Synthesis, Characteristics and Applications. Van Nostrand Reinhold, New York, 1991. 2. M. Mulder, Basic Principles of Membrane Technology. Kluwer, Dordrecht, Boston, London, 1991. 3. S. Nakao, T. Anazawa, T. Tsuru and S. Kimura, Influence of high permeation on concentration polarization in ultrafiltration. Presented at ISMMP'94, 5-10 April 94, Hangzhou, P.R. China. 4. J.A. Wesselingh and R. Krishna, Mass Transfer. Ellis Horwood, New York and London, 1990. 5. O. Kedem and A. Katchalsky, Thermodynamic analysis of the permeability of biological membranes to non-electrolytes. Biochim. Biophys. Acta, 27 (1958) 229. 6. A. Vemiory, R. Dubois, P. Decoodi and J.P. Gassee, Measurement of the permeability of biological membranes. J. Gen. Physiol., 62 (1973) 489. 7. P.M. Bungay, Transport principles m Porous membranes, in: P.M. Bungay, H.K. Lonsdale, M.N. de Pinho (Eds), Synthetic Membranes: Science, Engineering and Applications. NATO ASI Series, Series C: Mathematical and Physical Sciences, Vol. 181, 1986. p. 109. 8. T. Tsuru, M. Urairi, S. Nakao and S. Kimura, Negative rejection of anions in the loose reverse osmosis separation of mono- and divalent ion mixtures. Desalination, 81 (1991) 219. 9. W.M.Deen, Hindered transport of large molecules in liquid -filled pores. AIChE J., 33 (9) (1987) 1409 10. C. Guizard, A. Julbe, A. Larbot and L. Cot, Ceramic membrane processing, in: B.I. Lee and E.A.J. Pope (Eds.), Chemical Processing of Ceramics. Marcel Dekker, New York, 1994, p. 501. 11. J. Randon, A. Julbe, P. David, K. Jaafari and S. Elmaleh, Computer simulation of inorganic membrane morphology: 2. Effect of infiltration at the membrane support interface. J. Colloid Interface Sci., 161 (1993) 384.
12-- TRANSPORTAND FOULINGPHENOMENAIN LIQUIDPHASESEPARATION
615
12. P. Aimar, C. Taddei, J.P. Lafaille and V. Sanchez, Mass transfer limitations during ultrafiltration of cheese whey with inorganic membranes. J. Membr. Sci., 38 (1988) 203 13. A.D. Marshall, P.A. Munro and G. Tr~igardh, The effect of protein fouling in microfiltration and ultrafiltration on permeate flux, protein retention and selectivity: A literature review. Desalination, 91 (1993) 65. 14. C. Vetier, M. Bennasar and B. Tarodo de la Fuente, Study of the fouling of a mineral microfiltration membrane using scanning electron microscopy and physicochemical analysis in the processing of milk. ]. Dairy Res., 55 (1988) 381. 15. P.Freund and G.M. Rios, M6canismes de colmatage et performances compar6es de quatre membranes d'alumine: cas de l'ultrafiltration de g61atine. Can. J. Chem. Eng., 70 (2) (1992) 250. 16. Y. Matsumoto, S. Nakao and S. Kimura, Cross-flow filtration of solutions of polymers using ceramic microfiltration. Int. Chem. Eng., 28 (1988) 677. 17. G. Grund, C.W. Robinson and B.R. Glick, Protein type effects on steady-state crossflow membrane ultrafiltration fluxes and protein transmission. J. Membr. Sci., 70 (1992) 177. 18. J.L. Negrel, G.M. Rios and J.L. Cuq, Design and performance of new formed-in-place proteinic membranes for nanofiltration. Presented at ISMMP'94, 5-10 April 1994, Hang-
zhou, P.R. China. 19. C. Guizard, A. Larbot and L. Cot, A new generation of membranes based on organicinorganic polymers. Proceedings of the First International Conference on Inorganic Membranes, Montpellier, France, 1989, p. 55. 20. A. Julbe, C. Guizard, A. Larbot, L. Cot and A. Giroir-Fendler, The sol-gel approach to prepare candidate microporous inorganic membranes for membrane reactors. J. Membr. Sci., 77 (1994) 137. 21. R. Mellis, W.N. Gill and G. Belford, Fluid dynamics in a tubular membrane: theory and experiment. Chem. Eng. Commun., 122 (1993) 103. 22. S.M. Ahmed, Studies of he double layer at oxide solution interface. ]. Phys. Chem., 73 (1969) 3546. 23. J. Lyklema, The structure of the electrical double layer on porous surfaces. J. Electroanal. Chem., 18 (1968) 341. 24. J.W. Perram, R.J. Hunter and H.J.L. Wright, The oxide-solution interface. Aust. J. Chem., 27 (1974) 461. 25. S. Levine, J.R. Marrott, G. Neale and N. Epstein, Theory of electrokinetic flow in fine cylindrycal capillaries at high zeta-potentials. J. Colloid Interface Sci., 52 (1975) 136. 26. F.A. Morrison, JR and J.F. Osterle, Electrokinetic energy conversion in ultrafine capillaries. J. Chem. Phys., 43 (6) (1965) 2111. 27. R.J. Gross and J.F. Osterle, Membrane transport characteritics of ultrafine capillaries. J. Chem. Phys., 49 (1) (1968) 228. 28. J.C. Fair and J.F. Osterle, Reverse electrodialysis in charged capillary membranes. J. Chem. Phys., 54 (8) (1971) 3307. 29. H.K. Lonsdale and W. Pusch, Donnan-membrane effects in hyperfiltration of ternary systems. J. Chem. Soc., Faraday Trans. 1, 71 (3) (1975) 501. 30. A.W. Adamson, Physical Chemistry of Surfaces. Wiley, New York, 1988, 272 pp. 31. J. Israelachvili, Intermolecular and Surface Forces. Academic Press, London, 1985, 180 pp. 32. S. Saksena and A.L. Zydney, Pore size distribution effects on electrokinetic phenomena in semipermeable membranes. ]. Membr. Sci., 105 (1995) 203. 33. G.M. Rios, B. Tarodo de la Fuente, M. Bennasar and C. Guizard, Cross-flow filtration of biological fluids on inorganic membranes: a first state of the art, in: S.Thorne (Ed.),
616
34. 35. 36. 37.
38. 39.
40. 41.
42.
43. 44. 45.
46. 47. 48.
12 n TRANSPORTAND FOULINGPHENOMENAIN LIQUIDPHASESEPARATION
Developments in Food Preservation 5. Elsevier, London, New York, 1989, p. 131. G. Belfort, R.H. Davis and A.L. Zydney, The behavior of suspensions and macromolecular solutions in crossflow microfiltration. J. Membr. Sci., 96 (1994) 1. H.B. Winzeler and G. Belfort, Enhanced performance for pressure-driven membrane processes: the argument for fluid instabilities. J. Membr. Sci., 80 (1993) 35. B.B. Gupta, P. Blanpain and M.Y. Jaffrin, Permeate flux enhancement by pressure and flow pulsations in microfiltration with mineral membranes. J. Membr. Sci., 70 (1992) 257. M.Y. Jaffrin, L.H. Ding and J.M. Laurent, Kinetics of concentration polarization formation in crossflow filtration of plasma from blood: experimental results. J. Membr. Sci., 72 (1992) 267. Y. Wang, J.A. Howell, R.W. Field and D. Wu, Simulation of crossflow filtration for baffled tubular channels and pulsatile flow. J. Membr. Sci., 95 (1994) 243. J. Charpin, P. Bergez, F. Valin, H. Bamier, A. Maurel and J.M. Martinet, Inorganic membranes: preparation, characterization, specific applications. Mater. Sci. Monogr., 38C (1987) 2211. J. Murkes and C.-G. Carlsson, Crossflow Filtration, Theory and Practice. Wiley, New York, NY, 1988. M.-C. Aubert, M.-P. Elluard and H. Barnier, Shear stress induced erosion of filrtration cake studied by a flat rotating disk method. Determination of the critical shear stress of erosion. J. Membr. Sci., 84 (1993) 229. E. Vernon Ballou, T. Wydeven and M.I. Leban, Solute rejection by porous glass membranes. I. Hyperfiltration of sodium chloride and urea feed solutions. Environ. Sci. Technol., 5 (1971) 1032. E. Vernon Ballou and T. Wydeven, Solute rejection by porous glass membranes. II. Pore size distributions and membrane permeabilities. J. Colloid Interface Sci., 41 (1972) 198. K. Singh and A. Singh, Membrane potential and solute separation studies on zirconium phosphate membranes. ]. Membr. Sci., 82 (1993) 141. M.R. Hoogland, A.G. Fane and C.J.D. Fell, The effect of pH on the cross-flow filtration of mineral slurries using ceramic membranes. Proceedings of the First International Conference on Inorganic Membranes, Montpellier, France, 1989, p. 153. F.F. Nazzal and M. Wiesner, pH and ionic strength on the performance of ceramic membranes in water filtration. J. Membr. Sci., 93 (1994) 91. J.J. Porter and R.S. Porter, Filtration studies of selected anionic dyes using asymmetric titanium dioxide membranes on porous stainless tubes. J. Membr. Sci., 101 (1995) 67. J.J. Porter and S. Zhuang, Microfiltration of sodium nitrate and direct red 2 dye using asymmetric titanium dioxide membranes on porous ceramic tubes, to be published in J. Membr. Sci.
P. Eriksson, Nanofiltration extends the range of membrane filtration. Environ. Progr., 7 (1) (1988) 58. 50. M. Perry and C. Linder, Intermediate reverse osmosis ultrafiltration membranes for concentration and desalting of low molecular weight organic solutes. Desalination, 71 (1989) 233. 51. R. Rautenbach and A. Gr6schl, Separation potential of nanofiltration membranes. Desalination, 77 (1990) 73. 52. P. Schirg and F. Widmer, Characterisation of nanofiltration membranes for the separation of aqueous dye-salt solutions. Desalination, 89 (1992) 89. 53. R.W. Baker, F.R. Eirich and H. Strathmann, Low pressure ultrafiltration of sucrose and raffinose. I. Phys. Chem., 76 (2) (1972) 238. 49.
12 u TRANSPORTAND FOULINGPHENOMENAIN LIQUIDPHASESEPARATION
617
54. A.Y. Tremblay, Finely porous models and radially averaged friction factor, J. Appl. Polym. Sci., 45 (1) (1992) 159. 55. U. Merten, in: U. Marten (Ed.), Desalination by Reverse Osmosis. M.I.T. Press, Cambridge, MA, 1966, p. 15. 56. H. Medizadeh and J.M. Dickson, ]. Appl. Polym. Sci., 42 (1991) 1143. 57. K.S. Spiegler and O. Kedem, Thermodynamics of hyperfiltration (reverse osmosis): criteria for efficient membranes. Desalination, 1 (1966) 311. 58. T. Tsuru, S.-I. Nakao and S. Kimura, Effective charge density and pore structure of charged ultrafiltration membranes. J. Chem. Eng. Jpn., 23 (5) (1990) 64. 59. T. Tsuru, S.-I. Nakao and S. Kimura, Calculation of ion rejection by extended NernstPlanck equation with charged reverse osmosis membranes for single and mixed electrolyte solutions. J. Chem. Eng. Jpn., 24 (4) (1991) 511. 60. T. Tsuru, M. Urairi, S.-I. Nakao and S. Kimura, Reverse osmosis of single and mixed electrolytes with charged membranes: experiments and analysis. J. Chem. Eng. Jpn., 24 (4) (1991) 518. 61. X.L.Wang, T. Tsuru, S.-I. Nakao and S. Kimura, Electrolyte transport through nanofiltration membranes by the Space Charge Model and the comparison with TeorellMeyer-Sievers model. J. Membr. Sci., 103 (1995)117. 62. C. Bardot, E. Gaubert and A.E. Yaroshchuck, Unusual mutual influence of electrolytes during pressure driven transport of their mixtures across charged porous membranes. J. Membr. Sci., 103 (1995) 11. 63. G. Jacazio, R.F. Probstein, A.A. Sonin and D. Yung, Electrokinetic salt rejection in hyperfiltration through porous materials. Theory and experiment. J. Phys. Chem., 76 (26) (1972) 4015. 64. S. Alami-Younssi, A. Larbot, M. Persin, J. Sarrazin and L. Cot, Rejection of mineral salts on a gamma alumina nanofiltration membrane. Application to environmental process. J. Membr. Sci., 102 (1995)123. 65. G.M. Rios, R. Jouli6, S. Sarrade and M. Carles, Investigation of ion separation by microporous nanofiltration membranes, AIChE J., to be published. 66. C. Guizard, C. Mouchet, R. Vacassy, A. Julbe and A. Ayral, Zirconia nanofiltration membranes: I. Mechanism of pore formation and static characterization., in preparation. 67. C. Guizard, C. Mouchet, R. Vacassy, X. Bouisson and V. Thoraval, Zirconia nanofiltration membranes: II. Performance of the membranes, dynamic characterization with model solutes, in preparation. 68. R.P. Castro, Y. Cohen and H.G. Monbouquette, The permeability behavior of polyvinylpyrrolidone-modified porous silica membranes. J. Membr. Sci., 84 (1993) 151. 69. S. Sarrade, C. Bardot, M. Carles, R. Soria, S. Cominotti and R. Gillot, Elaboration of new multilayer membrane for nanofiltration. Proceedings of the 6th World Filtration Congress,
18-21 May 1993, Nagoya, Japan. 70. S.Sarrade, G.M. Rios and M.Carles, Dynamic characterisation and transport mechanisms of two inorganic membranes for nanofiltration. J. Membr. Sci., 114 (1996) 81. 71. M.G. Liu, R.Ben Aim and M. Mietton-Peuchot, Characterization of inorganic membranes by permporometry method: importance of non-equilibrium phenomena. Key Eng. Mater., 61/62 (1991) 603. 72. C. Guizard, A. Boy6, A. Larbot and L. Cot, A new concept in nanofiltration based on a composite organic-inorganic membrane. Rec. Progr. Gdn. Procddds, 6 (22) (1992) 27. 73. P. Lacan, C. Guizard, P. Le Gall, D. Wettling and L. Cot, Facilitated transport of ions
618
12-- TRANSPORTAND FOULINGPHENOMENAIN LIQUIDPHASESEPARATION through fixed-site carrier membranes derived from hybrid organic-inorganic materials.
J. Membr. Sci., 100 (1995) 99. 74.
M.Cheryan and M.A. Mehaia, Membrane bioreactors, in: W. Courtney McGregor (Ed.),
Membrane Separations in Biotechnology. Marcel Dekker Inc., New York, Basel, Vol. 10, 1986, p. 255. G.M. Rios, F. Lambert and J.C. Jallageas, Essais de mise en oeuvre d'un r6acteur d'ultrafiltration pour la catalyse biologique en milieu solvant: cas de l'est6rification enantios61ective par lipase du menthol. Entropie, 63 (1991) 31. 76. K. Nakamura, T. Hoshino, A. Morita, M. Hattori and R. Okamoto, Membrane separation of supercritical fluid mixture, in: T. Yano, R. Matsuno and K. Nakamura (Eds.), Developments in Food Engineering, 2. Blackie, London, New York, 1994, 820 pp. 77. S. Sarrade, C. Perre, M. Carles, R. Veyre and G.M. Rios, Nanofiltration coupled with supercritical carbon dioxide. Interest and preliminary studies. ICOM'93, 30 August-3 75.
Sept. 1993, Heildelberg, Germany. 78.
79. 80. 81.
82.
83. 84.
S. Sarrade, G.M. Rios, C. Perre and M. Carles, Performance of nanofiltration under supercritical fluid conditions, in: Y.H. Ma (Ed.), Proceedings of the Third International Conference on Inorganic Membranes, 10-14 July 1994, Worcester, MA, p. 129. G.M. Rios, H. Rakotoarisoa and B. Tarodo de la Fuente, Basic transport mechanisms of ultrafiltration in the presence of an electric field. J. Membr. Sci., 38 (1988) 147. C. Guizard, N. Idrissi, A. Larbot and L. Cot, An electronic conductive membrane from sol-gel process. Br. Ceram. Proc., 38 (1986) 263. C. Guizard, F. Legault, N. Idrissi, A. Larbot, L. Cot and C. Gavach, Electronically conductive mineral membranes designed for electro-ultrafiltration. J. Membr. Sci., 41 (1988) 147. M.J. Van Der Waal, P.M. Van Der Velden, J. Koning, C.A. Smolders and W.P.M. Van Swaay, Use of fluidised beds as turbulence promotors in tubular membrane systems. Desalination, 22 (1977) 465. G.M. Rios, H. Rakotoarisoa and B. Tarodo de la Fuente, Basic transport mechanisms of ultrafiltration in the presence of fluidized particles. J. Membr. Sci., 34 (1987) 331. F. Clavaguera, E. Rjimati, S. Elmaleh and A. Grasmick, Intensification of microfiltration by a circulating bed, in Proceedings of ICIM 2. Key Eng. Mater., 61/62 (1991) 569.
Fundamentals of Inorganic Membrane Science and Technology Edited by A.J. Burggraaf and L. Cot 9 1996, Elsevier Science B.V. All rights reserved
Chapter 13
Applications of ceramic membranes in liquid filtration C.A.M. Siskens (Formerly: Hoogovens Industrial Ceramics BV) Ministry of Transport, Public Works and Water Management, Road and Hydraulic Engineering Division, Delft, The Netherlands
13.1 INTRODUCTION
The number of applications of ceramic membranes is immense and ever increasing. Many references on the use of ceramic membranes can be found in the proceedings of the two International Conferences on Inorganic Membranes [1,2] as well as in the excellent book of Bhave [3]; others [4-10] highlight developments since 1988. Except for a single reference, e.g. [11,12], direct data on the extent of installed ceramic membranes is rare. BCC's 1994 study "Inorganic membranes: markets, technologies, players" [13] estimates the inorganic membranes to grow to about 15% of the total separation membrane/module sales. This means that, worldwide in 2003, the sales in inorganic membranes are estimated at US$ 228 million, of which 69% is in ceramic membranes. These figures constitute an adjustment to earlier expectations [14], stating US$ 363 million in 2000. Clearly, great care should be exercised in using these values as data on market volumes are rather incongruous. Furthermore, detailed knowledge about commercial applications seems to be restricted: in the description of tests many publications deal with the potential of inorganic membranes rather than with ongoing industrial applications. Moreover, many market oriented publications are 'lost' in journals which are not abstracted in major data-bases. Both factors diminish the insight in the real scope of ceramic membranes. The limited nature of industrial application of ceramic membranes can be inferred too from the rather short description in some new textbooks like those of Gasper [15], Ripperger [16] and Ho and Sirkar [17].
620
13 - - A P P L I C A T I O N S OF CERAMIC MEMBRANES IN LIQUID FILTRATION
This reticence contrasts with the often quoted, many advantages of ceramic membranes: chemical and thermal stability, narrow pore size distribution, high porosity, high flux, mechanical strength (enabling back flushing), micro biological resistance, long lifetime, etc. [3-11,14]. In practice it turns out that the points of chemical and thermal stability are successful, even under seemingly moderate circumstances, as they permit the ceramic membrane to be cleaned much more thoroughly (and harsher!) than polymer membranes. This constitutes an extra advantage because it substantially adds to the economy of use: higher average flux, lower cleaning frequency, longer lifetime of the membranes. Despite all these advantages a real breakthrough has not been accomplished, and this can hardly be attributed to the disadvantages of higher price of production and brittleness. More probably for this still rather new product, the success of application depends on other marketing factors such as elemental applications research, engineering development and guidance of the customer. The application of a technology is mainly governed by its costs versus its benefits. The economical place of any separation process then depends on the type and value of the materials to be treated. Based on this line of thought, the applications in this chapter are classified into three groups, viz.: treatment of wastes: cleaning of waste streams to enable their disposal, regeneration: enabling the recycling/reuse of material, processing: treatment of process streams. It appears that in certain cases the category 'wastes' coincides with the category 'regeneration', e.g. in cases where the permeate water is of sufficient quality to be reused, or where a retentate may serve as a raw material for another process. The material presented in this chapter is based mainly on open literature dealing with the use of commercially available ceramic membranes and on technical data as acquired in Hoogovens Industrial Ceramics BV (HIC), the author's former company. This review is not meant to be exhaustive, but reflects examples of the use that ceramic membranes have found in certain industrial applications. 13.2
TREATMENT
OF
WASTES
13.2.1 Wastes of Oily Emulsions
13.2.1.1 Compressor-condensate In oil lubricated compressors a condensate is formed [18]. Such condensate is an oil in water emulsion with typical oil contents of 0.5%. In various countries discharge of more than 10 ppm oil is prohibited, e.g. in Austria since the beginning of 1992. In Germany, for new investments for waste water cleaning, the use of chemicals to break emulsions is forbidden. Furthermore, incineration of oily waste streams requires the highest oil content possible.
13 - - A P P L I C A T I O N S OF CERAMIC MEMBRANES IN LIQUID FILTRATION
621
With ceramic membranes (typical pore size 0.2 ~tm) this legal limit of 10 p p m is easily obtained [18]. Average membrane flux is 150 1/m 2 h; values are reported between 125 and 6001/m 2 h [19]. The interval between cleanings or the maintenance interval can amount to 1000 h. By combination of the membrane unit with a static separator the concentration factor can be as high as 180, the oil content in the final concentrate can amount to over 90%. The pay-back time for such an installation would be less than two years at a cost of NLG 9 4 / m 3 effluent, and related to the present costs of treatment [18].
13.2.1.2 Centralised Treatment of Industrial Emulsions In many industrialised countries oily wastes are collected and treated in commercial or public emulsion treatment centres. The supply of oil emulsions varies very considerably in type of oils, concentration, contamination with other materials, etc. Following coarse pre-filtration and decantation, oil/water emulsions can be treated very successfully with ceramic membranes. The concentrate is returned to the decanter, and microfiltered again after removal of the free oil, until all oil is removed. The extracted water can be fed into a biological treatment plant, or discharged directly, depending on the composition of the original emulsion a n d / o r local regulations. In a typical example [20] 6.4 m 2 of 0.2 ~trn membranes are used in a pilot scale operation, yielding average fluxes between 100 and 1251/m 2 h in a installation working at 55~ The concentration factor for the membrane installation varies between 6 and 12. Due to the extreme fouling nature of the feed, periodic cleaning is compulsory, but can be restricted to once a week. The system has been in operation since August 1992. The pay-back time is less than 3 years. A flux of 2001/m 2 h with the Ceramesh metal/ceramic composite membrane (0.2 ~tm pore size) on a metal working emulsion is reported by Cowieson and Gallagher [21]; similar data are given for Carbosep membranes [22]. Y6ksel et al. [23,24] describe the use of organic demulsifiers (ternary and quaternary polyamines) to enhance the breaking of oily emulsions. This method is particularly suitable when the composition of the oily waste water is fairly constant but it entails extra costs and maintenance. Ceramic membranes perform much better than polymer membranes because the latter get blocked by the polyamines.
13.2.1.3 Bilge Water Treatment Bilge water is the waste stream of (salt) water, fuel, oil, fats, detergents and others as found in the engine room of ships. The oil content of such water can be as high as 50%, the further composition cannot be quantified. Discharge of bilge water is a serious pollution item. Separation systems based on differences in density are not able to reach the discharge limit of 15 ppm as set by the new
622
13 - - A P P L I C A T I O N S OF C E R A M I C M E M B R A N E S IN LIQUID F I L T R A T I O N
regulations of the International Maritime Organisation, especially not under seagoing circumstances. Bilge water can be treated with HIC's ceramic membranes on shore or on board of the ship. In all cases a permeate with less than 15 ppm is reached with 0.2 ~tm membranes, even in the presence of detergents. Fluxes vary between 50 and 100 1/m 2 h. The treatment of bilge water and emulsions resembles that of the treatment of oil field brines and produced water. Chen et al. [25], using ferric chloride and other chemicals to enhance the performance of Membralox 0.2, 0.5 and 0.8 ~tm membranes, describe permeate fluxes between 1400 and 3400 1/m 2 h. Without pretreatment however severe fouling occurred as well as break-through of oil. Zaidi et al. [26] report about the continuation of this work. They quote fluxes between 800 and 12001/m 2 h , but also mention substantial lower fluxes in long term pilot tests using 0.8 ~tm membranes. In addition they indicate a drop in permeate flux caused by conditions of low pH, the presence of sea water, corrosion inhibitor, oil slugs or flow variations.
13.2.1.4 Vegetable Waste Water In the production of olive oil large amounts of the so-called alpechine (Spanish) or margine (French) are produced as waste (vegetable waste water). Depending on the extraction process 1.2-1.7 m 3 of waste water is produced per ton of olives. The treatment of this stream is becoming important as discharge into surface water or as an agricultural fertiliser is no longer acceptable. Alpechine is characterised by a low pH, a low content of nutrients and a high content of low-biodegradable organics, thwarting aerobic treatment or anaerobic-aerobic treatment. In comparison with these methods and in comparison with evaporation a treatment consisting of pre-filtering followed by microfiltration with ceramic membranes and a polishing step (ultrafiltration plus reverse osmosis) produces very good results. The microfiltration step of fresh alpechine on 0.2 ~tm HIC membranes attains fluxes from 90-125 1/m 2 h at temperatures between 30 and 50~ In the polishing step COD was reduced to approx. 1700 mg/1. This method constitutes an important economical advantage over other methods: e.g. Mendia [27], describing different methods of treatment, the use of evaporation [28,29] or biological treatment [30] and earlier experiments with polymer membranes [31] with a combination of UF, RO and adsorption. 13.2.2 W a s t e s B a s e d on S e m i - s o l i d s
13.2.2.1 Fish Factory Effluent In fish processing plants a large quantity of water is used as cleaning and transport medium. The water becomes polluted with fats, proteins, bones and
13 - - A P P L I C A T I O N S O F C E R A M I C M E M B R A N E S IN LIQUID F I L T R A T I O N
623
blood. The aim of microfiltration with ceramic membranes is to minimalize the sludge production on the one hand and to produce a dischargeable water on the other hand [32]. Over biological treatments this has the advantage of winning back part of the valuable material instead of transforming it into a sludge which has to be disposed of in landfills. Filtrating with 0.2 ~tm HIC ceramic membranes a COD reduction of over 60% was reached, at a content of suspended solids in the permeate of less than 10 ppm. Permeate flux was at a level of 150 1/m 2 h at a process temperature of 25~ Comparable results were found by Quemeneur and Jaouen [32].
13.2.2.2 Manure The disposal of pig manure poses problems comparable to those of the vegetable waste water. Pre-filtering is even more important here because of the very coarse nature of some of the manure components. As manure from pigs has solids contents as low as 5 to 11%, the main goal is the reduction of the amount to be transported from areas with a manure-surplus to regions with a fertiliser shortage. Very important too, but in the economical sense, is the possibility for disposal of the concentrate of the separation process. Local factors like fertilising limits and the nearby availability of fields that can be fertilised are decisive. Test results: starting from pig manure with a solids content of ca. 11% a vibrating screen separates this into a feed stream for the microfiltration containing 6% solids. On this feed, HIC's ceramic 0.1 ~tm membranes reach average fluxes of 80-100 1/m 2 h at filtration temperatures of 80~ The concentration factor can range between 2.5 and 3. Operating costs are below the DEM 2 0 / m 3 quoted by Meindersma [24]. The combined concentrate of pre-filter and MF is about 55% of the original volume and contains approximately 20% solids; the clear permeate of the MF contains approximately 2% solids, typically dissolved substances.
13.3 R E G E N E R A T I O N
13.3.1 Recycling of Solids from Suspensions 13.3.1.1 Ceramics Industry A good example of the filtration of hard, abrasive materials is the application of ceramic membranes in the cleaning of waste water of the ceramic industry [33]. Waste water in this industry typically contains clay, sand, glazes, etc. The use of microfiltration allows for the return of solids to the production
624
13 E A P P L I C A T I O N S OF CERAMIC MEMBRANES IN LIQUID FILTRATION
process and the recycling of water. Starting from a feed stream with 3-7% solids (density 1.02-1.05 g / c m 3) the use of 0.2 ~tm membranes permits concentration to a concentration of 27% solids (density 1.2 g / c m 3) at a flux of 220 1/m 2 h. To achieve higher levels of solids concentration consecutive stages would be necessary. Similar flux values are reported by Hoogland et al. [34] in the filtration of SiO2-slurries with Norton Ceraflo membranes (0.2 ~tm). They clearly show that at a pH value of about 7, flux is at a minimum. However this minimum flux is much less time-dependent than flux at higher or lower pH: at pH = 7 the flux decreases from 390 to 225 1/m 2 h in the first hour, whereas at pH = 2.5 the flux decreases from 1630 to 550 1/m 2 h in the same time. 13.3.1.2 Paint and Ink
In the paint and ink producing and applying industries three different water uses can be distinguished [35]: cleaning operations in the production process of solvent-based paints and inks, carrier and solvent in water-based paints and inks, - water-curtains to catch over-spray from water-based paints in spray booths. The waste water resulting from cleaning paint production equipment contains a high quantity of pigments and solvents. In order to reduce the waste water volume as well as to recover the pigments, tests were performed with ceramic microfiltration. The applied ceramic membrane [35] has pores of 0.2 ~tm and forms a barrier for the pigments. The waste water is recirculated across the membrane until a sufficient concentration is reached (12% dry matter). This enables treatment in a filter press, for reuse of the pigments in the production process. The water fraction permeates through the membrane (flux: 100-250 l / ( m 2 h). The water is colourless and contains no pigments. The system is compact, reliable, and can be fully automated. Moreover, the waste water can be treated batch by batch without any problems. The specific operational costs are about NLG 30-35/m 3, which is considerably lower than the waste processing costs, but also lower than the costs of alternative methods of treatment (including electro-flotation). The increasing demand for solvent-free paints led to the introduction of water-based paints. Both in the production and in the application of this type of paint quite often a water-paint waste mixture results. Two examples are: leftovers, diluted with (cleaning) water and water used in water curtains in spray booths, which becomes increasingly contaminated with paint. Using ceramic membranes with pore sizes of 0.1 micron [35], it is possible to concentrate the paint particles to a very high degree (35-65% dry matter) and at the same time produce an effluent containing less than 0.1% paint. This can be -
-
13 - - APPLICATIONS OF CERAMIC MEMBRANES IN LIQUID FILTRATION
625
reused or directly discharged into open waters or treated further with reverse osmosis before discharge. The permeate flux through the membranes varies from 50 to 250 1/m 2 h, dependent mainly on paint-type and concentration. Temperature ranges between 20 and 50~ Average cleaning interval is 4 weeks. The specific overall treatment costs are in the same range as in the above mentioned example, NLG 30-35/m 3. As in case of the solvent based paints this compares very favourably with the costs of other treatment possibilities and even seem to be lower than costs with polymeric membranes [36].
13.3.2 Lifetime Extension of Cleaning Baths 13.3.2.1 Alkaline Degreasing Baths Degreasing baths remove oil and other pollutants from the surface of metal components before this surface is treated. In due time the contents (1-2 m 3) of the degreasing bath become polluted and have to be exchanged for fresh cleaning solutions. This poses several problems: oil and dirt have to be separated from the discarded bath, the bath has to be neutralised before disposal, - changing the bath incurs high costs. Most often these actions are performed by specialised firms. Typical lifetimes of degreasing baths amount to 1-2 weeks. By drawing a continuous stream from the degreasing bath, and circulating it over a microfiltration system and concentration tank, oil and dirt can be retained in the concentration tank. Microfiltration with HIC's 0.2 ~tm membranes yields average fluxes of 250 1/m 2 h at temperatures from 40 to 70~ the pH lies between 9 and 11. The permeate contains less than 100 ppm oil. This treatment extends the life time of the degreasing bath up to five times; pay back time is less than two years. Similar data are reported of Carbosep, Le Carbone Lorraine and Atech membranes [19,22,37,38]. In these processes the retention of the detergents has to be monitored in order to keep their concentration in the degreasing bath at the correct level. -
-
13.3.2.2 Industrial Washing Operations In industrial cleaning of laundry, wool, leather, feathers, etc. large amounts of water and detergents are used. In the washing process this solution becomes polluted with fats, proteins, metals, etc., causing a high chemical oxygen demand (COD) and metal content. The use of ceramic membranes for laundry is necessitated in those cases where there is a risk for chemical contamination of the laundry. In the case of wool, leather, feathers it is the typical processing of
626
13 - - A P P L I C A T I O N S OF CERAMIC MEMBRANES IN LIQUID FILTRATION
fats and proteins which necessitates the ceramic membranes. By microfiltration, followed by reverse osmosis, it is possible to regenerate the washing fluid so it can either be reused or discharged. Microfiltration with 0.2 ~tm ceramic membranes (HIC) yields average fluxes of 125-150 1/m 2 h at temperatures from 40 to 70~ Suspended solids and concentration of hydrocarbons are both reduced to less than 10 ppm in the permeate. By RO the COD is reduced to below 100 mg/1. Cleaning interval for the microfiltration installation is once a week. The feed to the microfiltration system has to be filtered over 100 ~tm screens to prevent clogging of the equipment. Menjeaud [39], treating 7 m3/h washing water from a laundry for the printing and mechanical industry, achieves permeate fluxes decreasing from 250 to 501/m 2 h as the concentration factor increases from 2 to 25. The concentrate has such a high COD that it can be used as combustible. An important factor in this use of membranes is the detergents/surfactants retention and fouling of the membrane. Although Akay and Wakeman's review [40] only deals with polymeric membranes, it thoroughly describes the various parameters influencing the behaviour of surfactants. Y/iksel et al. [23] indicate that ceramic membranes are much less prone to fouling by surfactants than polysulfone membranes. Maleriat and Schlumpf [41] show the dependence of the retention of a detergent (dodecyl benzene sulfonate) on its concentration. At values below the critical micelle concentration the retention is low, above it, retention increases with concentration. This behaviour is further complicated by temperature dependence: with increasing temperature fluxes increase and retention decreases.
13.3.3 Recycling in Chemical Processes 13.3.3.1 Cleaning of Organic and Inorganic Reagents A typical example of the application of ceramic membranes in chemical industry is the cleaning of mono ethanol amine. Mono ethanol amine (MEA) is used for the absorption of H2S from acid gasses but is polluted during this process by various organic compounds. Filtration of the MEA over 0.2 ~tm HIC ceramic membranes at an average flux of 32 1 / m 2 h produces a clean, transparent yellow liquid, free of solids. Filtration temperature is 37~ pH is about 11.5. Tests lasted successfully for over 700 h. Another example is the filtration of TiO2 from a waste stream in the so-called sulphuric acid process [42]. Using Le Carbone Lorraine membranes (0.2 ~tm) a stable average flux of 250 1/m 2 h is reached at 5 bar transmembrane pressure and 30~ A plant of almost 300 m 2 is laid out on a flux of 2001/m 2 h. Cleaning is performed with HF (2%), I hour a day.
13 m A P P L I C A T I O N S OF C E R A M I C M E M B R A N E S IN L I Q U I D F I L T R A T I O N
627
13.3.3.2 Galvanic Baths In one example, parts from a nickelling bath are rinsed [43] with water. This rinsing water contains approximately 150 ppm nickel which is precipitated as Ni(OH)2 by addition of NaOH. The slurry resulting from this treatment passes a sedimentation tank and a filter press. This lowers the overall nickel content of the filtrate to ca. 3 ppm. A further treatment with ceramic membranes of 0.2 ~tm lowers the nickel concentration in the permeate to 0.2 ppm, which is well below the Dutch legal discharge limit of 0.5 ppm. The retentate is fed back to the sedimentation tank. The system of 2.4 m 2 treats about 800 1/h. Over polymer membranes the ceramic membranes show the advantage of a much longer interval between cleaning, viz., once per week instead of every day. 13.4 P R O C E S S I N G
13.4.1 Treatment of Liquid Products 13.4.1.1 Fruit Juices The application of ceramic membranes in the production of fruit juices is a well established technique [3,6,12,14,44-50]. A very wide range of fruit juices is designated (apple, pear, peach, orange, grapefruit, pineapple, kiwi fruit, strawberry, cranberry, carrot, beet); the clarification of apple juice seems to be the main application [6,14,44-49]. Merin and Daufin [44] present a review of older data, economical restraints being the limiting fluxes of the membranes and the impact of the short operating season in the production of apple juice. Gillot et al. [47], using ZrO2 0.1 ~tm Membralox membranes, clarify apple juice at fluxes between 200 and 250 1/m 2 h, concentration factor 10. Baur et al. [ 4 8 ] u s e 6 8 m 2 of 0.2 ~tm Membralox membranes and achieve fluxes of 100 to 150 1/m 2 h at temperatures ranging from 40 to 60~ concentration reaches 70 ~ Brix. Gupta et al. [49] report fluxes of 100 1/m 2 h using Norton Ceraflo and Le Carbone Lorraine membranes of 0.2 ~tm. They introduce pulsations in the circulation flow in order to enhance these fluxes.
13.4.1.2 Beer Brewing Yeast rests in fermenting cellars in beer breweries typically have a composition of 90% beer and 10% solids, mainly yeast. The amount of this waste material is 2-3% of the annual output. It can be sold as cattle feed or discharged. In a system with 4 m 2 0.4 ~tm ceramic microfiltration membranes, beer recovery amounts to 42-62%; the concentrate contains 23% solid matter [51]. Fluxes in
628
13 ~ APPLICATIONS OF CERAMIC MEMBRANES IN LIQUID FILTRATION
this recovery process are about 40 1/m 2 h at a temperature of 15~ Production runs vary from 8 to 16 h. The recovered beer can be blended back into the fermenting or lager cellar (in amounts up to 5%). In the treatment of tank bottoms [44,45,48,50,52] Finnigan et al. [52] reach long-term fluxes of approximately 20 1/m 2 h; Baur et al. [48] report average fluxes of 18 1/m 2 h (cleaning included) treating 500 h / d a y with 72 m 2 0.2 ~tm Membralox membranes. According to them circulation velocity and transmembrane pressure should be adapted to the yeast concentration because the secondary membrane layer completely governs the filtration process. Loss of this layer leads to blocking of the membrane and the temporarily passing of unwanted components.
13.4.1.3 Beer and Wine Clarification Publications on clarification of alcoholic drinks like wine and beer deal mainly with the treatment of wine. Advantages of ceramic membranes over classical methods are the reduction of operating costs (reduction of filter aids, less loss of product) and a better clarification. Ceramic membranes last longer and can be back flushed. According to Castelas and Serrano [53] microfiltration with pore sizes over 0.4 ~tm does not influence the wine, whereas pore sizes of 0.25 ~tm and lower disturb the organoleptic characteristics of the wine. However the complete removal of bacteria can only be achieved by 0.2 ~tm. Fouling of the membranes (Membralox) with coarser pore sizes, limits fluxes to 40-601/m 2 h, 0.2 ~tm is less affected and retains a flux of 85 1/m 2 h. Red wines seem to have a stronger tendency to fouling than white wines. Bauer [46] reports a decrease in fluxes from 275 1/m 2 h bar down to 11 1/m 2 h.bar for red wine and down to 1101/m 2 h bar for white wine, using Le Carbone Lorraine membranes. Similarly, but less extreme, Horgnies [50], in a very detailed description of commercial systems with Millipore Ceraflo 0.2 ~tm membranes, finds fluxes of 80-100 1/m 2 h for red wine and 100-120 1/m 2 h for dry white wine. Baur et al. [48] treating wine, rich in colloids, with 3.6 m 2 0.2 ~tm Membralox membranes report fluxes around 40 1/m 2 h. Belleville et al. [54] give a full description of the chemical nature of the fouling species. The use of enzymes to enhance the filterability might also be effective for raising the economy in the application of ceramic membranes [55]. Another means of achieving economical operation could be flux enhancement by the application of pulsating flow as outlined by Jaffrin et al. [56,57]. The simultaneous introduction of pulses, and a rise in circulation velocity from 3 to 4.37 m / s and of the transmembrane pressure from I to 4.5 bar, increases the flux from 35 1/m 2 h to 50 1/m 2 h.
13 - - APPLICATIONS OF CERAMIC MEMBRANES IN LIQUID FILTRATION
629
In the clarification of beer by cross-flow microfiltration the paper by Tr~igardh and Wahlgren [58] seems to be one of the sporadic examples of this application. Here the use of 0.5 ~tm membranes (Membralox) is necessary to maintain the taste of the beer; 0.2 ~tm shows an unacceptable retention of proteins and colour. Bacteria were retained by the 0.5 ~tm membrane. An editorial in "Filtration and Separation" [59] highlights the reluctance of the beer brewers to change over from kieselguhr filtration to membranes, but judges that the examples of fruit juice and wine production show good market prospects. 13.4.1.4 Potable Water
Drinking water is a major necessity of life: many membrane processes have been developed to produce it a n d / o r enhance its quality. Filtration [16] aims at the removal of: - suspended particles, precipitates caused by water hardness a n d / o r salts, micro-organisms: algae, bacteria, fungi. Examples of the use of ceramic membranes in the production of potable water are quite numerous [42,60-65]. An interesting review is presented by Pou6t et al. [60] of some 15 installations working with ceramic membranes for the production of drinking water. Sizes of these installations, installed in France between 1984 and 1990, vary from 5 to 100 m3/h. Moncorg6 and Pascal [61] and Bauer et al. [42] describe the use of the carbon/carbon composite membranes of Le Carbone Lorraine in the filtration of drinking water. With 0.2 ~tm membranes the fluxes range between 1000 and 2000 1/m 2 h at trans-membrane pressures from I to 2 bars. The use of Kerasep membranes [65] (Rhone-Poulenc's alumina/alumina membranes, 0.2 ~tm pore size) leads to fluxes of 600-1200 1/m 2 h at 2 bar transmembrane pressure. Micro-organisms form a very important source of fouling: various authors [66,67] report a strong decline in flux, even as much as 70% of the original value, in the presence of micro organisms. Moulin et al. [64] use a coagulant and ozone to enhance the flux of their 0.2 ~tm ceramic membrane (Membralox). Using ozone and a concentration of 45 ppm coagulant, flux is approximately 15001/m 2h, with 110 ppm coagulant flux increases to 20001/m 2 h. The ozone treatment decomposes the organic material, so fouling is kept at a minimum. The combination of electro-coagulation, flotation and microfiltration is applied by Pou6t et al. [60]. In this case Membralox 1P19-40 membranes are used: a zirconia/alumina composite membrane of 50 or 100 nm pore size. With 100 nm and the use of electro-coagulation fluxes stabilise on 250-350 1/m 2 h. In the treatment of river water Mietton Peuchot and Ben Aim [68] use polyaluminium chloride as flocculant, raising the flux of Membralox 0.2 ~tm membranes from 200 towards 800 1/m 2 h. -
-
630
13 ~ APPLICATIONS OF CERAMIC MEMBRANES IN LIQUID FILTRATION
13.4.2 Treatment of Semi-solid Products 13.4.2.1 Proteins Most of the processing of proteins with ceramic membranes is in the field of the dairy industry; some work with other proteins will be presented in Section 13.4.3. Merin and Daufin [44] and Bhave [3] present a comprehensive review of the field, the main use of ceramic membranes being protein concentration by microor ultra-filtration and bacteria removal by microfiltration. For the latter the Bactocatch process, as described by Gillot et al. [47], Merin and Daufin [44] and Bhave [3] forms an important example. At an average flux of 700 1/m 2 h 99.7% of the bacteria are withheld without retaining the proteins. The production of casein [12,69] is a good example of the processing of proteins with ceramic membranes. Surel and Famelart [69] delineate this process: using either 0.1 or 0.2 ~tm Membralox membranes fluxes are 54 1/m 2 h in the absence of calcium, decreasing to 31 1/m 2 h after addition of up to I g/1 of calcium. Retention of both 0~- and [~-casein by the 0.1 ~tm membrane is better, and is further enhanced by the addition of calcium. The very serious problem of fouling by proteins is corroborated by many publications [41,70,71]. Various parameters influencing the fouling behaviour have been studied. Clark et al. [70] discuss the influence of protein concentration, trans-membrane pressure, cross flow velocity and pH. For pore sizes of 0.1 ~tm (Membralox membranes), filtering bovine serum albumin, the flux has a minimum at the pH of the protein isoelectric point. Dumon and Barnier [71] show that the amount of protein adsorption depends on previous adsorption. Contacting with citrate or phosphate lowers a subsequent protein adsorption; contacting with nitrate increases the protein adsorption. Rios et al. [72] show that with small pores (< 0.2 ~tm) protein fouling remains on the outside of the membrane, whereas with the larger pore sizes the pores become blocked by the intrusion of protein into these pores. Cleaning after fouling by proteins is an important issue for the economical application of membranes. Kerherve et al. [73], Gillot et al. [47] and Daufin et al. [74] describe cleaning cycles, necessary to re-establish initial values of fluxes. Efficient cleaning was achieved by means of NaOC1 (with or without an acid step comprising HNO3) and by means of NaOH, 'reinforced' with complexing agents and surfactants.
13.4.2.2 Whey As whey contains many nutrient compounds (lactose, proteins, minerals and some fat) its use as starting material for the manufacture of various specialty products expands.
1 3 - APPLICATIONS OF CERAMIC MEMBRANES IN LIQUID FILTRATION
631
Microfiltration of whey prior to ultrafiltration in the production of whey protein concentrates (WPC) was reported among others by Maubois et al. [75], van der Horst [76], and Wnuk et al. [77]. The microfiltration step also prevents fouling of the UF-membranes (either polymeric membranes or ceramic membrane): e.g. Daufin et al. [78] by phosphates and calcium. Ceramic membranes yield higher fluxes (up to 691/m 2 h) and better separation then polymeric membranes, resulting in WPCs with lower fat contents [76]. Experiments at HIC with its 0.5 ~tm ceramic membranes showed fluxes of well over 200 1/m 2 h. Special importance however is required for any aggregation step preceding the microfiltration: both Gesan et al. [79] and Daufin et al. [78] emphasise the influence of the controlled aggregation by the addition of calcium. Gesan et al. [79] describe the performance of 57 m 2 Carbosep M14 membranes in the defatting of rennet whey, stressing the point that the performance has to be improved by a better control of fouling. Daufin et al. [78] show that through this microfiltration step the UF-step (Carbosep M5, 10,000 D) performs very good, yielding fluxes of 50 to 120 1/m 2 h, even with very high protein contents. Surel and Famelart [69], in a study with 0.1 and 0.2 ~tm Membralox-membranes, show that an addition of calcium lowers the MF permeate flux from 54 1/m 2 h to 20 1/m 2 h at a velocity of 6 m/s. This flux is quite dependent on velocity: 35 1/m 2 h at 4 m/s, 541/m 2 h at 6 m / s and 631/m 2 h at 7.3 m/s. Analogous to this processing, Korolczuk and Mahaut [80] report the necessity to use ceramic membranes for the filtration of acid-coagulated milk in order to produce UF-fresh cheeses with good taste. Typical fluxes, using Carbosep M1 (cut-off 50,000 D) increase from 10 to 20 1/m 2 h at 40~ with decreasing concentration factor.
13.4.2.3 Sugars Punidadas et al. [81] describe detailed experiments on the refining of raw cane sugar. Use of 0.2 m 2 SCT membranes with pore sizes between 0.1 ~tm and 0.5 ~tm effectively removed almost 100% of the solutions turbidity and 50% of its colour. Average flux is 38 1/m 2 h; the higher values are reached at higher operating temperature (90~ Interestingly, it is shown that working with the smaller pore sizes requires use of high tangential velocity from the very start of the process, whereas pore sizes of 0.5 ~tm and larger perform better after building a secondary membrane layer by first applying a modest velocity. The microfiltration process on its own is not sufficient for the complete purification of thecane sugar, however it prevents the ion exchangers from fouling and poisoning [82]. One of the side streams in the production of glucose from corn starch contains a high concentration of glucose. This very sticky suspension is heavily
632
13 ~ APPLICATIONS OF CERAMIC MEMBRANES IN LIQUID FILTRATION
contaminated with fats, proteins, fibres and other insoluble material and usually is used for animal feed. Filtration of this suspension with 0.2 ~tm ceramic membranes (HIC) at 45-60~ and pH = 4.5, produced a clear permeate at an average flux of 1901 / m 2 h.
13.4.2.4 Paper and Pulp The manufacturing process of paper and pulp consumes enormous amounts of water. Reduction of these streams by recirculation of the process water in the plants is of great environmental benefit. Typical waste waters are the so-called white water and wash water of paper recycling plants, containing ink. Tests with ceramic microfiltration membranes (HIC) show fluxes in the range of 100-200 1/m 2 h; pre-filtering of coarser components is indispensable. Treatment of the MF-permeate with ultrafiltration, or direct ultrafiltration, removes most of the high-molecular weight components which interfere with the paper-making process [83]. It was shown that the brightness of the produced paper increased by the use of Carbosep M5 (10,000 D) or other ultrafiltration membranes [83]. Due to the fouling of polysulphone membranes in bleach plant effluents Afonso and Pinho [84] studied the use of Carbosep membranes for the combination of ultra-filtration (10,000 D) and microfiltration (0.14 ~tm). The introduction of microfiltration preceding ultrafiltration improves the performance of the latter, regarding limiting fluxes at given feed circulation velocities.
13.4.3 Biotechnology Speaking about biotechnology the topics of the use of membrane reactors and the filtration of yeast, enzymes and proteins are discussed most often. Sometimes it is difficult to discern biotechnology from applications in more established industries like dairy, etc. Besides that, in many papers biotechnology is mentioned in a rather general sense [6,11,85-87], perhaps indicating the freshness of these processes a n d / o r some reluctance in communicating details about the application. In the filtration of fermentation broths, lysed yeast [21,88-91] microfiltration is used to separate the yeast cells a n d / o r cell fragments. For the Ceramesh ceramic/metal composite membrane of 0.2 ~tm pore size a flux of 60 1/m 2 h is reported [21] for lysed yeast, at a temperature of about 55~ and a solids concentration of up to 16-17%. The same magnitude of flux and solids concentration are obtained with whole yeast suspensions. Using Kubota membranes in the range of 50 nm to 0.8 ~tm Narukami et al. [88] choose 0.8 ~tm for their work with fermentation broth. They report a stable flux of 20 1/m 2 h using suction (0.8 bar) on the permeate side as driving force, whereas the flux decreases as a
13 m A P P L I C A T I O N S OF CERAMIC MEMBRANES IN LIQUID FILTRATION
633
function of time to 12.5 1/m 2 h when one uses 0.8 bar pressure in the common way of operation. This difference is attributed to the compaction of the cake layer on the membrane. Chang et al. [89] filtrate a alcohol-distillery waste with 50 nm and 0.4 ~tm membranes from TIA (France). Although the 0.4 ~tm membrane has a higher initial flux, the flux of the 50 nm membrane is always higher in the long run. At a concentration factor of 2, the flux for the 50 nm pore size amounts to 2151/m 2 h, for the 0.4 ~tm pore size it is 1851/m 2 h. The dependence of flux on concentration factor also depends on the type of raw material for the distillation process. A further example of the separation of bio-mass is given by Maebashi [90] in the filtration of the sediment of soy sauce production. The flux of a rotating TOTO 0.1 ~tm disk membrane ranges from 25 to 8 1/m 2 h as the concentration factor increases from 15 to 200, at a transmembrane pressure of 2 bar. Imasaka et al. [91] study the effect of gas-liquid two-phase crossflow filtration of bakers yeast. They employ tubular TOTO 0.2 and 0.5 ~tm membranes. They show that this way of operation significantly reduces the specific energy of the separation process. The separation of proteins and enzymes is performed with ultrafiltration membranes. Branger et al. [93] use Carbosep M1 and M4 (40,000 and 20,000 Dalton respectively) for the separation of enzyme hydrolysates. The fluxes with these membranes compare favourably with polymeric membranes: 37-102 1/m 2 h vs. 7-41 1/m 2 h. The use of Schott's porous glass membranes (pore sizes from 10 to 90 nm) in the separation of proteins with molecular weights from 14,400 to 450,000 is illustrated by Langer and Schnabel [85] who show a decrease in retention with increasing pore size for different proteins. Due to the chemical nature of the membrane material, it lends itself to surface modifications, including the coupling of enzymes or the attachment of micro-organisms. The separation of proteins can be improved by chemical modification of the membrane surface [94]. Coating a Carbosep M5 membrane (10,000 D) with quaternized polyvinylimidazole raises the retention of tetracycline from 25% towards 76%. Unfortunately the flux declines at the same time from 32 1/m 2 h to 7.6 1/m 2 h. The porous ceramic membrane can be used to either separate biologically reacting material in reactors, or carry catalysts, microbes or enzymes to influence the desired reactions. An overview of the Japanese efforts for the establishment of membrane reactors in the "Aqua Renaissance '90 Project" are summarised by Kimura [95]; a very recent review was written by Zaman and Chakma [96]. The preparation of microporous membranes (pore diameters smaller than 2 nm) for the application in membrane reactors is described by Keizer et al. [97] and Julbe et al. [98], however without detailing the membrane reactor itself.
634
13 - - A P P L I C A T I O N S OF C E R A M I C M E M B R A N E S IN LIQUID F I L T R A T I O N
Various Japanese researchers s h o w the possibilities of i m m o b i l i s i n g an enz y m e or yeast o n / i n a ceramic m e m b r a n e [99-101]. In the first example [99], the e n z y m e is b o u n d by Nakajima et al. to the ceramic surface of the TOTO 50 n m m e m b r a n e by activating it first w i t h a silane-glutaraldehyde technique. Invertase is then b o u n d to this activated surface and converts 100% of the 10-50 w t % sucrose in the feed solution. Alternatively glucose-isomerase yields a fructose ratio of 42% in a 45 wt% glucose feed at a residence time of 1000 s. The p r o d u c t i v i t y of such systems is 10-fold higher than in conventional columns in w h i c h the e n z y m e is immobilised in beads. H o r i t s u [100] immobilises yeast cells on the surface of a ceramic carrier by the different charge of carrier and cells. Using this set-up, soy sauce, beer a n d sake are p r o d u c e d w i t h fermentation times m u c h shorter, up to 10 times, than in conventional processing. Kawase et al. [101] s t u d y this immobilisation beh a v i o u r by m e a s u r i n g zeta-potentials and find a neat correlation of this potential w i t h the n u m b e r of a d s o r b e d cells.
REFERENCES
1. L. Cot and J. Charpin (Eds.), Proceedings of the 1st International Conference on Inorganic Membranes, 3-6 July 1989, Montpellier, France. 2. A.J. Burggraaf, J. Charpin and L. Cot (Eds.), Proceedings of the 2nd International Conference on Inorganic Membranes, 1-4 July 1991, Montpellier, France. Trans Tech Publications, Z/irich, 1991. 3. R.R. Bhave, Inorganic Membranes: Synthesis, Characteristics, and Applications. Van Nostrand Reinhold, New York, 1991. 4. H.P. Hsieh, Inorganic Membranes. AIChE Symposium Series, Vol. 84, 1988, lp. 18. 5. L. Cot, Ch. Guizard and A. Larbot, Novel ceramic material for liquid separation process: present and prospective applications in microfiltration and ultrafiltration. Industr. Ceram., 8 (1988) 143--148. 6. J. Charpin and P. Rigny, Inorganic membranes for separative techniques: from uranium isotope separation to non-nuclear fields, in Ref. [1], pp. 1-16. 7. K.K.Chan and A.M. Brownstein, Ceramic membranes m Growth prospects and opportunities. Am. Ceram. Soc. Bull., 70 (1991) 703-707. 8. J. Charpin, A.J. Burggraaf and L. Cot, A survey of ceramic membranes for separations in liquid and gaseous media, in: P. Vincenzini (Ed.), Ceramics Today Tomorrow's Ceramics. Materials Science Monographs, Vol. 66D. Elsevier, New York, 1991, pp. D2623-2638. 9. S.I.Nakao, Current status of inorganic membrane in Japan, in Ref. [2], pp. 219-228. 10. L.M. Sheppard, Porous ceramics: processing and applications, in: K. Ishikazi, L. Sheppard, S. Okada, T. Hamasaki and B. Huybrechts (Ed.), Ceramic Transactions; Vol. 31: Porous Materials. The American Ceramic Society, Westerville, OH, 1993, pp. 3-23. 11. G. Cueille and M. Ferreira, Place of mineral membranes in the processes for bio-industry and the food-industry, in: Ref. [1], pp. 303-309. 12. D. Bonneau, G.E. Brinkman and J. Guibaud, Nouveaux produits et d6veloppements dans la gamme Membralox, in: Ref. [2], pp. 313-318.
13-- APPLICATIONSOFCERAMICMEMBRANESIN LIQUIDFILTRATION
635
13. Business Communications Company: Inorganic membranes: markets, technologies, players; February 1994; press release. 14. A. Crull, Prospects for the inorganic membrane business, in: Ref. [2], pp. 279-288. 15. H. Gasper, Handbuch der industriellen Fest/Fliissig-Filtration. H~ithig Buch Verlag, Heidelberg, 1990. 16. S. Ripperger, Mikrofiltration mit Membranen: Grundlagen, Verfahren, Anwendungen. VCH Verlagsgesellschaft, Weinheim, Germany, 1992. 17. W.S.W. Ho and K.K. Sirkar, Membrane Handbook. Van Nostrand Reinhold, New York, 1992. 18. J.W. Wouters, Goedkope olie/waterscheiding met keramische membranen. Polytechn. Tijdschr., May (1994) 46--47. 19. T.A. Peters, Aufbereitung von Abwasser und Prozet~fliissigkeiten durch Ultrafiltration und Crossflow-Mikrofiltration mit atech-SiC-Membraben. Filtrieren Separieren, 7 (1993) 312-316. 20. C. De Haan and J.W. Wouters, Emulsieverwerking met keramische membranen. MilieuMarkt, Jan./Feb. (1994) 20-21. 21. D.R. Cowieson and P.M. Gallagher, Recent developments in formation and applications of metal mesh/ceramic composite membranes, in: Ref. [2], pp. 23-32. 22. Anon., Carbosep-Ultrafiltration zur Emulsionsspaltung. Filtrieren Separieren, 5 (1991) 387-388. 23. L. Yfiksel, R. Sedelies, U. Mfiller and B. Spei, A new concept in the economical treatment of waste emulsions: chemical emulsion breaking and membrane filtration, in: Proceedings 5th World Filtration Congress, Nice, France, 1990, pp. 286-292. 24. G.W. Meindersma, Membraanprocessen in de praktijk. PT-Procestechniek, 44 (1989) 44-47. 25. A.S.C. Chen, J.T. Flynn, R.G. Cook and A.L. Casaday, Removal of oil, grease, and suspended solids from produced water with ceramic crossflow microfiltration. SPE Prod. Eng., (1991) 131-136. 26. A. Zaidi, K. Simms and S. Kok, The use of micro/ultrafiltration for the removal of oil and suspended solids from oilfield brines. Water Sci. Tech., 25 (1992) 163-176. 27. L. Mendia, P. Carbone, G. d'Antonio and M. Mendia, Treatment of olive oil waste waters. Water Sci. Tech., 18 (1986) 125-136. 28. M.C. Annesini, A.R. Giona, F. Gironi and F. Pochetti, Distillation process for vegetable water treatment. Effluent Water Treat. J., 24 (1984) 450-455. 29. G. DiGiacomo, V. Brandani and G. DelRe, Evaporation of olive oil mill vegetation waters. Desalination, 81 (1991) 249-259. 30. S.G. Velioglu, K. Curi and S.R. Camlilar, Activated sludge treatability of olive oil-bearing wastewater. Water Res., 26 (1992) 1415-1420. 31. P. Canepa, N. Marignetti, U. Rognoni and S. Calgari, Olive mills wastewater treatment by combined membrane processes. Water Res., 22 (1988) 1491-1494. 32. F. Quemeneur and P. Jaouen, Potentialities of inorganic membranes in fish processing i n d u s t r i e s - economic aspects, in: Ref. [2], pp. 585-588. 33. Netzsch Filtrationstechnik GmbH, Waste water avoidance in the ceramic industry by means of microfiltration. Interceram, 42 (1993) 44-45. 34. M.R. Hoogland, A.G. Fane and C.J.D. Fell, The effect of pH on the crossflow filtration of mineral slurries using ceramic membranes, in Ref. [1], pp. 153-162. 35. W.F.J.M. Nooijen and F. Muilwijk, Paint/water separation by ceramic microfiltration. Filtration Separation, 31 (1994) 227-229.
636
13 - - APPLICATIONS OF CERAMIC MEMBRANES IN LIQUID FILTRATION
36. E. VanKooij, Spoelwater van verfindustrie bruikbaar als grondstof, Proces Technologie, 3 (1993) 20-23. 37. Anon., Crossing the micron flow, Filtration Separation, 30 (1993) 19-20. 38. M. Hansmann, Dynamisch filtrieren, Industrie-Anzeiger, 39 (1993) 19-21. 39. C. Menjeaud, Treatment and regeneration of waste water produced by industrial laundry with inorganic membranes, in: Ref. [2], pp. 589-592. 40. G. Akay and R.J. Wakeman, Ultrafiltration and microfiltration of surfactant dispersions An evaluation of published research, Trans. AIChE (Chem. Eng. Res. Des.), 71 (1993) 411-420. 41. J.P. Maleriat and J.P. Schlumpf, Interactions solute-membrane lors de l'ultrafiltration d'un tensio-actif, in: Ref. [2], pp. 482-484. 42. J.M. Bauer, J. Elyassini, G. Moncorge, T. Nodari and E. Totino, New developments and applications of carbon membranes, in: Ref. [2], pp. 207-212. 43. Novem, Demonstratieprojecten energiebesparing: projectresultaat 412, 1993. 44. U. Merin and G. Daufin, Separation processes using inorganic membranes in the food industry, in: Ref. [1], pp. 271-281. 45. J. Guibaud, Some applications of Membralox | ceramic membranes, in ref. [1], pp. 343-348. 46. J.M. Bauer, Utilisation de composites carbone-carbone dans la fabrication de membranes minerales de microfiltration et d'ultrafiltration, in Ref. [1], pp. 361-366. 47. J. Gillot, R. Soria and D. Garcera, Recent developments in the Membralox ceramic membranes, in: Ref. [1], pp. 379-381. 48. W. Baur, L. Gottkehaskamp and D. Oechsle, Die Anwendung von keramischen Membranen bei der Querstrom-Filtration in der Getr~inkeindustrie. Filtrieren Separieren, 6 (1992) 141-147. 49. B.B. Gupta, P. Blanpain and M.Y. Jaffrin, Permeate flux enhancement by pressure and flow pulsations in microfiltration with mineral membranes. J. Membr. Sci., 70 (1992) 257-266. 50. M.C. Horgnies, D6veloppement et application de la filtration tangentielle sur c6ramique en agro-alimentaire. Liquides Mag. (1994) 49-53. 51. Anton Steinecker Maschinenfabrik GmbH, Freising (BRD), Crossfow-micro-filtration with ceramic membranes, Company Product Bulletin, 1993. 52. T. Finnigan, R. Shackleton and P. Skudder, Using ceramic microfiltration for the filtration of beer and recoevry of extract. Filtration Separation, 26 (1989) 198-200. 53. B. Castelas and M. Serrano, Utilisation des membranes dans le traitement du vin, in: Ref. [1], pp. 283-290. 54. M.P. Belleville, J.M. Brillouet, B. Tarodo de la Fuente and M. Moutounet, Cross-flow microfiltration of a red wine on a alumina membrane: investigation on fouling colloids, in: Ref. [2], pp. 477-480. 55. R. Urlaub, Wirkungsspezifisch und gut steuerbar: Enzyme in der Weinbereitung. Die Erniihrungsindustrie, 1992 (1992) 6-10. 56. M.Y. Jaffrin, R. Ben Amar and B.B. Gupta, Membrane fouling control in cross flow filtration of wine with mineral membranes, in: International Technical Conference on Membrane Separation Processes, Brighton, UK, 24-26 May 1989, Paper E2. 57. M.Y. Jaffrin, B.B. Gupta and P. Paullier, Energy saving pulsatile mode cross flow filtration. J. Membr. Sci., 86 (1994) 281-290. 58. G. Tr~igardh and P.E. Wahlgren, Removal of bacteria from beer using crossflow microfiltration, in: Ref. [1], pp. 291-295.
13 - - APPLICATIONS OF CERAMIC MEMBRANES IN LIQUID FILTRATION
637
59. Editorial, Beer filtration- Drinktec Interbrau. Filtration Separation, 30 (1993) 521. 60. M.F. Pou~t, F. Persin and M. Rumeau, Etude du couplage electrocoagulation-flottation pour limiter le colmatage des membranes en traitement des eaux, in: Ref. [2], pp. 237-242. 61. M.F. Pou~t, F. Persin, M. Gros and M. Rumeau, Etude de pretraitements avant ultra ou microfiltration tangentielle, in: Ref. [2], pp. 549-552. 62. G. Moncorg6 and G. Pascal, Neueste Entwicklung an Kohlefaser-KohlenstoffMikrofiltrationsmembranen und deren Anwendung. Filtrieren Separieren, 6 (1992) 156--160. 63. C. Moulin and M. Rumeau, Potabilisation d'eau par microfiltration tangentielle sur membrane minerale, in: Ref. [1], pp. 515-518. 64. C. Moulin, M.M. Bourbigot and M. Faivre, Interest of the ozone/coagulant combination for the potabilization of surface waters by crossflow microfiltration on mineral membranes, in: Ref. [2], pp. 229-236. 65. Rhone-Poulenc, Der Einsatz von Kerasep-Membranen in der Wasser- und Abwasserbehandlung. Filtrieren Separieren, 7 (1993) 267-268. 66. F. Duclert and M. Rumeau, Microfiltration d'eau sur membrane minerale; influence de la qualit6 de l'eau, in: Ref. [1], pp. 493-496. 67. S. Elmaleh and W. Naceur, Transport of water through an inorganic composite membrane. J. Membr. Sci., 66 (1992) 227-234. 68. M. Mietton Peuchot and R. Ben Aim, Improvement of crossflow microfiltration performances with flocculation. J. Membr. Sci., 68 (1992) 241-248. 69. O. Surel and M.H. Famelart, Microfiltration of sodium caseinate on ceramic membranes, in: Ref. [2], pp. 509-512. 70. W.M. Clark, A. Bansal, M. Sontakke and Y.H. Ma, Protein adsorption and fouling in ceramic ultrafiltration membranes. J. Membr. Sci., 55 (1991) 21-38. 71. S.Dumon and H. Barnier, Ultrafiltration of protein solutions on ZrO2 membranes. The influence of surface chemistry and solution chemistry on adsorption, J. Membr. Sci., 74 (1992) 289-302. 72, G.M. Rios and P. Freund, Basic studies on transport and fouling phenomena during protein UF and EUF on alumina membranes, in: Ref. [1], pp. 171-176. 73. F.L. Kerherve, S. Rio, U. Merin, J.P. Labb6, A. Qu6merais, F. Michel and G. Daufin, Nettoyage de membranes d'ultrafiltration de lactoserum et de lait, in: Ref. [1], pp. 419-423. 74. G. Daufin, U. Merin, F.L. Kerherve, J.P. Labb6, A. Qu6merais and Ch. Bousser, Effinciency of cleaning agents for an inorganic membrane after milk ultrafiltration, in: Ref. [2], pp. 553-556. 75. J.L. Maubois, J. Pierre, J. Fauquant and M. Piot, Industrial fractionation of main whey proteins. IDF Bull., 212 (1987) 154-159. 76. Van der Horst, Microfiltration in whey processing, in: Ref. [1], pp. 297-302. 77. R. Wnuk, N. Stroh and H. Chmiel, Inorganic membranes in the food and biotechnology industries; A study on fouling inorganic membranes, in: Ref. [1], pp. 479-482. 78. G. Daufin, J.P. Labb6, A. Quemerais, F. Michel, J. Fauquant and J.F. Radenac, Optimizing pH for improved defatted whey ultrafiltration using an inorganic membrane, in: Ref. [2], pp. 557-560. 79. G. Gesan, U. Merin, G. Daufin and J.J. Maugas, Performance of an industrial microfiltration plant for defatting rennet whey, in: Ref. [2], pp. 307-312. 80. J. Korolczuk, and M. Mahaut, Rheological properties of UF-fresh cheeses, in: Ref. [2], pp. 491-494.
638 81.
82.
83.
84. 85. 86.
87. 88.
89.
90.
91. 92. 93. 94.
95. 96. 97.
98.
99. 100.
13 ~ A P P L I C A T I O N S OF C E R A M I C M E M B R A N E S I N LIQUID F I L T R A T I O N
P. Punidadas, M. Decloux and G. Trystram, Microfiltration tangentielle sur membrane min6rale en c6ramique. Application au traitement du sucre roux. Ind. Alim. Agricol., (1990) 615-623. M. Decloux, E.B. Messaoud and M.L. Lameloise, Etude du couplage microfiltration tangentielle/6change d'ions en raffinerie de sucre de canne. Ind. Alim. Agricol., (1992) 495-502. J. Nuortila-Jokinen, T. Uusluoto and M. Nystr6m, Removal of disturbing substances by ultrafiltration of make-up waters in the pulp and paper industry. Paper Timber, 76 (1994) 256-261. M.D. Afonso and M.N. Pinho, Membrane separation processes in pulp and paper production. Filtration Separation, 28 (1991) 42-44. P. Langer and R. Schnabel, Porous glass UF-membranes in biotechnology, in: Ref. [1], pp. 249-255. V.A. Lyalin and V.D. Alpem, Filtration tangentielle sur les membranes inorganiques: comment augmenter son rendement en biotechnologie et industrie alimentaire, in: Ref. [2], pp. 123-130. Anon., Membranes: Projet Eureka pour Tech-Sep. Informations Chimie (1992) 137-139. Y. Narukami, A. Kayawake, M. Shioyama, Y. Okamoto, K. Tokushima and M. Yamagata, Ceramic membrane filtration of methane fermentation broth, in: Ref. [1], pp. 267-270. I.S. Chang, K.H. Choo, C.H. Lee, U.H. Pek, U.C. Koh, S.W. Kim and J.H. Koh, Application of ceramic membrane as a pretreatent in anaerobic digestion of alcohol-distillery wastes, J. Membr. Sci., 90 (1994) 131-139. N. Maebashi, Ceramic membranes and application to the recovery of soy sauce, in: K. Ishikazi, L. Sheppard, S. Okada, T. Hamasaki and B. Huybrechts (Ed.), Ceramic Transactions; Vol. 31: Porous Materials. The American Ceramic Society, Westerville, OH, 1993, pp. 81-87. T. Imasaka, N. Kanekuni, H. So and S. Yoshino, Gas-liquid two-phase cross-flow filtration by ceramic modules. Kagaku Kogaku Ronbunshu, 15 (1989) 638-644. W.M. Clark, A. Bansal, M. Sontakke and Y.H. Ma, Protein adsorption and fouling of ceramic membranes during ultrafiltration, in: Ref. [1], pp. 415-418. J.L. Branger, R. Audinos, J. Noguera and M. Chignac, Ultrafiltration concentration of enzyme hydrolysates by mineral membranes, in: Ref. [1], pp. 243-248. B. Chaufer, M. Rollin, A. Grangeon and J. Dulieu, Tetracycline removal or concentration with an inorganic ultrafiltration membrane modified by a quatemarized polyvinylimidazole coating, in: Ref. [2], pp. 249-254. S. Kimura, Japan's Aqua Renaissance '90 Project. Water Sci. Tech., 23 (1991) 1573-1582. J. Zaman and A. Chakma, Inorganic membrane reactors, J. Membr. Sci., 92 (1994) 1-28. K. Keizer, V.T. Zaspalis and A.J. Burggraaf, Passive and catalytically active membranes for affecting chemical reactions, in: P. Vincenzini (Ed.), Ceramics T o d a y - - Tomorrow's Ceramics. Materials Science Monographs, Vol. 66D, Elsevier, New York, 1991, pp. D2511-2524. A. Julbe, C. Guizard, A. Larbot, L. Cot and A. Giroir-Fendler, The sol-gel approach to prepare candidate microporous inorganic membranes for membrane reactors, J. Membr. Sci., 77 (1993) 137-153. M. Nakajima, N. Jimbo, H. Nabetani and A. Watanabe, Use of ceramic membrane for enzyme reactors, in: Ref. [1], pp. 257-266. H. Horitsu, A new approach that uses bioreactors with inorganic carriers (ceramic) in
13 -- APPLICATIONSOF CERAMICMEMBRANESIN LIQUIDFILTRATION
639
the production of fermented foods and beverages, in: in: K. Ishikazi, L. Sheppard, S. Okada, T. Hamasaki and B. Huybrechts (Ed.), Ceramic Transactions; Vol. 31: Porous Materials. The American Ceramic Society, Westerville, OH, 1993, pp. 381-389. 101. M. Kawase, Y. Kamiya and M. Kaneno, Porous ceramic carrier for bioreactor, in: K. Ishikazi, L. Sheppard, S. Okada, T. Hamasaki and B. Huybrechts (Ed.), Ceramic Transactions; Vol. 31: Porous Materials. The American Ceramic Society, Westerville, OH, 1993, pp. 391-400.
This Page Intentionally Left Blank
Fundamentals of Inorganic Membrane Science and Technology Edited by A.J. Burggraaf and L. Cot 9 1996, Elsevier Science B.V. All rights reserved
Chapter 14
Feasibility of the application of porous inorganic gas separation m e m b r a n e s in s o m e large-scale chemical processes Henk M. van Veen, Maarten Bracht, Edwin Hamoen and Peter T. Alderliesten Fossil Fuels Department, Inorganic Membrane Group, Netherlands Energy Research Foundation, ECN, P.O. Box 1, 1755 ZG Petten, The Netherlands
14.1 INTRODUCTION During the last decade there has been intensified activity in research and development of ceramic membranes for gas separation applications. In several studies it is said that the market for these membranes will expand very rapidly in the near future [1-3]. This market growth will be due to advantages such as high permeation and membrane stability as compared with other membrane separation technologies. During the first years of inorganic membrane development, R&D was mainly focused on the membrane as the product, and research was driven by materials development and materials scientists. Research was carried out by universities, while research institutes and especially (end-user) industry were hardly involved. The main reason for this was that a lot of fundamental knowledge was needed before these membranes could be implemented in the foreseen market.
642
14
-
-
APPLICATION OF POROUS INORGANIC GAS SEPARATION MEMBRANES
Nowadays somewhat more attention is paid to the application and use of these membranes in processes and under process conditions. Industry is now getting more involved in R&D. In several review articles, membrane development and possibilities of inorganic membranes in gas separation applications [4-8] and especially in membrane reactor applications [9-16] have been summarised. In most of the literature the use of inorganic membranes for gas separation and reactor applications are considered to be very promising. However, research is still strongly focused on the membrane as a material and much less on the membrane process. As a result, insufficient data, especially on testing under realistic circumstances, are available on the real possibilities of inorganic membranes in large-scale processes. There is a need for such data and extensive technical and economic evaluations of membranes in different possible applications should be made, preferably using a multidisciplinary approach. Aspects such as chemical engineering and mechanical engineering are as important as materials engineering to introduce inorganic gas separation membranes into commercial processes. Furthermore, in order to introduce these membranes into the market successfully all aspects starting from fundamental material development to marketing strategies must be considered, depending, of course, on the state of development. If these aspects are taken into account it will become clear that the introduction of inorganic membranes in petrochemical and energy production processes is more difficult than first expected. The aim of this chapter is to show that a multidisciplinary approach, focusing on materials, processes and modelling as depicted in Fig. 14.1, is needed to judge the techno-economic feasibility of inorganic membranes in large-scale processes. This will be done by discussing examples of the potential use of porous inorganic membranes in three different membrane reactor applications.
Materials
/
'l
Processes
Modelling
Fig. 14.1. Disciplines to be c o n s i d e r e d .
14 -- APPLICATIONOF POROUS INORGANICGAS SEPARATIONMEMBRANES
643
In all three, hydrogen separation will take place: the dehydrogenation of propane to propylene, the dehydrogenation of ethylbenzene to styrene, and the water-gas shift reaction. Membrane characteristics such as permeation, selectivity and separation factor are given throughout this chapter. The definitions for these characteristics are given in the appendix.
14.2 B A C K G R O U N D I N F O R M A T I O N
14.2.1 Materials
Hydrogen selective inorganic membranes can be mesoporous (2 nm < pore diameter < 50 nm; ceramic, glass or Carbon) microporous (pore diameter < 2 nm; ceramic, carbon or zeolite) or dense (ceramic or metal). These membranes can be used from ambient temperatures up to about 600~ for mesoporous materials, up to about 500~ for microporous inorganic membranes and up to about 800~ for dense inorganic membranes [14-16]. These temperatures are only a rough indication, because of the different materials which can be used and the test conditions at which the membranes have to operate. Typical characteristics of both porous and dense inorganic membranes are given in Table 14.1. Only applications with porous ceramic membranes will be dealt with in this chapter. TABLE 14.1 Typical characteristics of inorganic gas separation membranes Membrane system
Pore diameter + (thickness)
Temp. (~
Gas mixture 1
Permeation 2 ( m o l / m 2 s Pa)
Permselectivity 3
Mesoporous alumina [17]
4 nm (3 btm)
25 250 475
H2/C3H8
6 x 10-5 4 x 10-5 3 x 10-5
3.0 3.7 4.0
Mesop~rous glass: Vycor .... (Toshiba) [18]
4.5 n m (300 l.tm)
20
H2/N2 and H2/CO2
7.4 x 10-8
Knudsen
Mesoporous carbon [19]
several nm (12-18 btm)
400
H 2 /C O-C O2- 2 x 10-6 H2S
3.5:CO* 4.5:CO2"
Microporous SiO2 on A120 3 by polymeric sols [20]
appr. 10 ~, (100 nm)
25 100 200
H2/C3I--I6
14 62 156
7 x 10-7 10 x 10-7 11 x 10-7
(continued)
644
14 -- APPLICATIONOF POROUSINORGANICGAS SEPARATIONMEMBRANES
TABLE 14.1 (continuation) Membrane system
Pore diameter + (thickness)
Temp. (~
Gas mixture 1
Permeation 2 (mol/m 2 s Pa)
Permselectivity 3
Microporous SiO2 on A1203 by polymeric sols + CVD [21]
appr. I nm (5 ~tm)
50 250
H2/N2 H2/N2
1.8 x 10-8 5.4 x 10-8
44 200
Hollow fibre microporous glass (PPG) [22]
4-8/~ (5 ~In)
204 260 316 371
H2/CO H2/CO H2/CO H2/CO
8.7 x 10-9 11.0 x 10-9 10.9 x 10-9 10.2 x 10-9
325* 205* 147" 101"
Microporous carbon molsieve; hollow fibre [23]
appr. 5/~ (6 ~xn)
20 20 200 500
O2/N2 He/N2 H2/CH4 H2/CH4
6 x 10-8 3 x 10-7 1.07 x 10-7 1.16 x 10-7
8 20 57* 35*
Silicalite on ceramic disc [24]
appr. 4.5/~ (5 ~lm)
20 20 20 20
H2/N2 H2/n-C4H10 N2/n-C4H10 N2/i-C4H10
2.3 x 10-7 2.3 x 10-7 2 x 10-7 2 x 10-7
3.1 146 15 55
Dense SiO2 by CVD dense (5 lirn) modification of microporous silica on alumina [25]
50 250 50 250 270
H2/N2 H2/N2 H2/CI-I4 H2/CH4 H2/N2
3 x 10-9 2.2 x 10-8 3 x 10-9 2.2 x 10-8
26 250 17.5 166 47*
Metal: Pd alloy on ceramic [26]
dense (6-8 ~tm)
440
H2/N2
1.6 x 10-6
>1000
Metal: P t / P d (80/20) on alumina [27]
dense (?)
100 200 300
H2/N2
1.2 x 10-7 3.6 x 10-7 7.5 x 10-7
5.6 37 200
1 The fastest permeating compound is mentioned first. 2 Permeation of fastest permeating compound. 3 If marked * then these figures are real separation factors. 9Separation factor is a function of the process variables and process circumstances.
I n o r g a n i c g a s s e p a r a t i o n m e m b r a n e s n o r m a l l y c o n s i s t of a s u b s t r a t e , o n w h i c h o n e or m o r e i n t e r m e d i a t e l a y e r s a n d a t o p l a y e r or g a s s e p a r a t i o n l a y e r h a v i n g K n u d s e n d i f f u s i o n s e l e c t i v i t y ( p o r e s of a b o u t 4 n m i n d i a m e t e r ) is a p p l i e d . T h e s e m e m b r a n e s w e r e d e v e l o p e d d u r i n g t h e l a s t t e n y e a r s i n m a i n l y flat a n d t u b u l a r c o n f i g u r a t i o n a n d b y u s i n g a l u m i n a as t h e b a s e m a t e r i a l . T h e y a r e
14 m APPLICATION OF POROUS I N O R G A N I C GAS SEPARATION MEMBRANES
645
now available on a semi-commercial scale. Because of the low price of organic membranes and the rather low selectivity of inorganic Knudsen diffusion membranes it will be hard to find commercial applications for these membranes, unless they can be used under conditions where organic membranes would not be able to operate, e.g. high temperature or chemically harsh applications. The Knudsen diffusion gas separation layer can be modified by e.g. sol-gel, cvd, or crystallisation techniques to enhance the selectivity, but this decreases the permeation. Silica is the material mainly used for modification. However, data on reproducibility and stability are still scarce. The large scale use of high selective inorganic membranes and these membranes at high temperatures, up to at least 600~ will probably last another 5-10 years. On a laboratory scale (maximum membrane surface area of about 50 cm 2) these high selective membranes are now available, although stability can be a problem in certain atmospheres. Only a few years ago it was recognised that research and development should also be focused on high temperature gas tight sealing, membrane systems/modules and decreasing of costs by e.g. the increase of membrane surface area to volume ratio. On a laboratory scale membrane sealing technology is now available up to temperatures of about 600~ [28,29]. Some work has been reported on the increase of membrane surface area to volume ratio for ceramic gas separation membrane systems [30]. However, difficulties are foreseen in scaling up and controlling this technology [30,31].
14.2.2 Membrane Reactors
Besides the application of inorganic membranes in stand-alone gas separation units, attention is focused on more process-integrated applications. In such configurations the separation function of the membrane can be used to shift the equilibrium of a chemical reaction by selective removal of one or more components on the product side of the reaction in a so-called membrane reactor. Four basic catalytic membrane reactor configurations, when the membrane and reactor are in the same physical unit, can be distinguished [32]: - a catalytic membrane reactor (CMR), in which the membrane is permselective to one or more components and is catalytically active; a catalytic non-permselective membrane reactor, where the membrane acts as the catalyst, but is not selective to any of the components; a packed bed or fluidized bed membrane reactor (PBMR or FBMR), in which the selective membrane is surrounded by a packed bed or fluidized bed of catalyst particles; - a packed bed or fluidized bed catalytic membrane reactor (PBCMR or FBCMR), in which the selective and catalytic active membrane is also surrounded by a packed bed or fluidized bed of catalyst particles. -
-
14 - - APPLICATION OF POROUS I N O R G A N I C GAS SEPARATION MEMBRANES
646
CO2
CATALYST 9
FUEL
9
9
9
9
9
CO
"i
~
9
o
9 'qlml~._o .,@.
@
9
H20
~e -
9
9
i
,O
q
9
9
9
9
9
9
9
9 9
9
9
9
9
9 9
9
9
9
9
9
9
@
9
9
9
9 9
9
9
9
%%'.~% .:. ';;, : i ~ " . "" = ~ ' ~ ~ : : , '
9
+
2
9
H2
9
21 9
~.
'!
9
9
9
9
9
9
9
9
o
9 9
9
9
9 9
9
9
9
9
9
O tOtO~I[O
9
9
9o . ;,"~
4 0
9
9
9
RICH STREAM
9
9
9 9
9
~'-.- -io9
9
9
9
~'~ C O
E E~T~ IN ~M
21 9 9
9
REACTION
9
+
9
H2
9
"'''''''!*-'"
(
RICH STREAM
9
9
9
9 9
9
9 9
9
9
9
9
_L_J 9
GAS
9 9
9
q 9B ' O 9 jl oil
e
, , ~ :,., , . - . - .9..'%". :~. - "o . . . . , ,9 9"j
f
R E A C T O R SHELL/ F i g . 14.2. S c h e m a t i c o f t h e w a t e r - g a s
shift membrane
reactor.
A schematic of a PBMR, in this case for the water-gas shift reaction, is given in Fig. 14.2. Of course the catalytic reactor and the membrane unit can also be separated from each other, but can still be used to enhance the yield of a catalytic process, as will be s h o w n in Section 14.3 (see also Fig. 14.5).
14.2.3 Membrane Process Modelling In the various feasibility studies presented in this chapter, models of membrane separation and membrane reactor systems play an important role. Models are being used for various reasons: not only because there is a lack of experimental data, or the calculations concern non-existing, ficfive membranes, they are also used to conveniently represent available data. In the various studies, different types of models have been used. However, the basis of all the models used is the same and will be discussed here. In a membrane permeator unit two important phenomena are encountered: transmembrane transport and flow around the membrane. In a membrane reactor a third p h e n o m e n o n is of importance: chemical reaction In the feasibility studies relatively simple models have been used because these concern mostly a rough estimate of the possibilities. For high selective and microporous types of membranes, permeation through the membrane is ass u m e d to occur only via diffusion which obeys Fick's law. In the case of the Knudsen diffusion membranes the contribution of the non-separating viscous flow through the membrane is also accounted for. The basis of the flow models are ordinary differential mass balances for each component on either side of the membrane. The mechanism for permeation is substituted in the mass balances. When reaction occurs the kinetic expression is also added to the balances. The chemical reaction is assumed to take place in
14 - - APPLICATION OF POROUS INORGANIC GAS SEPARATION MEMBRANES
647
close proximity to the membrane surface (passive membrane reactor system), and not in the membrane pore itself (active membrane reactor system). The latter case will lead to a completely different and more complicated description of the phenomena. Further steady-state conditions are assumed, so the p a r a m e ters are time-independent. The model as used here is treated extensively in Refs. [29,57]. Further information can be found in Refs. [61,68-71]. In the membrane reactor m a n y parameters influence the performance of the system. By making the model equations (mass balances)dimensionless [61], parameters are grouped so that a few dimensionless groups appear which describe the process. The physical meaning and their definitions are given in Table 14.2. A kinetic expression of the power law type for the reaction rate is assumed. The P e number is an important parameter which has an influence on the performance of the membrane process. Permeation and surface area are coupled via the P e number. In the equation QH2 is the permeation of the fastest permeating component (usually H 2 in this study). In membrane gas separation processes P e is usually between 0.1 and 1.0. For new applications P e = 0.5 can be taken as a first guess. The actual performance of the systems depends on m a n y more parameters than the P e number only, i.e. membrane selectivity, pressure drop, sweep gas flow to feed gas flow ratio, composition of the feed. TABLE 14.2 Dimensionless numbers and their meaning Parameter
Meaning
Definition
Peclet number
Ratio of total feed rate and maximum possible transmembrane flow rate
Pe =
Ratioof maximum conversion and the total feed rate (dimensionless residence time)
Da =
Ratio of permeation of H 2 and component i (permselectivity)
Si= Qi-
Damk6hler number
Si
~)
Ratio of permeate and feed side pressure
~ot AmQH2Pf V(Pf)~ k ~ot(RT) ~
QH 2
pP
Molar ratio of total sweep gas flow and total F~ot feed flow Y=
G
648
14 - - APPLICATION OF POROUS I N O R G A N I C GAS SEPARATION MEMBRANES
For the purpose of two of the studies described in this chapter a membrane separation model based on the characteristics given above has been implemented in the flow sheeting package ASPEN PLUS TM. This package allows the use of self-made user sub-routines and is therefore suitable for the implementation of the membrane model. The advantage of the use of the flow sheeting package is that the sensitivity of the total system performance to changes in membrane parameters can be determined quickly and that optimum process configurations can be found more easily. 14.3 GAS
SEPARATION
APPLICATIONS
FOR INORGANIC
MEMBRANES
In this section some examples of inorganic gas separation membranes in membrane reactor applications will be discussed. A first indication of the technical and economic feasibility of these membranes in dehydrogenation reactions and in the water-gas shift reaction will be given.
14.3.1 Dehydrogenation of Propane This section is written in close cooperation with Kinetics Technology International B.V. in Zoetermeer and Holland Industrial Ceramics in Velsen-Noord, The Netherlands [33]. 14.3.1.1 Introduction
Steam cracker plants based on naphtha a n d / o r gas-oil feedstocks are the major source of locally produced propylene in Europe and the Far East. In the United States approximately 90% of propylene comes from steam crackers and refinery operations. The balance comes from catalytic dehydrogenation units. The growth rate of propylene use is expected to be 3-4% worldwide. With the more conventional sources of propylene such as steam cracker operations and refinery operations, it is not possible to supply sufficient propylene for this growing demand. However, at the price levels of mid 1993 the economics of propane dehydrogenation are not very attractive. In recent decades various processes have been developed for catalytic dehydrogenation of propane to propylene [34-37]. These processes can be divided into two groups: - processes with an adiabatic reactor concept, and - processes with an isothermal reactor concept. Current commercial processes for catalytic dehydrogenation of propane to propylene are based on adiabatic reactor systems. Typical examples are: - the Catofin process (Lummus/Air Products); - the Oleflex process (UOP)
14 -- APPLICATIONOF POROUS INORGANICGAS SEPARATIONMEMBRANES
649
the fluidized bed dehydrogenation process (FBD) (Snamprogetti/Yarsintez) Recent developments in catalytic dehydrogenation have led to nearly commercial processes, using an isothermal reactor concept. Examples are: - the STAR process (Phillips) [35,37]; - the LINDE process (Linde/BASF) [36]. The potential benefits which can be achieved by using ceramic membranes in comparison to conventional propane dehydrogenation processes such as Oleflex and Catofin will be discussed here. -
14.3.1.2 Thermodynamics of propane dehydrogenation Besides several side reactions, the following main endothermic reactions are of importance in the dehydrogenatiort of propane to propylene" C3H8 ~ C3H6 + H2
(14.6)
C3H 8 ~
(14.7)
C2H 4 + CH 4
By selectively removing hydrogen from the reaction mixture, the reaction can be shifted beyond the original thermodynamic equilibrium. In this way reaction limitations can be overcome and the propylene yield enhanced. In Fig. 14.3 the equilibrium conversion for these two reactions as a function of temperature is given. From this figure it is concluded that: reaction (7) is more
60 C 3 H 8 - > C2H4 + CH4
50 O
-9 40
o
30
X 2o
H6 + H2
m 10 0 200
300
400
500
600
Temperature ~ Fig. 14.3. Thermodynamic equilibrium at I bar.
700
650
14 - - A P P L I C A T I O N OF P O R O U S I N O R G A N I C GAS S E P A R A T I O N M E M B R A N E S
675 ~ 4 O
"~ 3 650 ~
o 2
625 ~
0
100
200
300
400
500
Residence time (ms) Fig. 14.4. Influence of temperature on thermal cracking.
favoured from a t h e r m o d y n a m i c point of view; and for high propylene yields a high temperature is needed. Another important reaction which can take place at high temperatures is thermal cracking, which sets an upper limit to the reaction temperature. Therefore, a high-selective catalyst is necessary which only promotes the dehydrogenation and not the cracking reaction. The upper limit temperature of the cracking reaction has been determined by thermodynamic calculations using the p r o g r a m m e SPYRO | The results are given in Fig. 14.4 which shows that a temperature above 625-650~ leads to important thermal cracking reactions, which reduces the selectivity towards propylene, but also leads to increased coke formation, which deactivates the catalyst.
14.3.1.3 Adiabatic reactor concepts; reactor modelling evaluation In this modelling study only a packed bed m e m b r a n e reactor has been dealt with, because the regeneration of the catalyst and m e m b r a n e can be done separately, and also it will be easier to match the catalyst and m e m b r a n e surface necessary. Both Catofin and Oleflex use an adiabatic reactor concept. The Oleflex process uses four reactor beds in series, which as such is more suitable for addition of a ceramic m e m b r a n e separation unit than the Catofin process which uses a parallel reactor system. A comparison between the Oleflex process as a base case and an Oleflex process equipped with ceramic membranes is m a d e for the following cases:
14
-
-
651
APPLICATION OF POROUS INORGANIC GAS SEPARATION MEMBRANES
FeedPI"IReact~~---~modul 111.~IMembrane e l~,mo.du! e~ 2 = ~~Permeate....
IMembranemodul ~11 "=IReact~ .....e4,' .3,k,,
= IMembrane~1= !Reach~ 31~'=
,~Permeate
~Permeate
Fig. 14.5. Process flow diagram includhlg a membrane module after each reactor.
1. A 'Knudsen diffusion selective' m e m b r a n e after the first, second and third reactor (see Fig. 14.5). The permeation of the pure gases is inversely proportional to the square root of the molecular masses. 2. A 'Knudsen diffusion selective' m e m b r a n e after the third reactor only. 3. 'Ideal' membranes, which remove all the hydrogen formed in the reaction, after the first, second and third reactor. 4. 'Ideal' membranes, which remove all the h y d r o g e n formed in the reaction, after the first, second and third reactor and with increased outlet temperature. In cases 1 to 4 part of the reactor effluent is split off by the m e m b r a n e as permeate. The retentate stream, depleted in hydrogen, is then fed to the next reactor. After the fourth reactor m e m b r a n e permeate and reactor effluent are mixed again to be treated further in the d o w n s t r e a m section of the process. Permeation characteristics of 'Knudsen diffusion' membranes, consisting of a support and two consecutive layers, have been used to calculate the performance of the ceramic m e m b r a n e reactor, see also Section 14.2.1 [17,31]. The pore size of the separation layer of these membranes is 4 n m in diameter [31,38]. Ideal m e m b r a n e s which remove all the hydrogen formed do not exist (possible Pd-based membranes will come close to the required characteristics), but are used as a basis for calculating the m a x i m u m possible increase in conversion and selectivity. Two semi-quantitative models describing the reactor and m e m b r a n e performance were used to evaluate the overall performance. The reactor was modelled using the flow-sheeting package PRO II. A membrane model was used which describes both the transport through the membranes and transport along the membrane. These models are described in Ref. [33]. Based on the Oleflex process the following boundary conditions were chosen for the calculations: Plant capacity: Pressure after first reactor: Pressure after second reactor: Pressure after third reactor: Permeate pressure: Residence time in reactor: Permeate flow:
150,000 t / y e a r propylene 1.7 bar 1.5 bar 1.3 bar 1.1 bar 0.5 s 10% of feed flow
The results of the calculations are given in Table 14.3.
652
14-- APPLICATIONOFPOROUSINORGANICGASSEPARATIONMEMBRANES
TABLE 14.3 Adiabatic Oleflex based reactor performance (all figures on weight bases)
Conversion (%) Selectivity (%) Yield (%) Tin (~ Tout(~ Membrane area (m2)
Base case
Case 1 Case 2 Case 3 Case 4 Knudsen + Knudsen + Ideal + reactors Ideal + reactors reactors 1,2,3 reactor 3 1,2,3, constant T 1,2,3,higher temp.
47.0 73.9 34.7 650 595 -
42.2 74.0 31.2 650 594 475
46.1 74.0 34.2 650 595 313
49.4 75.7 37.4 650 583 -*
54.0 74.5 40.3 685 595 -*
* Membrane area not calculated because no estimation of the permeation for ideal membranes has been made.
F r o m Table 14.3 it is clear that in process configurations w i t h K n u d s e n diffusion selective m e m b r a n e s a d r o p in yield is obtained, as c o m p a r e d w i t h the base case. A p p a r e n t l y , the use of K n u d s e n diffusion m e m b r a n e s u n d e r the chosen conditions in these configurations is not attractive d u e to the relatively large a m o u n t of p r o p a n e p e r m e a t i n g t h r o u g h the membrane. With 'ideal' m e m b r a n e s (Cases 3 and 4) positive effects are observed. In Case 3, w i t h the same heat input as in the base case, the increase in yield is limited. H o w e v e r , in Case 4, w i t h a higher inlet temperature, higher yields are obtained and m a x i m u m profit of the m e m b r a n e is made. F r o m this w e can conclude that: m e m b r a n e s w i t h a selectivity higher than K n u d s e n diffusion are needed; the process conditions s h o u l d be changed in order to increase the m e m b r a n e separation performance, and the d e h y d r o g e n a tion reaction kinetics seem fast e n o u g h to react on the h y d r o g e n removal, at the chosen residence time of 0.5 s. In order to increase the m e m b r a n e separation performance there are t w o possibilities: (1) increase the m e m b r a n e permselectivity, to values higher than for K n u d sen diffusion; (2) increase the driving force for separation across the m e m b r a n e (a higher d r i v i n g force for separation means a m e m b r a n e process w i t h higher separation factors, at the same m e m b r a n e permselectivity) by: increasing the feed pressure, increasing the a m o u n t of h y d r o g e n in the feed; using a sweep gas at the p e r m e a t e side; reducing the permeate pressure. Since the driving force for h y d r o g e n transport is low, a m o d e r a t e increase in m e m b r a n e selectivity (to a permselectivity of 10 for H 2 vs C3H 8 a n d C3H6) has s h o w n to have h a r d l y any influence on the performance. Furthermore, the
14 m APPLICATION OF POROUS INORGANIC GAS SEPARATION MEMBRANES
653
membrane surface needed will increase because it is assumed that the increase in selectivity is obtained by a decrease in permeability of all the components, except hydrogen whose permeability is constant. An increase in membrane surface means an increase in costs. An increase in feed gas pressure is not attractive since the reaction conversion drops significantly with an increase in the feed gas pressure. As can be expected, the use of extra hydrogen in the feed has a negative effect on the conversion. Potential sweep gas candidates are steam and propane. Unfortunately steam will permeate in reverse through the membrane (when the membrane has a rather low selectivity) and deactivate the catalyst in the next reactor. Propane as a sweep leads to a significant change in the hydrogen and propane quantities in the retentate. The propane recycle which is required to use propane as a sweep gas, leads to a significant increase in utility consumption. Finally, we have calculated the effect of permeate pressure reduction. In Table 14.4 the results for the base case and Case 2 are compared with a new case 5, in which Knudsen diffusion membranes have been used only after the third reactor and in a process having a permeate pressure of 0.3 bar. Permeate pressure reduction leads only to a marginal improvement in yield. The only possibility of using inorganic membranes in an adiabatic reactor concept for dehydrogenation of propane is to use membranes with a selectivity much higher than Knudsen diffusion, in combination with a reduced permeate pressure. In this case, hardly any reactant will be lost through the membrane and the driving force for hydrogen transport will be high enough. Results of calculations for this combination will be reported in future. Possibilities for the use of inorganic membranes in an isothermal concept may lead more easily to a technically feasible process, because extra heat for propane conversion is available. Detailed flow sheeting calculations for the TABLE 14.4. Adiabatic Oleflex based reactor performance (all figures on weight bases) Base case
Case 2 K n u d s e n + reactor 3
Case 5 K n u d s e n + reactor 3
Conversion (%)
47.0
46.1
46.6
Selectivity (%)
73.9
74.0
74.0
Yield (%)
34.7
34.2
34.5
Tin (~
650
650
650 593
Tout (~ M e m b r a n e area (m 2)
595 -
595 313
Feed pressure (bar)
-
1.3
1.3
Perm. pressure (bar)
-
1.1
0.3
60
654
14 - - A P P L I C A T I O N OF POROUS I N O R G A N I C GAS S E P A R A T I O N MEMBRANES
integrated process are not yet available. However, to obtain a first indication of the economic feasibility of this concept, laboratory-scale membrane data are being used for performance estimation, see Section 14.3.1.4. In general, membranes to be applied should be stable under working conditions. Also, coke formation on the membranes should not lead to dramatic reduction of permeation and selectivity and regeneration with steam should not be a problem [39]. Another technical constraint can be the connection of membranes to the metal housing.
14.3.1.4 Isothermal reactor concepts; economic evaluation An isothermal reactor concept incorporating a ceramic membrane is more attractive compared to an adiabatic reactor concept from a thermodynamic point of view. In this concept we assumed a reactor with reactor tubes located in a direct-fired heater and operated in a cyclic way to remove coke formed on the catalyst. Parallel bed and heaters have been assumed [35-37]. On behalf of KTI an experimental programme on these reactor concepts has been started at the University of Southern California (USC). Some of the experimental results, concerning the use of Knudsen diffusion membranes are available in the literature [32,40]. These data have been used to calculate the economics of an isothermal propane dehydrogenation membrane reactor concept and are compared with the commercial Oleflex and Catofin processes, based on an adiabatic concept. The experimental circumstances of these lab-scale experiments, especially residence time, pressures and gas composition are not the same as in commercial, large-scale processes. However, we do not expect these differences to have a great influence on the results of the work presented here. Two process flow diagrams have been developed for a ceramic membrane reactor process: - the CMRL process: a process based on the commercial Oleflex process with a low propane conversion and Knudsen diffusion membranes - the CMRH process: a process based on the commercial Catofin process with a high propane conversion and Knudsen diffusion membranes The operating characteristics of these processes are given in Table 14.5. The design capacity of the plant is 150,000 M T / y e a r polymer grade propylene, which is equivalent to a production of 18,750 k g / h (8,000 h/year). The basis of the economic evaluation is the comparison of operating and investment costs for a membrane reactor with those for a conventional dehydrogenation plant. The return on investment (ROI) and the propylene production costs of the different processes have been calculated. The results are summarised in Table 14.6. Details of the calculations are reported in Ref. [33]. In the calculations a propane price of 130 $/tonne and a propylene price of 330 $ / t o n n e has been assumed [33].
14-- APPLICATIONOFPOROUSINORGANICGASSEPARATIONMEMBRANES
655
TABLE 14.5 Operating characteristics
Reactor type Conversion (%) Selectivity (wt%) H2/feed ratio Pressure (bar) Inlet temp. (~ Outlet temp. (~ LHSV (h-1) Reactor volume (m3)
Oleflex [34]
Catofin
CMRL [32,33,40] CMRH [32,33,40]
adiabatic 35.0 77.0 0.8 1.6-1.1 625 520-580 2.5 55.7
adiabatic 54.9 69.4 0.0 0.5 650 520-600 1.2 82.0
isothermal 38.0 89.0 0.2 1.15 560 560 2.5 44.4
isothermal 53.4 78.6 0.2 1.15 580 580 2.5 36.1
TABLE 14.6 Production cost breakdown (in US$)
Propane feed Co-product credit Utilities, catalyst, chemicals Fixed expenses Full production costs Dep recia tion Accounting production costs Selling price Overall margin ROI resulting from overall margin (%)
Catofin
Oleflex
CMRL
CMRH
203 -55.8 60.3 41.1 248.7 66.2 314.9 330 15.1 1.5
183.0 -47.3 66.8 42.7 245.2 69.2 314.4 330 15.6 1.4
171.6 -28.8 67.1 37.6 247.5 59.8 307.3 330 22.7 2.4
177.9 --44.3 59.3 37.2 230.1 59.0 289.1 330 40.9 4.4
F r o m Table 14.6 it can be seen that Catofin a n d Oleflex give a b o u t the s a m e ROI. This ROI is not v e r y attractive. The C M R L gives a n ROI of a b o u t 2.4%. The C M R H case gives an absolute increase in ROI of 3% p o i n t s c o m p a r e d to c o m m e r c i a l adiabatic processes. A sensitivity analysis of the ROI on b o t h the feed costs a n d the p r o d u c t v a l u e s is p e r f o r m e d . For the Oleflex a n d C M R H case these results are s u m m a rised in Figs. 14.6 a n d 14.7 w h i c h indicate that the ROI of s u c h a p r o p a n e d e h y d r o g e n a t i o n unit is not attractive w h e n the price difference b e t w e e n prop a n e a n d p r o p y l e n e is less t h a n a b o u t 250-300 $ / t o n n e . At m i d 1993, price levels of 330 $ / t o n n e p r o p y l e n e a n d 130 $ / t o n n e p r o p a n e , the process is n o t e c o n o m i c a l l y viable. Historical price levels s h o w t h a t a price difference of 300
14 - - APPLICATION OF POROUS INORGANIC GAS SEPARATION MEMBRANES
656
11
$ / ton
,' . . . .
10~-
110
9P
120
8
130 140
6-
5~ 4~ 3r 2 ~10 -1
C3H8 Price I
'
-2
,
-3 ~~ "i
.5 [,
,
9
.
0
.
|
9
290 300 3~0 a20 330 340 3so 3;0 370 380 390 C3H6 Price [$1tonj
Fig. 14.6. Influence of propane/propylene price on Oleflex return on investment.
$/ton
15 14 13 12 11 10 9 8 7
110 120 13o 1 o
5 4 3 2 1 0 -1 -2 290
300
310
320
330
340
350
360
370
380
390
C 3 H 6 Price IS/toni
Fig. 14.7. h l f l u e n c e o f p r o p a n e / p r o p y l e n e
price on CMRH return on investment.
$ / t o n n e has not been encountered d u r i n g the last 3 years. It is concluded that a ceramic m e m b r a n e reactor based on K n u d s e n diffusion m e m b r a n e s can give i m p r o v e m e n t s in an isothermal reactor concept a l t h o u g h the difference in price level b e t w e e n feedstock and p r o d u c t is too small to give an economically viable process.
14 - - A P P L I C A T I O N OF P O R O U S I N O R G A N I C GAS S E P A R A T I O N M E M B R A N E S
657
Following the results of the adiabatic reactor concept it is expected that high selective membranes will further improve the economics. However, it should be recognised that the process conditions in an isothermal concept are more severe than in an adiabatic concept. In particular, decoking conditions can be a problem in using high selective membranes. Detailed calculations on the isothermal membrane reactor concept are being performed and will be reported in future.
14.3.1.5 General conclusions propane dehydrogenation The selectivity of Knudsen diffusion membranes is not high enough to give a technically and economically feasible ceramic membrane reactor process for the dehydrogenation of propane to propylene based upon an adiabatic reactor concept. Measures such as an increased driving force or a moderately increased selectivity do not lead to positive results, because the driving force for hydrogen separation under the chosen process conditions is not high enough. Probably the only possibility is the combination of a high driving force (sweep gas or low permeate pressure) and a very high selective membrane. The use of ceramic membranes in an isothermal reactor concept shows better prospects. This process, in combination with high selective membranes and the necessary membrane boundary conditions are being studied, and the results will be reported in future. Propane and propylene prices are the main actors in the introduction of a dehydrogenation process in general, thus also for processes based upon membrane reactors. At a price difference (propylene-propane) of 300 $/tonne or less membrane based dehydrogenation processes will hardly be economic feasible.
14.3.2 Dehydrogenation of Ethylbenzene to Styrene 14.3.2.1 Introduction Next to ethylene, propylene and vinylchloride, styrene is one of the most important monomers for the production of plastics. The worldwide demand for styrene in 1992 was 18.2 million tonnes and is expected to grow annually with 3-5% to 23.9 million tons in 2000 [42]. Recent production statistics show an annual production of about 1.3 million tons of styrene in the Netherlands. Approximately 75% of this is produced at DOW Benelux in Terneuzen by catalytic adiabatic dehydrogenation of ethylbenzene [42]. The conversion of the endothermic reaction by which styrene is produced from ethylbenzene is mainly limited by temperature and thermodynamic equilibrium. The conversion to styrene increases with temperature, decreases with pressure and with dilution of an inert component like steam.
658
14 APPLICATIONOFPOROUSINORGANICGASSEPARATIONMEMBRANES -
H
-
H
H\
I
~
I
/H
./O~C\
H " - - C . - - C "--H
I H -
+
Fig. 14.8. The dehydrogenation of ethylbenzene to styrene.
When producing styrene from ethylbenzene several reactions besides the main reaction take place. Six reactions are of importance; these include the production of toluene, benzene, ethylene and methane and the thermal cracking of ethylbenzene (coking) [43]. This last reaction is the main reason for the upper temperature limit of 630~ On the other hand, high temperatures favour the dehydrogenation reaction, so the process takes place between approximately 570 and 630~ The dehydrogenation reaction is presented in Fig. 14.8. As with the dehydrogenation of propane, removing hydrogen from the reaction mixture may shift the conversion beyond the reaction equilibrium to the product side, obtaining higher selectivities to and yields of styrene. In the literature several experiments and some modelling results are presented about the possibilities of membrane reactors in the dehydrogenation of ethylbenzene. The results vary from a small increase in yield and selectivity [39,44] to very large increases in yield up to 20% [45--49]. In this study the feasibility of implementing ceramic membranes on an industrial scale in the styrene production process is treated. Therefore, a model has been set up in the flowsheeting package ASPEN PLUSTM,which describes a styrene process production plant. Some modelling has been done with different types of membrane reactors in different reactor section configurations to investigate the influence on the performance of the production of styrene.
14.3.2.2 Conventional process description This work focuses on the reactor section of the styrene production process because it is the most promising part for the implementation of membranes. The reactor section of this process is shown in Fig. 14.9 [50]. The process uses two radial reactors in series with one preheater and one interstage heater. Steam is used as an energy carrier (adiabatic reactor) and diluent [43,50,51]. Reactor temperatures and pressures are 570--630~ and 1.5 bar, respectively. Total hydrocarbon mass flow (96 wt% ethylbenzene) is 95,000 kg/h. The steam/hydrocarbon ratio is 2. Typical conversion, selectivity and yield numbers are 71, 92 and 66%, respectively. Definitions are given in the appendix. Reaction equations and kinetics are taken from literature [43,51].
14 - - APPLICATIONOF POROUSINORGANICGASSEPARATIONMEMBRANES
659 Raw styrene
Ethylbenzene
A
F e e ~ @ ~
C = Heat exchanger
B Fig. 14.9. Reactor section of the styrene production process.
14.3.2.3 Implementation of membranes The packed bed ceramic membrane reactor configuration (PBMR) has been chosen as the reactor set-up (see Section 14.2.2). In the PBMR configuration three possible sub-configurations can be envisioned for a specific sweep gas in combination with a hydrogen or oxygen selective membrane for the dehydrogenation of ethylbenzene. These sub-configurations are shown in Fig. 14.10. In sub-configuration (A) hydrogen will permeate through the hydrogen selective membrane tube under the influence of a pressure difference over the membrane and it will be carried away with an inert sweep gas (steam). The partial pressure of hydrogen in the reaction mixture will decrease and the equilibrium will shift to the product side. In sub-configuration (B) the permeated hydrogen will be swept away with air. Hydrogen will be burned and the heat generated by this exothermic reaction flows through the membrane to the reaction mixture. In this way the reactor will get an isothermic character and therefore higher conversions. The third sub-configuration (C) uses oxygen permeable membranes instead of hydrogen permeable membranes. Again air is used as an oxygen source in Catalyst
Sub-conflgumtlon A
Sub-configumtlon B .
,
Steam
~,\
'
Fer
o
~ o
Steam___~,.
o
~
/(
' o o o~
9
/
o o P~du~ o
"~'12
i
o
-~ -~
Air
Fe(~l
Air
.
.
.
.
.
.
.
.
.
.
'
' '
i... o
~'-
o o
~
~
"o o o
o o P[oduct o
o'-P
~LI2 ill
Sul~conflguratlon C
Air
Feed I ~ ~ 9
o
o" o~
_oo~o~ Oo
Air
02
i
/
i
i
i
i
i
/ ' Wall Permeate side
Fig. 14.10. Membrane reactor sub-configurations.
....
660
14 - - A P P L I C A T I O N OF POROUS I N O R G A N I C GAS S E P A R A T I O N M E M B R A N E S
the annular space of the reactor. Now oxygen will permeate through the membrane into the reaction mixture were it will burn the hydrogen formed. In this way there will be less loss of heat compared to the second case. Sub-configuration (C) is a principally different process: oxidative dehydrogenation. The most important disadvantage of the last two configurations is that not only will hydrogen be burned, but also hydrocarbons such as styrene and ethylbenzene. It is assumed that in sub-configuration (B) the membranes do not have an infinite hydrogen selectivity. For this reason we have chosen to focus our investigation on the first reactor sub-configuration (A). Although, especially sub-configuration (C), the oxidative dehydrogenation process seems very promising if a catalyst active only for H 2 oxidation (and not CxHy oxidation) is developed. By implementing the membrane reactor in the process according to sub-configuration (A), the conventional process is changed as little as possible. Again several configurations for the reactor section are possible. The first is to implement membranes in the first reactor (see Fig. 14.9) and leave the rest of the process intact. The second possibility is to implement the membranes in the second reactor and leave the first intact. Another option is to leave both reactors intact and to implement membranes between the two reactors. Because of the expected high costs of ceramic membranes, implementation in both reactors will probably be too expensive in relation to the possible advantages. For modelling the styrene process in ASPEN PLUS TM, several assumptions have been made: the radial flow reactors are estimated by plug flow reactors; all reactions are catalytic and only the main reaction is reversible; under these conditions coking is negligible; - the pressure drop in the reactors is negligible; heat transfer through the membrane is posed ideal; the sweep gas flows concurrently with the reactant gas; a small part of the steam which is originally used as carrier gas and heat carrier, is now used as sweep gas. The total amount of steam used, stays the same; the standard pressure at the permeate side of the membrane reactor is 0.1 bar. In modelling we used (if possible) permeations and selectivities based upon real measurements on different types of membranes [17,26,27,29,31,38] (see also Section 14.2.1). These membranes are: - Knudsen diffusion membranes, as developed by ECN; - microporous, highly selective membranes, as developed by ECN; and - palladium membranes [26,27]. We also formulated a non-existing, hypothetical membrane to see what would be ultimately possible. The performance of these membranes is in -
-
-
-
-
-
-
14 - - A P P L I C A T I O N OF POROUS I N O R G A N I C GAS SEPARATION MEMBRANES
661
TABLE 14.7 Permeations and permselectivities of the membranes used Knudsen
H2 Ethylbenzene Styrene Toluene Benzene H20
Microporous
Palladium
Hypothetical
Q*
S*
Q*
S*
Q*
S*
Q*
S*
13.10-6 1.8.10-6 1.9.10-6 2.0.10-6 2.2.10-6 4.6.10-6
1 7 6.9 6.5 5.9 2.8
1.10-6 2.10-8 2.10-8 2.10-8 2.10-8 1.10-7.
1 50 50 50 50 10
0.8.10-6 5.10-9 5.10-9 5.10-9 5.10-9 8.10-9
1 160 160 160 160 100
1.10-6 2.10-9 2.10-9 2.10-9 2.10-9 1.10-8
1 500 500 500 500 100
*Q = Permeation (mol/m 2 s Pa) and S = permselectivities (PermH2/Permx). principle the same as of m i c r o p o r o u s membranes. The only difference is that the permselectivity of h y d r o g e n in relation to the other c o m p o n e n t s is ten times higher. The p e r m e a t i o n of h y d r o g e n however, remains the same [29,31]. In Table 14.7 the p e r m e a t i o n s and permselectivities are given for the chosen m e m b r a n e types. These data have been used to perform the modelling. In this investigation Pe = 0.5 is taken as a s t a n d a r d for calculations (see Section 14.2.3). No further optimisation t o w a r d s m e m b r a n e surface area has been carried out.
14.3.2.4 Results With Pe = 0.5, it has been calculated that u n d e r the chosen conditions in all configurations of the reactor section a m e m b r a n e surface area of a p p r o x i m a t e l y 43,000 m 2 is required for m i c r o p o r o u s and p a l l a d i u m m e m b r a n e s and 3,300 m 2 for K n u d s e n diffusion membranes.
Sub-configuration (A) Results of the i m p l e m e n t a t i o n of all four types of m e m b r a n e s in only the first reactor (PBMR) are given in Table 14.8. I m p l e m e n t a t i o n of these m e m b r a n e s decreases the performance of the reactor because: a part of the steam that is used for dilution and energy carrier in the conventional m o d e l is n o w used as sweep gas; less dilution and e n e r g y i n p u t has a negative effect on the d e h y d r o g e n a t i o n ; in this early stage not m u c h h y d r o g e n has been f o r m e d that can be t r a n s p o r t e d t h r o u g h the membrane; and - the high partial pressure of ethylbenzene will enhance the p e r m e a t i o n of this reactant t h r o u g h the m e m b r a n e w i t h the consequence that there is less ethylbenzene left to react to styrene. -
-
662
14 - - A P P L I C A T I O N OF P O R O U S I N O R G A N I C GAS S E P A R A T I O N M E M B R A N E S
TABLE 14.8
Results of the simulations with sub-configuration (A) Membrane
Implementation of membranes in the first reactor Yield (%)
Conversion (%)
Selectivity (%)
N o membrane
43
46
95
Knudsen Microporous Palladium Hypothetical
38
40
94
40
42
95
40
42
95
40
43
95
Yield of styrene 3.67
=,,
,,,.
,,,,
.,,
..
..
..,
.=
=,,,
,,=
,4.. 0.65 >-
0.63 0.1
0,3
0.5
0.7
0.9
1.1
P e r m e a t e p r e s s u r e (bar]
',,-~Convent~nal
reactor ~ - Memloran~reactor,
}
Selectivity to styrene 0.93.=,
>,
0.93
..,= ~ O,925
0.92
,.
.
.
_--
.
.
~
.
.
~
.
.
.
,-
.
.
.,
.
.
~
.
.
.
-.
.
.
~
~
I 0.1
0.3
0.5
0.7
0,9
,1
P e r m e a t e pressure (bar] ': - ~ - Conventzonal reactor ~
Mernl0ranereactor
']
Fig. 14.11. Yield and selectivity as a function of the permeate pressure. For m e m b r a n e s i m p l e m e n t e d in the second reactor only the results of microporous m e m b r a n e s will be discussed in detail, because p a l l a d i u m m e m b r a n e s gave almost the same results and the performance is better than that of K n u d s e n diffusion membranes. The yield and selectivities have been calculated at different permeate pressures and are plotted in Fig. 14.11. The results for the conventional reactor are obtained w i t h o u t a m e m b r a n e i m p l e m e n t e d in the process.
14 B APPLICATION OF POROUS I N O R G A N I C GAS SEPARATION M E M B R A N E S
663
With decreasing permeate pressures, the yield increases to the same level as in the conventional reactor and the selectivity increases to a higher level. In this stage of the process the reaction approaches the equilibrium closer than in the first reactor, so the negative effect of permeating ethylbenzene is less and the positive effect of hydrogen permeation is larger. The increase in yield, with decreasing permeate pressure is due to the suppression of the hydrogenation reaction in which styrene reacts to ethylbenzene. The increase in selectivity to styrene is explained by the suppression of the side reactions to, e.g., toluene and benzene, due to the lower partial pressure of hydrogen. Thus, implementation of membranes leads to the expected effects, but these are too small to compensate the negative effect of less steam in the reactant gas (partly used in this case as sweep gas). The results from simulations with a membrane unit placed between the two conventional reactors are comparable to the above-mentioned results, because the average partial pressures are almost the same. The simulation results are not as promising as expected beforehand and not as good as those reported in literature [45-49]. To find an explanation for our results, we first simulated the implementation of membranes having a permselectivity which is thousand times higher than that for microporous membranes, but which has the same hydrogen permeability. To create an even more ideal environment for extremely selective removal of hydrogen from the reactant gas, the permeate pressure has been set to 0.005 bar. Although the hydrogen partial pressure in the reactant gas was negligible (meaning that the hydrogen transport through the membrane was as large as possible), the increase in yield and selectivity was no more than 2 percentage points. However, a large increase was expected under these conditions. From this it can be concluded that the application of membranes under these circumstances will not lead to an improvement in the performance of the process. In order to explain this, the influence of the kinetics of the main reaction on the performance of the membrane reactor has been studied, for microporous membranes implemented in the second reactor. The reaction rate of the main reaction is successively multiplied by a factor 2 and 10, and as a consequence the reaction equilibrium is reached much faster. Under these circumstances increases are found in both yield and selectivity for the conventional dehydrogenation reactor without membranes. The results of the calculations are presented in Table 14.9 in which the differences in yield and conversion are given in percentage pointswith respect to the conventional case. The higher yields and conversions for the PBMR compared to the conventional reactor are due to the fact that the conversion is no longer limited by the kinetics, as in the previous cases, but by the permeation of hydrogen. It can be concluded that, if the reaction rate of the dehydrogenation process is faster than in current processes, implementation of microporous membranes
664
14 -- APPLICATIONOF POROUS INORGANICGAS SEPARATIONMEMBRANES
TABLE 14.9 Yield and selectivity in a PBMR as a function of reaction kinetics, compared to those in a conventional reactor Reaction rate
Conventional
2 times faster
10 times faster
Yield
_+0 % point
+5 % point
+12 % point
Selectivity
+ 1 % point
+1.5 % point
+2 % point
will give a significant improvement in yield and selectivity. The ratio of permeation and reaction rate is very important when selecting membranes for this application. This is comparable to that reported for the cyclohexane dehydrogenation process [52]. 14.3.2.5 Discussion
The results found in this study are less promising then those reported in literature [45-49]. There are several reasons for this difference. In some publications experiments have been reported in which process conditions a n d / o r feed compositions have been used that are not realistic or feasible on an industrial scale but do have a large impact on the performance of the membrane reactor. Also, when results are reported from modelling this process, incorrect assumptions were sometimes made, e.g. side-reactions which have a large influence on the performance of this process have been neglected [47]. In other publications a very large heat input is taken, which leads to a more or less isothermal reactor, and as a consequence to higher conversions [45,46,48]. Even if implementation of membranes on technical basis is possible, one of the major problems of implementing ceramic membranes in this styrene production process is the enormous membrane surface area required, which does not appear economically viable. Other membrane reactor configurations with a higher surface area to volume ratio may reduce the required module volume and simplify the installation. The biggest problem, however, is that these large membrane surface areas lead to very high costs. We assumed that in about ten years installation of one square metre of microporous gas separation membranes will cost about as much as $ 5,700. In the case treated, a surface area of 43,000 m 2 of microporous membranes is required, which will cost a total of 250 million dollars. The possible gain is 1% in selectivity, which gives an annual profit of 4.5 million dollars. Even when the reaction rate of the main reaction is enhanced by a factor ten, the possible profit of 9 million dollars does not lead to an interesting ROI. It should be realized that extra costs for, e.g., process integration are not yet included.
14 - - A P P L I C A T I O N OF P O R O U S I N O R G A N I C GAS S E P A R A T I O N M E M B R A N E S
665
14.3.2.6 Conclusions
In the present concept of styrene dehydrogenation implementation of inorganic membranes is not feasible. Application of Knudsen diffusion membranes with a low permselectivity to hydrogen leads to a considerable permeation of ethylbenzene and thus, to lower yields. Microporous and palladium membranes give better results, but worse than a conventional case, because the conversion is limited by reaction kinetics. The ratio of permeation rate to reaction rate is very important in selecting membranes in a membrane reactor process in which equilibrium shift is foreseen. From the simulations with higher reaction rates it follows that implementation of ceramic membranes can lead to higher yields and selectivities. However, even under these conditions the profit from extra styrene yield does not compensate the costs of the membranes. For profitable implementation of inorganic membranes, a high-selective membrane with a higher permeability than the membranes now available is necessary, in combination with higher reaction rates. 14.3.3 Water-Gas Shift Membrane Reactor 14.3.3.1 Introduction
The water-gas shift (WGS) reaction is an important reaction in many commercial processes where hydrogen has to be generated or where CO must be converted. In the WGS reaction carbon monoxide together with steam is converted to carbon dioxide and hydrogen. The reaction is a reversible chemical reaction, usually assisted by a catalyst (see Eq. (14.8)). CO + H20 ~
CO 2 +
H2
AH = -41.1 k J / m o l
(14.8)
Hydrogen is a very important gas in many areas of industry [50,53]. Currently, hydrocarbons are the main source for large-scale production of hydrogen [42,54]. Most of the hydrogen for industrial purposes is produced from natural gas and oil. The contribution of coal is limited at present but is expected to grow considerably in the future. In addition, biomass is expected to be a growing hydrogen source in the future. All these processes a n d / o r feedstocks produce a gas mixture containing mainly hydrogen and carbon monoxide (syn gas). The hydrogen and carbon monoxide concentration will vary depending on the feedstock and process. The WGS process is being used to adjust the syn gas compositions for further syntheses or to enhance the hydrogen content for hydrogen production. Coal particularly is thought to play a major role in future world energy supplies and possibly also as a feedstock supplier for the chemical industry [54].
666
14 - - A P P L I C A T I O N OF POROUS I N O R G A N I C GAS S E P A R A T I O N M E M B R A N E S
Through integrated coal gasification combined cycle (IGCC) power plants, coal can be converted in a clean and efficient way into electricity with syn gas as an intermediate product. In future IGCC options, syn gas can be partially converted into secondary gaseous products and be partially used to generate power. Hydrogen is an obvious secondary product for such a system [55]. A possible problem in future energy generation from coal is the emission of large quantities of CO2. The rapidly increasing concentration of greenhouse gases in the atmosphere has already triggered the development of clean coal technologies for power generation worldwide. In the long term even the introduction of the highly efficient IGCC systems might not be sufficient to ensure the use of coal for power production, and further measures to decrease the emission of greenhouse gases, CO2 in particular, might be necessary. The options to do so in an IGCC system also compare favourably with other large-scale coal-based combined cycles. The generation of hydrogen (WGS process) also plays an important role here [56,57]. The attainable conversion with the WGS reaction depends on how the chemical equilibrium is set. The equilibrium constant Kp decreases as the temperature increases. This implies that the CO conversion decreases with increasing temperature. In many cases a high hydrogen yield is the objective of the WGS application. Increased hydrogen yield and reduced carbon monoxide content can be obtained in several ways. In principle it is desirable to carry out the reaction at low temperatures. This can be achieved by: (1) cooling during the reaction by heat transfer, or inert gas addition; 2) intercooling through execution of the reaction in several steps (reactors). Other methods to increase the level of conversion are: (3) increasing the steam to carbon monoxide ratio; (4) forcing equilibrium displacement to the product side; this should be achieved by continuous removal of either hydrogen or carbon dioxide directly at the place where it is formed. Generally, in a conventional WGS system a two-step shift is used to obtain high CO conversion rates. In the first high-temperature shift reactor the major part of the CO is converted at high activity, whereas in the second shift reactor the rest of the CO (closely up to the thermodynamic equilibrium) is converted at low temperature and also low activity. Steam to carbon monoxide ratios above the stoichiometric ratio (higher than 2) are generally being used to attain the desired carbon monoxide conversion, but also to suppress carbon formation on certain catalysts. Different types of catalysts exhibit shift activity. Commercially available high temperature (HT) FeCr and low temperature (LT) CuZn are generally employed. Among some interesting new developments is the Co/Mo based catalyst. This type is completely insensitive to sulphur and certain formulations are claimed to possess good activity at both high and low temperatures. The
14 - - APPLICATION OF POROUS INORGANIC GAS SEPARATION MEMBRANES
667
steam/carbon ratio is set by equilibrium considerations and carbon formation suppression. Of the methods to increase the CO conversion mentioned, the first three possibilities are accompanied by severe penalties with respect to energy use, exploitation and investment costs. The fourth possibility seems less affected by such drawbacks and is therefore preferred [57]. Inorganic gas separation membranes with their unique properties can be used to selectively remove hydrogen in a membrane reactor. A schematic of the combination of membranes and the WGS reaction has already been shown in Fig. 14.2 (Section 14.2.2). With such a PMBR reactor it is possible to enhance the CO conversion of the reaction and concurrently separate hydrogen from the reaction mixture, and furthermore have a separate CO2 rich stream. The membrane reactor replaces two unit operations, has an enhanced hydrogen yield and will save steam, and therefore has the potential of energy efficiency improvement. The hydrogen produced can either be sold as an end-product or consumed directly as feed stock in down-stream hydrogen consuming processes, in e.g. the petrochemical industry. A first step to explore the potential of a WGS inorganic membrane reactor is to assess its technical and economic feasibility. The potential and exact lay-out of such a reactor is thought to be strongly dependent on the upstream raw gas production and gas treatment processes and the respective downstream processes. Therefore the techno-economic feasibility of the application can only be judged after a detailed investigation of the performance of the reactor against the background of the specific detailed characteristics of the process chains envisaged. Some typical applications of the water-gas shift membrane reactor that are currently being foreseen have been very briefly mentioned in the Introduction. One specific application will be dealt with in greater detail to illustrate the assessment of the feasibility of the reactor system.
14.3.3.2 WGS membrane reactor for C02 emission control The potential of the WGS membrane reactor in CO2 control in IGCC installations has been studied in greater detail [57]. The possibilities of the reactor and demands set for the membranes have been determined by carefully assessing the process integration options, by experimental membrane characterisation and by using a membrane reactor model.
Process integration Various possible process flow schemes have been proposed. The C O 2 r e moval generally takes place from the coal gas [56]. Conventional approaches generally consist of a separate multistage water-gas shift (WGS) conversion of the fuel gas, followed by a low temperature CO2 removal process. Hydrogen is
668
14 - - APPLICATION OF POROUS INORGANIC GAS SEPARATION MEMBRANES
l
Steam
IVlain c x ) m l o o r e ~ s :
Coal
P '
k Steam turbine
................................................................
, I~
Air
.I' ..............
y
Fig. 14.12. L a y o u t of an IGCC with CO2 control using a WGS m e m b r a n e reactor.
the only fuel component left after the WGS conversion and is fed to the gas turbine to convert to water only. However, the conversion of the WGS reaction is limited by its chemical equilibrium and the low temperature CO2 removal makes an additional cooling step necessary. Application of the membrane reactor can enhance the equilibrium production of H 2 from fuel gas and establish a separation between H 2 and CO2 at an elevated temperature. In Fig. 14.12 the layout of an IGCC with CO2 emission control is shown. The layout is similar to an ordinary IGCC except that the gas leaving the gas cleaning section is now fed to the membrane reactor. Nitrogen from the air separation unit is available as sweep gas for the reactor. Unconverted CO and unseparated hydrogen are catalytically burnt and the heat is utilized in the steam turbine. In this scheme CO2 is recovered separately from the other components and is ready available for e.g. disposal or re-use. The availability of sweep gas (02 blown gasifier), the low requirements of the purity of the product streams, as well as the high pressure of the feed gas leaving the gasifier makes this process very favourable for the application of membranes as a separation technique. Hydrogen from the membrane reactor is converted in a gas turbine with a high efficiency. The process efficiency will increase when the hydrogen production (CO conversion) and recovery (on the permeate side) from the membrane reactor is raised. CO2 abatement increases with increasing recovery of carbon components on the retentate side of the membrane. The performance of the reactor can be measured in terms of these three parameters. The boundary conditions for the membrane reactor in the total system depends upon final
14 - - APPLICATION OF POROUS INORGANIC GAS SEPARATION MEMBRANES
669
performance of, amongst others, the membrane reactor itself. For an initial insight, it is desirable to carry out sensitivity analyses with the membrane reactor. For this purpose the boundary conditions around the membrane reactor will be estimated in the first instance. Ceramic membranes Inorganic membrane development is still in progress [57] (see also Section 14.2.2). Microporous silica membranes have been developed at several universities and research institutes. Membrane selectivities of 15 and 20 for the separation of H 2 f r o m C O 2 have been reported. Even higher selectivities for H 2 arid CO, C H 4 and N 2 have been measured [20,57]. Most measurements reported in the literature have been performed on a laboratory scale. However, it has been shown that it is possible to upscale these microporous ceramic membranes to, at least, bench scale [31,57]. With other membranes such as noble (Pd) metal membranes and dense ceramic membranes very high and almost infinite selectivities for hydrogen are possible [58]. The permeation of these membranes is generally smaller than the permeation of microporous membranes. Microporous carbon membranes have been developed [59] but their possibilities in high temperature hydrogen separation are still unclear, although it is believed that there are opportunities. Scaling-up of these membranes seems possible from a technical point of view. All these membrane types are potentially suitable for application in the WGS membrane reactor concept, provided their endurance is sufficient. Results and discussion The initial parameters used for the membrane reactor sensitivity analysis are shown in Table 14.10. These parameters are a first guess of the boundary conditions of the total process. The conversion in the reactor is plotted in Fig. 14.13 against the Da number which can be regarded as a dimensionless residence time. From this plot it follows that the conversion in the membrane reactor equipped with high selective membranes can exceed the values possible with an ordinary plug flow reactor. From the graph it is clear that the conversion increases with increasing TABLE 14.10 Initial s i m u l a t i o n p a r a m e t e r s
Tf pf
= 623 K = 36 b a r
pP
= 21 b a r
Y
= 1.035
H20/CO
= 1.5 (after s t e a m injection)
670
14 - - APPLICATION OF POROUS I N O R G A N I C GAS SEPARATION MEMBRANES
1 O 0
"~
--= ........
=---: .......................
9. . . . . .
~. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
~......................
.:. . . . . . . . . . . . . . . . . .
95
i_...a
~9
90
~
85 80
tt
S ~ = 15
0
5
10
15
20
Da
~~Plug
flow reactor ~ M e m b r a n e
reactor[
Fig. 14.13. Conversion vs. Da. ,..., 1oo ~- 95 ~ 90 ~ 0 o
85 80
~
75
~
70
"~
65 60 55 50, 0.1
D a - 12 S i - 15 0.3
0.5
0.7
0.9
1.1
Pe
{~CO
conversion ~
H2 recovery -=a,-C recovery I
Fig. 14.14. Conversion and recovery vs. Pe. Da as w o u l d be expected. Conversion in an o r d i n a r y reactor reaches a certain m a x i m u m d u e to the establishment of the chemical equilibrium. In a m e m b r a n e reactor conversion keeps increasing as a result of continuous h y d r o g e n permeation. This is an i m p o r t a n t aspect of the m e m b r a n e reactor, because the steam excess can be r e d u c e d which leads to a favourable e c o n o m y of the process. The influence of the Peclet n u m b e r is s h o w n in Fig. 14.14. Pe is reciprocally p r o p o r t i o n a l to the m e m b r a n e surface. Decreasing the Pe n u m b e r increases h y d r o g e n recovery and as a consequence the CO conversion. W h e n m o r e m e m b r a n e surface is available, also more carbon dioxide and carbon m o n o x i d e p e r m e a t e s t h r o u g h the m e m b r a n e and the carbon recovery decreases. A n i m p o r t a n t question for the application of m e m b r a n e s is w h a t the desired selectivity for the m e m b r a n e s has to be. The influence of the m e m b r a n e selec-
14 - - A P P L I C A T I O N
OF POROUS INORGANIC
GAS SEPARATION
671
MEMBRANES
,..., 100 ......a
90
o
80
~
70
.e
60
Da = ! 2 Pe =0.4
/ /
~" 50
! i i
4o 0
10
20
30
40
50
Si l. ~" CO conversion, ~
H2 recovery ~
C recovery t
Fig. 14.15. Conversion, recovery vs. selectivity 9 tivity on the performance of the reactor is s h o w n in Fig. 14.15. CO conversion a n d C recovery increase w i t h rising selectivity. The h y d r o g e n recovery is m o r e or less constant as a result of the Pe n u m b e r that is held constant for the various simulations. CO conversion rises slightly due to the decreased losses of reactants w i t h higher selectivity. Carbon recovery increases strongly with rising selectivity. T h o u g h the performance of the reactor increases w i t h rising selectivity, the incremental increase in conversion and especially the C-recovery decreases with g r o w i n g selectivity. The gain in performance is marginal w h e n the selectivity surpasses 40. The current m e a s u r e d H 2/CO2 permselectivity of scaled up m e m b r a n e s is 15 [57]. With this value a good recovery of both H 2 and CO2 is possible as appears from Fig. 14.15. For a p o w e r plant including a m e m b r a n e reactor with m e m b r a n e s w i t h a selectivity of 15 the efficiency of the total system has been d e t e r m i n e d t h r o u g h flow sheet calculations. In these calculations the requirements and the d e m a n d s of the m e m b r a n e reactor and the rest of the system m u s t match, so one or m o r e iterative calculations is necessary to optimise the total system. The results of the calculations after optimisation are presented in Table 14.11 in w h i c h three TABLE 14.11 Results of power plant efficiency calculations Process
Process efficiency (%)
IGCC with membrane reactor for CO2 removal IGCC with conventional CO2 removal IGCC without CO2 removal
42.8 40.5 46.7
672
14 - - A P P L I C A T I O N OF POROUS I N O R G A N I C GAS S E P A R A T I O N M E M B R A N E S
systems are compared. Obviously a system without CO2 removal exhibits the highest efficiency. However, the efficiency penalty induced is significantly lower when CO2 is removed using the membrane reactor system instead of removal with a conventional technique. The conventional technique used here is a wet scrubbing process in which CO2 is absorbed on a physical sorbent. J
_
14.3.3.4 Full-scale process considerations
When the membranes are used on an industrial scale, a considerable amount of surface area will be necessary to process the gas stream involved. A typical surface area necessary is 1500 m 2 for a 300 MWe class power plant. For ceramic membranes this is a rather large surface area. Considering that permselectivity is already good for this application, it seems reasonable to direct research towards enlargement of the permeation or explore module concepts with a high surface area to volume ratio (e.g. monolytic systems) next to selectivity improvement. When membranes are produced in a tubular geometry, which seems the most feasible currently, all membranes have to be sealed separately. This favours tubes with large diameters to reduce the number of seals. On the other hand, the smaller the tube diameter, the higher the specific surface area attainable in a module. High pressure, high temperature membrane sealing is an important aspect of the full scale module and this hurdle has been taken for laboratory and bench scale [16,28,31,57]. The membranes can be sealed gas-tight to a stainless steel tube by a special joining technique. Experiments will be carried out initially for the so-called passive reactor concept in which a high selective membrane is surrounded by catalyst. Dead end tube configuration, in which only one end of the membrane tube is connected and the other end is closed [14], seems favourable since it needs one ceramic to metal joint less than two-side connected tubes. A drawback of this option is the large force that will act upon the dead end side of the membrane when the process works with a considerable pressure drop as in this application. These aspects show that it is important to realise for which application the membranes are being developed and to consider scaling up in an early stage. 14.3.3.5 Conclusion
Through membrane reactor model calculations it has been shown that membranes can enhance the conversion of a WGS membrane reactor and concurrently separate hydrogen from carbon dioxide. This system can be used to control the release of CO2 to the atmosphere from a IGCC power plant. Through process
14 w A P P L I C A T I O N OF P O R O U S I N O R G A N I C GAS S E P A R A T I O N M E M B R A N E S
673
flow sheet calculations it has been shown that the efficiency of CO2 control using the membrane reactor is significantly higher than when a conventional technique (i.e. wet washing with a sorbent) is applied. When selectivity of the membranes can be increased, it does not seem to be necessary to surpass approximately 40 for the process under consideration, because the gain in reactor performance seems marginal. Enlargement of the permeation is an important aspect on the other hand, so that the total surface area necessary for the full scale application can be reduced. This example shows that knowledge of the demands and requirements of the application are also very important in the development of membrane material.
14.4 CONCLUSIONS Three examples of the use and feasibility of inorganic membranes in reactor applications have been discussed. Although several references give a very positive indication on the technical possibilities of the use of inorganic membranes in reactor applications, it has been shown that measurements under realistic conditions and calculations involving the complete process can show the opposite. A multidisciplinary approach is needed to study the feasibility of inorganic membranes in (membrane reactor) applications. A combination of membrane and materials know-how and an insight into application opportunities and process economics is necessary to discuss the techno-economic feasibility of inorganic membranes. Furthermore, measurements and calculations should be performed on a realistic basis and scale in order to obtain reliable data on the performance of the membranes. As yet, insufficient realistic data on longterm membrane stability and coke formation on the membranes are available. In general the membrane reactor examples show that knowledge of the demands and requirements of each foreseen membrane application is very important in the choice and development of the membrane materials. Application of ceramic membranes can improve the return on investment in the propane dehydrogenation process. Probably the only possibility for a technically and economically feasible propane dehydrogenation process, able to enhance the ROI enough to make the investment worthwhile, is the combination of a high driving force (sweep gas or low permeate pressure) and a very high selective membrane. The isothermal reactor concept shows better prospects than the adiabatic concept. At a price difference smaller than 300 $/tonne between propylene and propane the propane dehydrogenation process based upon membranes will hardly be economically viable. The present concept of implementation of ceramic membranes in the styrene process is not feasible, because:
674
14 - - A P P L I C A T I O N OF P O R O U S I N O R G A N I C GAS S E P A R A T I O N M E M B R A N E S
- the permselectivity to hydrogen of Knudsen membranes is too low, which leads to a significant loss of ethylbenzene from the reaction side which can no longer take part in the reaction; - the lower permeation of high selective microporous and palladium membranes leads to very high surface areas, which makes the application unattractive; - the reaction rate of the dehydrogenation of ethylbenzene to styrene is to low: the kinetics and not the permeation are the limiting step in the membrane reactor. From simulations with higher reaction rates it has been shown that implementation of ceramic membranes can lead to higher yields. However, even under these conditions the profit from the extra styrene yield does not compensate for the costs of the membranes. For profitable implementation of inorganic membranes, a high-selective membrane with a higher permeability than the membranes now available is necessary, in combination with higher reaction rates. It has been shown that membranes can enhance the conversion of a watergas shift membrane reactor and concurrently separate hydrogen from carbon dioxide. The efficiency of CO2 control using the membrane reactor with a H 2 / C O 2 selectivity of 15 is significantly higher compared to a conventional technique (i.e. wet washing with a sorbent). It is not necessary to exceed a selectivity of approximately 40 for H 2 / C O 2 for the process under consideration, because further increase in reactor performance seems marginal. Enlargement of the permeation is an important aspect on the other hand, so that the total surface area necessary for the full-scale application can be reduced. In all three applications discussed, the stability of the membranes in these high temperature processes and the design of suitable modules still needs much research and development.
Acknowledgements The research work described here has been funded in part by the Dutch Organization for Energy and Environment (NOVEM), the Dutch Ministry of Economic Affairs (EZ), and the Commission of the European Union. Kinetics Technology International BV is gratefully acknowledged for their helpful discussions and calculations. The authors would also like to express their thanks to Prof. R. Pruschek, Dr. G. Oeljeklaus and R. Kloster of the University of Essen, G. Haupt of Siemens AG Power Generation (KWU), Dr. H. van den Berg of Dow Benelux BV, and Dr. L. van der Ham of the University of Twente. K. Hemmes, G. Leendertse and E. Delnoij are gratefully thanked for their help in modelling and setting up the membrane models.
14 - - APPLICATION OF POROUS I N O R G A N I C GAS SEPARATION MEMBRANES
List of Symbols and Abbreviations
A Da F k
G
P Pe
Q R Si T V Y
membrane surface area (m 2) Damk6hler number flow rate (mol/s) reaction rate coefficient (mol/m3) 1-~ (l/s) equilibrium constant pressure (Pa) Peclet number permeation (mol/msPa) gas constant (J/molK) permselectivity (ratio H2/component i permeation) temperature (K) volume reactor (m 3) molar ratio sweep flow vs. feed flow sum of powers in power law expression ratio of permeate and feed side pressure
Superscripts
? m P s
Subscripts i
tot
feed membrane permeate sweep
component i total
Abbreviations
CMR CMRL CMRH FBCMR FBMR HT IGCC LHSV LT PBCMR PBMR ROI WGS
catalytic membrane reactor catalytic membrane reactor low conversion catalytic membrane reactor high conversion fluidized bed catalytic membrane reactor fluidized bed membrane reactor high temperature integrated coal gasification combined cycle liquid hourly space velocity (h-1) low temperature packed bed catalytic membrane reactor packed bed membrane reactor return on investment water-gas shift
675
676
14 - - A P P L I C A T I O N OF P O R O U S I N O R G A N I C GAS S E P A R A T I O N M E M B R A N E S
APPENDIX
The p e r m e a t i o n is defined as the (pure gas) flow ( m o l / s ) t h r o u g h the m e m brane per surface area and per pressure difference over the m e m b r a n e . The permselectivity is defined as the ratio of the p e r m e a t i o n of p u r e gases. The separation factor is defined as" y 1-y
1- x x
(14.9)
in w h i c h y = concentration fastest p e r m e a t i n g c o m p o n e n t on p e r m e a t e side; x = concentration fastest p e r m e a t i n g c o m p o n e n t on feed side. The conversion a n d selectivity are given for the p r o p a n e d e h y d r o g e n a t i o n reaction. For the ethylbenzene d e h y d r o g e n a t i o n and w a t e r - g a s shift reaction the same definitions can be used for the respective r e a c t a n t s / p r o d u c t s . The conversion is defined as: conversion =
mass flow propane i n - mass flow propane out mass flow propane in (14.10)
The selectivity can be expressed on a molar basis or mass basis: select. (mol) =
tool flow propylene o u t - mol flow propylene in
select. (mass) =
mol flow propane i n - mol flow propane out (14.11) mass flow propylene o u t - mass flow propylene in mass flow propane i n - mass flow propane out (14.12)
The yield is defined as c o n v e r s i o n , selectivity, on mol or mass basis.
REFERENCES
1. 2. 3.
4.
5.
A. Crull, Prospects for the inorganic membrane business, in: A.J. Burggraaf, J. Charpin and L. Cot (Eds.) Proceedings ICIM2-91. Key Eng. Mater., 61/62 (1991) 279-288. Business Communications Co. Inc., Inorganic Membranes: Markets, Technologies, Players. BCC Report No. RGB-112R, Jan. 1994. H.E. Johnson and B.L. Schulman, Assessment of the Potential for Refinery Applications of Inorganic Membrane Technology ~ An Identification and Screening Analysis. DOE/FE61680-H3, Final report, May 1993. H.P. Hsieh, Inorganic Membranes. Am. Inst. Chem. Eng. Symposium Series, No. 261, Vol. 84, 1989, pp. 1-18. Ceramic Membranes ~ New Horizon in Membrane Separation. Emerging Technologies no.
14 - - A P P L I C A T I O N OF P O R O U S I N O R G A N I C G A S S E P A R A T I O N M E M B R A N E S
677
36, Technical Insights, Inc., 1989. 6. B.Z. Egan, Using Inorganic Membranes to Separate Gases: R&D Status Review. Report ORNL/TM-11345, Nov. 1989. 7. A.J. Burggraaf, Key points in understanding and development of ceramic membranes, in: Y.H. Ma (Ed.), Proceedings of the 4th International Conference on Inorganic Membranes, July 10-14, Worcester, MA, USA, 1994, pp. 1-16. 8. A.J. Burggraaf, K. Keizer, R.S.A. de Lange, Z. Vroon and V.T. Zaspalis, Ceramic membranes for separation and reactions, in: R. Ballmoos et al. (Eds.), Proceedings of the 9th International Conference on Zeolites, Montreal, 1993, pp. 47-70. 9. J.N. Armor, Catalysis with permselective inorganic membranes. Appl. Catal., 49 (1989) 1-25. 10. H.P. Hsieh, Inorganic Membrane Reactors - A Review. Am. Inst. Chem. Eng. Symposium Series, No. 268, Vol.85, 1989, pp. 53-67. 11. H.P. Hsieh, Inorganic membrane reactors. Catal. Rev. Sci. Eng., 33 (1991) 1-70. 12. J. Shu, B.P.A. Grandjean, A. van Neste and S. Kaliaguine, Catalytic palladium-based membrane reactors: a review. Can. J. Chem. Eng., 69, (1991) 1036. 13. K.R. Westerterp, Multifunctional reactors. Chem. Eng. Sci., 47 (1992) 2195. 14. G. Saracco and V. Specchia, Catalytic inorganic membrane reactors: present experiences and future opportunities. Catal. Rev. Sci. Eng., 36 (1994) 305. 15. J. Zaman and A. Chakma, Review: inorganic membrane reactors. J. Membr. Sci., 92 (1994) 1. 16. G. Saracco, G.F. Versteeg and W.P.M. van Swaaij, Current Hurdles to the Success of High-Temperature Membrane Reactors. J. Membr. Sci., 95 (1994) 105. 17. H.M. van Veen, J.P.B.M. Tol, C.W.R. Engelen and H.J. Veringa, High temperature gas separation with alumina membranes, in: A.J. Burggraaf, J. Charpin and L. Cot (Eds.), Key Eng. Mat., 61/62 (1991) pp. 593-598. 18. Y. Shindo, K. Obata, T. Hakuta, H. Yoshitome, N. Todo, and J. Kato, Permeation of hydrogen through a porous vycor glass membrane. Adv. Hydrogen Energy, 2 (1981) 325. 19. A.S. Damle and S.K. Gangwal, Catalytic carbon membranes for hydrogen production, in: Proceedings of the lOth Annual Gasification and Gas Stream Cleanup Systems Contractors Review Meeting, Morgantown, 1990, pp. 322-329. 20. R.S.A de Lange, Microporous Sol-Gel Derived Ceramic Membranes for gas Separation Synthesis, Gas Transport and Separation Properties. Thesis, Twente University, 1993. 21. S. Kitao, H. Kameda, and M. Asaeda, Gas Separation by thin porous silica membrane of ultra fine pores at high temperature. Membrane (Maku), 15 (1990) 222. 22. D.L. Roberts, I.C. Abraham, Y. Blum and J.D. Way, Gas Separations Using Ceramic Membranes m Final Report. Report: DOE/MC/25204-3133, May 1992. 23. (a) J.E. Koresh and A. Softer, Molecular sieve carbon permselective membrane. Part I: Presentation of a new device for gas mixture separation. Sep. Sci. Technol., 18 (1983) 723. (b) J.E. Koresh and A. Softer, The carbon molecular sieve membranes. General properties and the permeability of CH4/H2 mixture. Sep. Sci. Technol., 22 (1987) 973. 24. M.D. Jia, K.V. Peinemann and R.D. Behling, Ceramic Zeolite Composite Membranes. Preparation, Characterization and gas Permeation. J. Membrane Sci., 82 (1993) 15. 25. S. Kitao and M. Asaeda, Gas separation performance of thin porous silica membrane prepared by sol-gel and cvd methods, in: A.J. Burggraaf, J. Charpin and L. Cot (Eds.), Key Eng. Mat., 61/62 (1991) 267-272. 26. Johnson Matthey Technology Centre, Thin Supported Pd-alloy Membranes for Hydrogen Purification. Report: ETSU F/02 / 00034 / REP, 1995.
678
14-- APPLICATIONOF POROUSINORGANICGASSEPARATIONMEMBRANES
27. M. Konno, M. Shindo, S. Sugawara and S. Saito, A composite palladium and porous aluminium oxide membrane for hydrogen separation. J. Membr. Sci., 37 (1988) 193. 28. (a) F.M. Velterop and K. Keizer, Development of high Temperature Resistant Module for Ceramic Membranes, in A.J. Burggraaf, J. Charpin and L. Cot (Eds.), Key Eng. Mat., 61/62 (1991) 391-394. (b) F.M. Velterop, Ceramic membranes m technological implementation at high temperatures. Chem. Plants Process., Nov. (1993) 66. 29. M. Bracht, A. Bos, P.P.A.C. Pex, H.M. van Veen and P.T. Alderliesten, Water-gas shift membrane reactor, in: W.R. Bowen et al. (Eds.), Proceedings of Euromembrane "95, Bath. Vol. 2, 1995, pp. 425-430. 30. R.L. Goldsmith, Low-Cost Ceramic Membranes and Supports for Gas Separation. Report of Ceramem Corp, Waltham, USA. 31. P.P.A.C. Pex et al., Materials aspects of microporous inorganic gas separation membrane manufacturing, in: W.R. Bowen et al. (Eds.), Proceedings of Euromembrane "95, Bath, Vol. 1, 1995, pp. 295-300. 32. Z.D. Ziaka, R.G. Minet and T.T. Tsotsis, A high temperature catalytic membrane reactor for propane dehydrogenation. ]. Membr. Sci., 77 (1993) 221. 33. Technical and economic evaluation study for the use of ceramic membrane reactors for the dehydrogenation of propane to propylene. Confidential NOVEM Report No. 33105/0090, by KTI, ECN and HIC, Jan. 1994. 34. (a) P.R. Cottrell et al., UOP Technology Conference, 1992. (b) G.M. Horneday et al., Advances in Petroleum Chemistry and Refining, Vol. 4, Ch. 10, 1961, 482 pp. 35. P.R. Sarthy et al., Hydrocarbon Process., 43 (2) 1993. 36. H. Bolt et al, Dehydrogenation process for propane and isobutane. Presented at the AIChE Spring Meeting, Houston, Texas, April 7-11, 1991. 37. R.D. Dunn et al., The Phillips steam active reforming (STAR) process for the dehydrogenation of C3, C4 and C5 paraffins. Presented at the 1992 NPRA meeting, New Orleans, LA, March 22-24, 1992. 38. H.M. van Veen, R.A. Terpstra, J.P.B.M. Tol and H.J. Veringa, Three layer ceramic alumina membranes for high temperature gas separation applications, in: Proceedings of the First International Conference on Inorganic Membranes, Montpellier, 1989, pp. 329-335. 39. G.R. Gallaher, T.E. Gerdes and P.K.T. Liu, Experimental evaluation of dehydrogenations using catalytic membrane processes. Sep. Sci. Technol., 28, (1993) 309. 40. Z.D. Ziaka, R.G. Minet and T.T. Tsotsis, Propane dehydrogenation in a packed-bed membrane reactor. AIChE J., 39 (1993) 526. 41. P. Trambouze, Energy saving in the petrochemical industry, in: P.A. Pilavachi (Ed.), Energy Efficiency in Process Technology. Elsever Applied Science, London, 1993, pp. 61-72. 42. A. Chauvel and G. Lefebvre, Petrochemical Processes 1. IFP Publications, Paris, 1989, pp. 361-372. 43. J.G.P. Sheel and C.M. Crowe, Simulation and optimization of an existing ethylbenzene dehydrogenation reactor. Can. J. Chem. Eng., 47 (1969) 183. 44. J.C.S. Wu and P.K.T. Liu, Mathematical analysis on catalytic dehydrogenation of ethylbenzene using ceramic membranes. Ind. Eng. Chem. Res., 31 (1992) 322. 45. B.K. Abdalla and S.S.E.H. Elnashaie, A membrane reactor for the production of styrene from ethylbenzene. J. Membr. Sci., 85 (1993) 229. 46. B.K. Abdalla and S.S.E.H. Elnashaie, Catalytic dehydrogenation of ethylbenzene to
14 w APPLICATIONOF POROUSINORGANICGASSEPARATIONMEMBRANES
679
styrene in membrane reactors. AIChE J., 40 (1994) 2055. 47. Y.L. Becker, A.G. Dixon, W.R. Moser and Y.H. Ma, Modelling of ethylbenzene dehydrogenation in a catalytic membrane reactor. J. Membr. Sci., 77 (1993) 233. 48. F. Tiscareno-Lechuga and C.G. Hill Jr., Experimental studies of the non-oxidative dehydrogenation using a membrane reactor. Appl. Cat. A: Gen., 96 (1993) 33. 49. J.C.S. Wu et al., Dehydrogenation of ethylbenzene to styrene using commercial ceramic membranes as reactors. Sep. Sci. Technol., 25 (1990) 1489. 50. Ullmans Encyclopedia of Industrial Chemistry, 5th edn., 1993. 51. G.F. Froment and K.B. Bischoff, Chemical Reactor Analysis and Design. Elsevier, New York, 1990, pp. 501-508. 52. M.P. Harold, C. Lee, A.J. Burggraaf, K. Keizer, V.T. Zaspalis, and R.S.A. de Lange, Catalysis with inorganic membranes. MRS Bulle., April (1994) 34-39. 53. Petrochemical Handbook 1993. Hydrocarbon Process., 72 (1993) 162. 54. M. Steinberg, H.C. Cheng, Modern and Perspective Technologies for hydrogen production from fossil fuels. Int. J. Hydrogen Energy, 14, (1989) 797. 55. G.A. Mills, Coproduction of Hydrogen and Electricity: Catalytic Applications. ACS Div. Fuel Chem., 39 (1994) 1162. 56. P. Riemer, Carbon Dioxide Capture from Power Stations. IEA Greenhouse Gas R&D Programme, 1993. 57. An Attractive Option for C02 Control in IGCC Systems: Water-gas Shift with Integrated Hydrogen~carbon Dioxide Separation (WIHYS) Process. Final Report, CEC project number JOU2 CT92-0158. P.T. Alderliesten and M. Bracht (eds.). Partners: Netherlands Energy Research Foundation, Univ. of Essen (Germany), Univ. of Limerick (Ireland), National Technical Univ. of Athens (Greece) and Siemens AG (Germany), 1996. 58. H.M. van Veen, Applicatiestudie keramische membranen voor g a s s c h e i d i n g - Onderzoek en ontwikkelingen op het gebied van hoge temperatuur gasscheiding met anorganische membranen: status tot november 1993 (Application Study Ceramic Membranes for gas S e p a r a t i o n - R&D Aimed at high Temperature gas Separation with Inorganic Membranes - - State of the art to November 1993, in Dutch), Confidential ECN report, Report number ECN-CX-94-056, June 1994. 59. Microporous Carbon Membranes for Gas Separations. Brite Euram Project BREU-CT920568. Partners: BP Research and Eng. Centre (England), NRC Democritos (Greece), SCT (France) and Imperial College (England). 60. R.L. Keiski, T. Salmi, V.J. Pohjola, Development and verification of a simulation model for a non isothermal water-gas shift reactor. Chem. Eng. J., 48 (1992) 17. 61. C.R.F. Lund and L.A. Bernstein, Membrane reactors for catalytic series and series-parallel reactions. J. Membr. Sci., 77 (1993) 155. 62. G.R. Gavalas, Hydrogen Separation by Ceramic Membranes in Coal Gasification. Final report, DOE/MC/26365-3423, 1993. 63. E. Kikuchi, S Uemiya, N. Sato, H. Inoue, H. Ando and T. Matsuda, Membrane reactor using microporous glass-supported thin film of palladium; application to the water-gas shift reaction. Chem. Lett., 3 (1989) 489. 64. S. Uemiya, N. Sato, H. Ando and E. Kikuchi, The water-gas shift reaction assisted by a palladium membrane reactor. Ind. Eng. Chem. Res., 30 (1991) 585. 65. D.J. Edlund, W.A. Pledger, B.M. Johnson, D.T. Friesen, presented at the 5th Annual Meeting of the NAMS, Lexington, KE, 1992, paper 11F. 66. L. Lloyd, D.E. Riddler, M.V. Twigg, The water-gas shift reaction, in: M.V. Twigg (Ed.), Catalyst Handbook. Wolfe Publishing, 1989, pp. 283-338.
680 67.
14m APPLICATIONOF POROUSINORGANICGASSEPARATIONMEMBRANES
R. Wurster and W. Zittel, Energy Technologies to Reduce CO2 Emissions in Europe: Prospects, Competition, Synergy. Conference proceedings, IEA-OECD, Paris, 1994. 68 J. E. Hogsett, W.H. Mazur, Estimate membrane system area. Hydrocarbon Proc., 62 (1983) 52. 69. A.S. Kovvali, S. Vemury, K.R. Krovvidi and A.A. Khan, Models and analyses of membrane gas permeators. ]. Membr. Sci., 73 (1992) 1. 70. Y. Shindo, T. Hakuta, H. Yoshitome, H. Inoue, Calculation methods for multicomponent gas separation by permeation. Sep. Sci. Technol., 20 (1985) 445. 71. R. Rautenbach and W. Dahm, Simplified calculation of gas-permeation hollow fiber modules for the separation of binary mixtures. J. Membr. Science, 26 (1986) 319.
Subject Index abrasive materials, 623 acetylacetone (acacH), 243 acidity, effects of, 304 activated gas permeation, 378 activation energy, 393 additives effect on microstructure, 290 effect on stress, 290 adsorption competitive, 381 multicomponent, 40 t-curve, 48 adsorption isotherms determination of, 43 types of, 36 ageing, 303, 308 agglomerate forming, 153 aggregates, 154 alkalinity, 315 y-alumina, 240 film, 321 membranes, 233, 289, 296-297 supporting membrane, 300 supports, 306, 309 amphiphilic systems, 245 amphoteric behaviour of metal oxides, 584 anodic alumina membranes, 539 apparent activation energy, 347, 385 applications, 641 architecture, 21 asymmetric structures, 23, 27 atomic force microscopy, 307 BaAl~2019, 235 back diffusion, 364 BaTiO3 membrane, 236
beer brewing, 627 bending strength, 136 BET isotherms, 40, BICUVOX, 470 bilge water, 621 binary silica-metaloxide, 305 binder polymer, 206 binders, 122, 203, 214 blocking electrodes, 491 boehmite, 240 Bosanquet equation, 358 Brownian agglomeration, 161 Brownian movement, 160 brownmillerite, 502 structure, 499 bubble point, 99, 134 burst pressure, 136 cake filtration, 154 calcia-stabilized zirconia, 465 capillary colloidal filtration, 183 capillary condensate permeation, 352 capillary condensation, 350, 373 capillary pressure, 272, 352 capillary suction, 351 carbon membranes, 312, 354, 546 carbon supports, 321 carbonate formation, 512 Carman-Kozeny, 187, 212, 575 catalysts, 666 cationic surfactant, 249 ceramic membranes applications of, 619 m commercially available, 620 ceramic nanofilters, 240, 596 ceramic paste preparation, 121 681
682 ceramics industry, 623 ceria-doped ZrO2--Y203, 475 CETS, 555, 556 CH3OH dehydrogenation, 540 CH4 reforming, 537 Chapman-Enskog relation, 359 characterisation, 67 charge density, 603 charge disproportionation, 484, 493 charge-pH diagram, 233 chemical diffusion coefficient, 451,492 chemical energy transmission systems (CETS), 555 chemical modification of the membrane surface, 633 chemical processes, recycling in, 626 chemical resistance, 33 chemical stability of y-alumina membranes, 234 chemical vapour deposition, 310, 375 chemically recuperated gas turbine (CRGT), 557 chemisorption, 36 clarification, 628 cluster-cluster model, 302 clusters, 230, 238 CMR, 540 configuration, 531 CNMR model, 550 CO2 removal, 558 coagulant, 629 coal gasification combined cycle, 666 coating flow dynamics, 190 coating thickness, 189, 190 coke formation, 654 colloidal filtration, 210 colloidal particles, 232 colloidal processing, 142, 163 colloidal ceramic processing, 150 colloidal sols, 232 colloidal stability, 210 colloidal suspensions, 229
SUBJECTINDEX commercially available membranes, 31 compact layer, 151 contact time, 151 withdrawal speed, 151 competitive adsorption, 381 competitive Langmuir adsorption, 387 compressor-condensate, 620 concentrated suspension, 173, 175 concentration factor, 623 concentration polarization, 570 condensation rate, 299 condensation reactions, 301 configurational diffusion coefficient, 388 connectivity, 72 contact angle, 196, 198 advancing, 198 receding, 198 contact time, 212 continuous stirred tank reactor, 504 convective flow, 603 conversion, 651, 670 correlation factor, 492 coupled membrane processes, 608 coupling effects, 355 cracking, 208, 279 cracking phenomena, 294 criteria for the selection of materials, 510 critical cracking thickness (CCT), 177 critical point, 272, 275 critical stress, 275,276, 280 critical thickness, 296 cross-flow microfiltration, 590 crystallisation, 315 CVD techniques, 538 Damk6hler number, 647 Darcy's law, 158 dead end pores, 335 deagglomeration, 214 Deborah number, 161 Debye length, 586 Debye-H~ckel screening length, 456
SUBJECTINDEX decomposition, 508 defects, 178-181 defect chemistry, 472 deflocculants, 123, 131 degreasing baths, 625 dehydrogenation, 550 of ethylbenzene, 643, 657 of propane, 643, 648 dense ceramic membranes, 435 dense inorganic membranes, 643 density, 70 density states, 488 depolarization, 591 detergents/surfactants retention, 626 dewetting, 155, 181, 190 diffusion coefficients, 359, 390 chemical, 451, 492 intrinsic, 384 tracer, 491 dip-coating, 183 Donnan effect, 603 Donnan exclusion, 588 drainage, 190 drying, 175 characteristics of membranes, 287 Constant rate period, 272 first falling rate period, 273 forces, 153 front, 274, 294 process, 271 rate, 274 second falling rate period, 274 stress, 276, 288 zone, 292 dual-phase composites, 470 dual-phase membrane, 438 Dubinin-Raduschkevich adsorption, 388 equation, 43, 52 dust, 181 Dusty Gas Model, 355, 359
683 edge-effects, 504 electrical double layer, 585 electro-osmosis, 594 electro-ultrafiltration, 610 electrokinematic flow, 588, 603 electrokinetic radius, 588 electroless plating, 538 electrolytic domain, 464 electronic conductivity, 492 electronic stoichiometry, 493 electrophoresis, 610 electrostatic interaction, 164 ellipsometry, 94 emulsion treatment centres, 621 emulsions, 621 enzymes, 632 equations of state, adsorption isotherms from, 41 erbia-stabilized bismuth oxide, 467 ethylbenzene, 657-658 extended defects, 495 extrusion, 119 facilitated transport, 608 FBMR configuration, 531 fermentation broths, 632 film coating, 189, 215, 262 film formation, 260 film thickness, 190 fish processing, 622 flat supports, 120 flow pulsations, 591 flow sheeting, 648 fluid flow measurements, 102 formed-in-place membranes, 580 fouling, 61, 575, 622 fractal concept, 238 fractal dimension, 299, 301 fractal geometry, 72 fracture, 280 fruit juices, 627
684 galvanic baths, 627 gas adsorption, 78 gas permeability, 103 gel structure, 154 Gibbs-Thomson equation, 273 Gouy layer, 586 grain boundaries, 508 grain boundary diffusivity, 508 gyration radius, 301 Henry constant, 385 Henry's law, 38, 346 high temperature NMR, 500 hollow fibres, 29 Horvath-Kawazoe equation, 55 hybrid installations, 11 hybrid membranes, 606 hydraulic permeability, 588 hydraulic pore radius RH, 246 hydraulic radius, 51 hydraulic resistance, 576 hydrodynamic model, 349 hydrodynamic of micro- and ultrafiltration systems, 590 hydrodynamics, 570 hydrogen recovery, 670 hydrolysis, 301 hydrostatic pressure difference, 351 image analysis, 77 immersion calorimetry, 84 immobilising an enzyme or yeast, 634 impurity phase, 467 initial layer formation, 260 ink, 624 ink-bottle pores, 50 inorganic membranes, 642 interaction forces, 162 interconnectivity, 26 intergrowth, 497 intergrowth structures, 511 intrinsic diffusion coefficient, 384
SUBJECTINDEX ionic conductivity, 454 ionic pre-exponential term, 496 isoelectric point, 594 isotopic exchange, 459, 468 Kelvin equation, 50, 350 kinetic demixing, 511 kinetics, 658, 663 Knudsen contribution, 342 Knudsen diffusion, 357 Knudsen equation, 338 Knudsen number, 337 Knudsen permeabilities, 343 Knudsen permeation of mixtures, 357 LaA111018, 235 lamellar systems, 248 Langmuir adsorption constant, 384 Langmuir isotherm, 36 lanthanum oxychloride porous thin film, 242 large micropores, permeation in, 387 late transition metal-containing perovskites, 492 layer thickness, 212, 215, 306 layered structures, 141 m dip-coating, 142 m porous, 142 substrate, 141 support structure, 141 suspensions, 142 withdrawal coating, 142 liquid adsorption, 61 liquid crystal phase, 249 liquid displacement techniques, 99 liquid permeability, 102 lubricants, 123 macromolecules, 156 macropores, 71 macroporous support, 119 manure, 623
SUBJECTINDEX market penetration, 10 market situation, 2 mass fractal dimension, 238 mass fractal, porosity of, 299 mass transport, 570 maximum packing, 171 Maxwell-Stefan equations, 386, 572 mean velocity, 338 membrane applications, prospects for, 12 membrane architecture, 335 membrane bioreactor, 608 membrane compaction, 578 membrane concepts, 436 membrane cut-off, 596 membrane fouling, 577 membrane geometry, 582 membrane reactors, 11, 633, 645, 658, 667, 673 applications, 642 membrane separation, 577 membrane thickness, 8 characteristic, 456 membrane transport, 572 membranes from RuO2-TiO2, 235 membranes, types of, 21 mercury porosimetry, 78 mesopores, 71 mesoporous alumina membranes, 539 mesoporous inorganic membranes, 643 mesoporous structures, 229 mesoporous textures, 248 metal alkoxides, 237 metal salts, 232 microdomains, 498, 501 micropore filling, 58 micropore size distribution, 53, 57 micropore volume filling, 58 micropores, 71 microporous ceramic membranes, 669 microporous inorganic membranes, 643 microporous material, 231 microporous membranes, 16, 298, 555
685 ---- highly selective, 660 obtained by chemical vapour deposition, 310 microporous silica membranes, 57, 253 microporous structure, 240 microporous top layers, 239, 240 microporous volume, 248 microscopy, 74 microstructural development, 275 microstructure, 162 MIEC membrane, 437 migration enthalpy, 495 mixed conduction, 438 mixed-conducting oxide membranes, 435 mixin g, 123 mobile turbulence promoters, 611 modelling, 549, 555, 650, 661 modelling equations, 464 modelling membrane processes, 646 modification technologies, 15 modified membranes, 354 modified structures, 26 molecular flow, 348 multi-valent dopants, 472 multichannel monolithic elements, 29 multiphase reactors, 542 multiple step coating, 267 nanofiltration, 595 nanophase ceramics, 240 NEMCA, 548 Nernst-Einstein, 458, 490 Nernst-Planck equation, 575 non-ionic surfactants, 246 non-Newtonian, 215 nonstoichiometry, 483 nuclear magnetic resonance, 87 nucleation, 315 observed rejection, 571 oily emulsions, 620 operating costs, 623
686 ordered microporosity, 249 ordered porous texture, 249 ordering local, 501 vacancy, 495, 497 organic additives, 268 oxidative coupling of methane, 507 oxygen desorption, 488 oxygen flux, 464--466 difficulties in measuring, 504 oxygen permeability data, 440 oxygen permeation, factors controlling, 448 oxygen pumps, 438, 469 oxygen transport equations, 489 packed bed membrane reactor, 650 packed structures, 275 paint, 624 palladium membranes, 660 paper and pulp, 632 partial conductivity, 454 partial electronic conductivity, 463 partial molar entropy, 487 partial oxidation, 543, 549, 553 particle compact, 173 particle packing, 152 advancing contact angle, 198 agglomerate forming, 153 aggregates, 154 binder polymer, 206 binders, 203, 214 Brownian agglomeration, 161 Brownian movement, 160 cake filtration, 154 capillary colloidal filtration, 183 Carman-Kozeny, 187, 212 coating flow dynamics, 190 coating thickness, 189, 190 colloidal filtration, 210 colloidal processing, 163 colloidal stability, 210
SUBJECTINDEX colloidally stable, 210 compact, 207 m concentrated suspension, 173, 175 contact angle, 196 contact time, 212 cracking, 208 critical cracking thickness (CCT), 177 Darcy's law, 158 deagglomeration, 214 Deborah number, 161 m defects, 178-181 dewetting, 155, 181, 190 dip-coating, 183 drainage, 190 drying, 175 drying forces, 153 dust, 181 dynamic contact angle, 198 electrostatic interaction, 164 film thickness, 190 film-coating, 189, 215 gel structure, 154 interaction forces, 162 layer thickness, 212, 215 macromolecules, 156 maximum packing, 171 microstructure, 162 non-Newtonian, 215 particle compact, 173 Peclet number, 169 pinholes, 178 polyelectrolytes, 203, 214 polymer solutions, 205 polymer thickener, 214 polymeric interaction, 166 pore diameter, 207 porosity, 207 random packing, 171 receding contact angle, 198 rheological properties, 158 rheology, 156, 171,215 shear erosion, 212
SUBJECTINDEX shear flow field, 160 shear induced agglomeration, 161 shear induced diffusion, 188 sintering, 175 sintering stress, 176 slip-casting, 183 sol, 158 spreading parameter, 197 stability of liquid coatings, 200 stability ratio W, 162 structure, 207 suspension, 158, 169, 171 m thickeners, 203 thickening polymer, 157 Van der Waals attraction, 163 viscosity, 212 wetting, 154, 190 work of adhesion, 197 work of cohesion, 197 work of wetting, 197 PBCMR, 532, 540 model, 549 PBMR, 531, 540 Pd-alloy membrane reactors, 534 Peclet number, 169, 574, 599, 601, 647, 670 penetration rate, 299 peptization, 229, 233 percolation, 72, 494 threshold, 470 permeability, 334, 594 coefficient, 59, 360, 390, 573 maximum, 392 measurements, 502 permeance, 333 permeation, 8, 333, 338, 645, 660 permporometry, 104 permselectivity, 365 perovskite -brownmillerite two-phase region, 503 membranes, long-term stability of, 511 space, 497 m stability, 488
687 m structure, 482 phase transformations, 282 physical adsorption, 35 physisorption, 78 pinholes, 27, 178 plasticizers, 122, 132 Poiseuille-type law, 338 polyelectrolytes, 203, 214 polymer solutions, 205 polymer thickener, 214 polymeric gels, 230, 248 polymeric interaction, 166 polymeric sols, 237 polymeric specimens, 301 pores m blocking, 373 characteristics, 335 clogging, 261 diameter, 207 growth, 297 hydraulic radius, 71 narrowing, 311 shapes, 23, 72 size, 71,308 size distribution, 49, 293 types of, 25 porosity, 70, 207, 461,307 porous structure, 67 porous substrates, 150 coating technique, 150 colloidal ceramic processing, 150 m dispersion technology, 150 positron lifetime spectroscopy, 97 potable water, 629 pre-filtering, 622 precursor chemistry, 300 preferential sorption, 369 Present-De Bethune model, 361 process integration, 664, 667 propylene, 648 protein deposition, 579 proteins, 630
688 proton conduction, 512 pulsate flows, 591 radiation scattering, 91 random packing, 171 random point defect chemistry, 490 real rejection, 571 reflection coefficient, 573, 608 reflection conditions, 362 rejection, 603 measurements, 98 of salt mixtures, 601 relaxation experiments, 492 relaxation methods, 461 reliability, 6 reliability factor, 5 repairing defects, 311 return on investment, 654, 673 Reynolds number, 584 rheological properties, 158 rheology, 156, 171,215 rotating disc, 593 roughness, 29 roughness effects, 270 saddle point, 495 Saito-Foley equation, 55 salt rejection, 598, 604 sealing, 645, 672 of the membrane, 32 selectivity, 393, 651, 660 selectivity coefficients, 58 separation factor, 8, 364, 366, 390 shape factor, 339 shape selectivity, 389 shear erosion, 212 shear flow field, 160 shear induced agglomeration, 161 shear induced diffusion, 188 silica, 375 silica materials, 248, 253 silica membranes, 236, 301,306, 309
SUBJECTINDEX synthesis route of, 303 silica microporous membranes, 300 silica supported membranes, 239 silica-titania, 375 silica-titania microporous membranes, 300 silica-titania/zirconia membranes, 306 silicalite layer thickness, 321 sintering, 175, 281 mechanisms, 282 stress, 176 SiO2 membrane, 60 SIO2/A1203, 239 SiO2/TiO2, 239 SiO2/TiO2 membrane, 60 SiO2/ZrO2, 239 size exclusion, 380 slip casting, 183, 264 slip coefficient, 345 slip flow, 345 slurry Preparation, 131 small polaron mechanism, 473, 493 sol, 158 sol-gel process, 227, 539 solid oxide electrolytes, 462 solid oxide membranes, 546 sorption, 390 space charge (SC) model, 599 specific surface area, 71,248, 603 spreading parameter, 197 stability, 6, 15 of liquid coatings, 200 ratio W, 162 stage cut, 367 stainless steel supports, 317 static lattice simulation, 495 stereology, 74 Stern layer, 586 sticking probability, 299 streaming potential, 588, 603 stress diagram, 284
SUBJECTINDEX during calcination, 291, 294 in supported films, 279 levels, 296 measurements, 283 model, 292 relaxation, 293, 294, 296 structures, modified, 26 styrene, 657 sugars, 631 supercritical fluid extraction, 609 supports, 5, 27 support system, 143 particulate materials, 143 requirements, 146 quality of, 27, 270 technology, 15 surface acoustic waves, 96 surface area, 7 determinations, 46 surface diffusion, 345, 347 surface exchange kinetics, 506 surface homogeneity, 148 surface modification, 506 surface oxygen exchange, 455 surface processes, 9 surface roughness, 146, 461 layer thickness, 147 support requirements, 146-150 surfactant molecules, 246 suspension,. 158, 169, 171 syn gas, 665 tank bottoms, treatment of, 628 tape casting, 119, 130, 133 tape drying, 133, 134 technical and economic feasibility, 648 techno-economic feasibility of inorganic membranes, 673 template agents, 245, 251 template approach, 310 template effect, 231 template ligands, 302
689 template molecules, 300, 313 template removal, 319 terbia-doped ZrO2-Y203, 478 thermal stability, 297 of y-alumina membranes, 235 thermodynamic factor, 383 thermoporometry, 84 thickeners, 203 thickening polymer, 157 thickness, characteristic, 458 three-phase boundary, 471 titania -alumina composite membranes, 236 -doped ZrO2-Y203, 477 membranes, 236, 289, 298 microporous layers, 243 -zirconia ultrafiltration membranes, 235 Toerell-Meyer-Sievers (TMS) model, 599 tortuosity, 26, 72 definition of, 341 tracer diffusion coefficient, 491 transition flow, 339 transition region, 341 trapping of defects, 453 tubular configurations, 120 ultrafiltration, 590 ultrasonic methods, 95 vacancy diffusion coefficient, 491 vacancy ordering, 495, 497 Van der Waals attraction, 163 vegetable waste water, 622 viscosity, 212, 340, 354 viscous flow, 341 Poiseuille, 334 Vycor glass membranes, 537 Wagner equation, 449, 490 warping, 278 washing operations, 625 water-gas shift reaction, 643, 665
690 wetting, 154, 190 whey, 630 Wicke-Callenbach cell, 363 work of adhesion, 197 work of cohesion, 197 work of wetting, 197 yeast, 632 zeolite chemistry, 312 zeolite layers, 320
SUBJECTINDEX particle sizes, 319 thicknesses, 319 zeolite membranes, 313, 376 synthesis, 317 zeolite precursor solution, 317 zeolite-type membranes, 542. zero point of charge (zpc), 585 zeta potential, 587 zirconia, 241, 243, 298 membranes, 290 ZSM5 zeolites, 377
