Sei sulla pagina 1di 15

Journal of Membrane Science 320 (2008) 42–56

Contents lists available at ScienceDirect

Journal of Membrane Science


journal homepage: www.elsevier.com/locate/memsci

Permeation of gases through microporous silica hollow-fiber membranes:


Application of the transition-site model
Philip Molyneux ∗
Macrophile Associates, 33 Shaftesbury Avenue, Radcliffe-on-Trent, Nottingham NG12 2NH, UK

a r t i c l e i n f o a b s t r a c t

Article history: In a previous paper [P. Molyneux, “Transition-site” model for the permeation of gases and vapors through
Received 29 September 2007 compact films of polymers, J. Appl. Polym. Sci. 79 (2001) 981–1024] a transition-site model (TSM) for the
Received in revised form 1 February 2008 activated permeation of gases through compact amorphous solids was developed and applied to organic
Accepted 2 March 2008
polymers; the present paper examines the applicability of the TSM to permeation through microporous
Available online 18 March 2008
silica. The basis of the TSM theory for amorphous solids in general is outlined; the present extension to
inorganic glasses has revealed that the transition sites (TS) of this theory, which are the three-dimensional
Keywords:
saddle-points critical in the molecular sieving action, equate to the doorways long recognized in per-
Activation energy
Hollow-fiber membranes
meation through amorphous silica and other inorganic glasses. The TSM, which views permeation as a
Microporous silica primary process, is contrasted with the conventional sorption–diffusion model (SDM) for permeation. It is
Molecular sieving pointed out that in the SDM, the widely accepted analysis into two apparently distinct factors – sorption
Permeability theory (equilibrium) and diffusion (kinetic) – has the fundamental flaw that these factors are not independent,
Transition-site doorways since both involve the sorbed state. By contrast, the TSM focuses on the permeant molecule in only
two states: as the free gas, and as inserted in a doorway D; hence the characteristics of these doorways
– (unperturbed) diameter  D , spacing , and the thermodynamic parameters  (force constant) and 
(entropy increment) for the insertion process – can be evaluated. The theory is applied to literature data
[J.D. Way, D.L. Roberts, Hollow fiber inorganic membranes for gas separations, Sep. Sci. Technol. 27 (1992)
29–41; J.D. Way, A mechanistic study of molecular sieving inorganic membranes for gas separations, Final
Report submitted to U.S. Department of Energy under contract DE-FG06-92-ER14290, Colorado School of
Mines, Golden, CO, 1993, www.osti.gov/bridge/servlets/purl/10118702-ZAx4Au/native/1011872.pdf; M.H.
Hassan, J.D. Way, P.M. Thoen, A.C. Dillon, Single component and mixed gas transport in silica hollow fiber
membrane, J. Membr. Sci. 104 (1995) 27–42] on the permeation through microporous silica hollow-fiber
membranes (developed by PPG Industries Inc.) of the nine gases: Ar, He, H2 , N2 , O2 , CO, CO2 , CH4 and
C2 H4 , over the temperature range 25–200 ◦ C. The derived Arrhenius parameters for the permeation of
these gases (excepting He) lead to estimates of the four doorway-parameters:  D , 125 pm; , ca. 30 nm;
, 0.43 nN; , 1.7 pN K−1 ; these values lie within the ranges of those obtained with the glassy organic
polymers. Some “secondary effects”, shown particularly by CO and CO2 , are interpreted as host–guest
interactions at the doorway. The behavior of He is anomalous, the permeation rising linearly with tem-
perature. This study confirms that the TSM may be applied to gas permeation by activated molecular
sieving for this type of inorganic membrane.
© 2008 Elsevier B.V. All rights reserved.

1. Introduction between the gas and that held within by the solid; and diffu-
sion as the movement of the sorbed gas molecules within the
The migration of gases in and through solids has many prac- solid.
tical and theoretical implications, particularly in the use of films In current thought, the primacy in migration is given to diffusion
as barriers, and in separation processes [1,2]. In these processes, [3–6]. With solubility as the equilibrium-based process, perme-
we can distinguish permeation as the transfer of the gas through a ation is then viewed as the composite of solubility and diffusion,
layer of the solid; solubility or sorption as the equilibrium process leading to the sorption–diffusion model (SDM). By contrast, in a
previous paper [7] dealing with migration of gases and vapors
in organic polymers, a model is presented which gives primacy
∗ Tel.: +44 115 933 4813; fax: +44 115 933 4813. to permeation; this was termed the transition-site model (TSM).
E-mail address: molyneux@easynet.co.uk. In the present paper this model is applied to simple gases per-

0376-7388/$ – see front matter © 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.memsci.2008.03.001
P. Molyneux / Journal of Membrane Science 320 (2008) 42–56 43

meating through a microporous form of amorphous silica, using this may be replaced by a cubic lattice (lattice parameter ), with
literature data on hollow-fiber membranes. Because of the nov- the yz plane as the plane of flux, and the migration taking place
elty and unusual approach of the TSM, it is useful to firstly outline down the x-axis in the direction of the concentration gradient in
the theory and its general application (Sections 2.1–2.5), and then the solid and the pressure difference across it. In a full picture,
the results of its application to organic polymers (Sections 2.6–2.7), there would be extensions of these lattice features into the y- and
before considering its specific application to permeation through z-directions.
silica (Section 3). The values of absolute entropy and molecular Combining this with the solubility equilibrium for the inser-
diameter required in the treatment of the experimental data are tion of the molecule into the sorption site trough at (or near)
also discussed (Appendixes A and B). the “upstream” surface, and likewise for the desorption from the
“downstream” face of the membrane, gives the potential energy
2. The transition-site model for permeation profile shown in Fig. 1B. Here it is assumed that the sorption process
is exothermic; however, the picture discussed is not importantly
2.1. Lattice picture of the migration process altered if this sorption process is endothermic. On the SDM, the
migration then takes place by a series of activated jumps between
To illustrate the relations and concepts dealt with in this paper, successive locations of the sorbed molecules.
Fig. 1 shows the main features of the lattice picture that is used This lattice picture is well established, originating apparently
in the SDM and has likewise been adopted for the present TSM. with the pioneering work of Barrer on activated migration in
The migration process requires there to be “free volume” regions diverse media [8,9] and used since then in numerous molecular
that take up the sorbed guest molecules, and that interconnect via modeling studies [10–14]. In fact, the lattice picture used here mim-
more restricted locations. For an amorphous isotropic solid, these ics that of interstitial migration in crystalline solids (for example,
will be located and oriented statistically through the bulk of the atomic species such as H, C and N in crystalline Fe); however, the
solid, as fixed in position when (for example) the material solid- present TSM does not seem to have been applied to such processes
ified. However, as shown in Fig. 1, for mathematical tractability [4,6].

Fig. 1. Schematic of the activated diffusion and permeation processes for an amorphous solid film (thickness L), showing the four environments for the permeant gas molecule
G: , molecule in the upstream gas phase at pressure p0 ; , sorbed molecule; , molecule in transition-site doorway D; , molecule in the downstream gas phase, at
pressure pL . (A) Pressure gradient: linear gradient of pressure px (at general depth x) within the film (ideal permeation for low p). (B) Energy profile: The amorphous solid
is modeled by a cubic lattice (spacing ), with yz (perpendicular to the plane of the diagram) as the plane of flux, and with migration occurring in the x-direction (left to
right). The troughs are the preferred sorption sites, and the peaks are the transition-site doorways D. Ordinate energy axis indicates the enthalpy change of sorption HS , the
diffusion activation energy ED and the permeation activation energy EP . (C): “Free volume” and “doorways”: Sequence of environments across the film—permeant molecule
diameter  G and unperturbed transition-site doorway diameter  D as labeled; the complete lattice would have corresponding features in the y- and z-directions.
44 P. Molyneux / Journal of Membrane Science 320 (2008) 42–56

In parallel with the energy profile shown in Fig. 1, there must obtained from P0 on the entropy effects in the migration process,
also be an entropy profile; this is seldom considered, but evidently and on the spacing  between transition-sites (Fig. 1).
must be taken into account to give the true, free energy descrip- A more general point in connection with these energy levels is
tion of the processes involved. In this case, the profile would have that they allow the permeation activation energy EP to be zero or
troughs where molecules are restricted; a further difference is that negative if ED becomes equal to or less than HS , so long as ED has
in contrast to the energy situation where the energy level is zero for a positive value as required for the jump process.
all dilute gases, in the case of entropy the starting levels would be
different following the absolute entropies of the gases considered. 2.3. Critique of the sorption–diffusion model (SDM) for analyzing
This has to be taken into account in interpreting the van’t Hoff and permeation
Arrhenius prefactors as discussed below. In previous cases where
entropy effects have been considered [9,10,14], they have dealt only The conventional and widely accepted belief is that permeation
with the diffusive jump process. must be considered as composite, that is, the resultant of sorption
(equilibrium) and diffusion (kinetic) effects. This seems to be sup-
2.2. General relations ported, for example, by the form of Eq. (1), and by the simplicity of
the units for diffusion coefficient (m2 s−1 ) compared with those for
On the basis of “ideal conditions” that is, low pressure p of permeability coefficient (mol m m−2 s−1 Pa−1 ).
the permeant gas and low concentration c of dissolved perme- However, as detailed in the previous paper [7], there are a num-
ant, the three determining parameters—the Henry’s Law solubility ber of instances where permeation is a simpler process than would
coefficient, S,1 the diffusion coefficient D, and the permeability be expected if it were the composite of independent effects. In this
coefficient P are related [2] by section, this is taken a step further by showing that, despite the
almost universal view that diffusion must be considered as the pri-
P = DS. (1) mary process in solid-state migration, there are a number of aspects
that raise doubts as to its validity for this purpose. These may be
It is this equation that is used as the basis for the conventional
considered under three headings.
SDM, which considers that permeation must be a composite quan-
tity, depending on both the sorptive and the diffusive properties of
2.3.1. The sorption and diffusion processes are not independent
the system.
The primary point of weakness of the SDM arises in the uni-
In considering the molecular bases of these processes, the
versally assumed analysis of permeation into the two assumedly
temperature dependence of the three parameters is of particular
distinct aspects, sorption (equilibrium factor) and diffusion (kinetic
importance. These generally take the similar van’t Hoff (sorption)
factor). Any analysis of a physical or chemical process requires that
and Arrhenius (permeation and diffusion) forms. In the light of Eq.
the proposed component effects are independent; for example, in
(1), this gives for the corresponding prefactors
analyzing the factors controlling the volume of a fixed amount of an
P0 = S0 D0 (2) ideal gas, these two factors would be the pressure and the (absolute)
temperature, each of which may be varied independently. However,
and for the energy quantities in the case of the SDM, the two coefficients S and D are clearly not
independent, since they both involve the sorption state. For the
EP = HS + ED (3)
sorption coefficient S relates to the equilibrium between the gas
where EP and ED the activation energies for permeation and diffu- state and the sorption state; while the diffusion coefficient D relates
sion, and HS is the isosteric enthalpy change of sorption. to the ease of a jump from the sorption state across the intervening
In Eq. (3), if the sorption process is exothermic (so that barrier (doorway), or (on transition-state theory) to the equilibrium
HS is negative), then the value of the permeation activation between the sorption state and the transition state (Fig. 1B).
energy EP becomes the difference between the numerical values Of course, this is not to argue against the classical picture of
of HS and ED ; for example, in the case of H2 O with poly(ethyl permeation being a sequence of processes that we call “sorption”,
methacrylate) (PEMA), the studies of Stannett and Williams [15] “diffusion” and “desorption”; nor is it to argue with the mathemat-
gave HS = −34 kJ mol−1 (exothermic) and ED = 36 kJ mol−1 so that ical validity of Eq. (1). But neither of these leads to, or necessarily
EP = 2 kJ mol−1 (compare Fig. 1B). requires, the conventional SDM analysis.
As a general point, inasmuch as each permeant has spe-
cific Arrhenius and van’t Hoff parameters, the plots for different 2.3.2. The sorption sites are heterogeneous
permeants may cross. This is particularly important with the Considering the sorption of a single simple-molecule permeant
increasing emphasis on the use of membranes over wide tempera- in an amorphous solid, such as an organic polymer, or an inorganic
ture ranges. For example, the recent work of Wilks and Rezac [16] glass, because of the disordered molecular structure of the host
showed, for the competitive permeation of CO2 and H2 through matrix, the permeant molecules will be held in a variety of sites
poly(dimethylsiloxane) (PDMS) membranes, that the selectivity and locations. For example, the variation of solubility and partial
factor is 4.2 at 35 ◦ C (CO2 faster), but that this has fallen to 0.8 (H2 molar volume with uptake indicates a Gaussian distribution for the
faster) at 200 ◦ C. Thus the comparison of the permeation behav- sorption sites both with glassy organic polymers [17a] and with
ior at a single temperature is not sufficient to elucidate migration vitreous silica [17b]. It is therefore evident that the starting point
behavior. for the diffusive jump (Fig. 1) is not well defined.
Furthermore, any molecular interpretation should seek to
explain and correlate not only the activation energy, but also the 2.3.3. It is the pressure gradient and not the concentration
prefactor. In particular, using the TSM [7], valuable information is gradient that is the true driving force for transport
The deficiencies in taking the concentration gradient as the
driving force, are seen from the molecular viewpoint, when we
1
This symbol for solubility coefficient, used here in conformity with customary
compare the behavior of two permeants O2 and H2 O, which have
practice, should not be confused with that for entropy (S0 , etc.) used later in the similar molecular diameters (293 pm and 289 pm from critical
paper. constants—see Appendix A and Ref. [7]), in a polar polymer such as
P. Molyneux / Journal of Membrane Science 320 (2008) 42–56 45

PEMA [15]. For a given partial pressure of the penetrant, the uptake be located on the average at this midpoint. Finally, because of the
of the H2 O is much greater than that of the O2 because the H2 O will Boltzmann factor exp(−E/RT) that controls the fraction of perme-
hydrogen bond to the ester groups in the polymer chains. However, ant molecules with the required excess energy E, in making the
for the same concentration gradient, the actual amount of H2 O free jump the penetrant molecule will expend the minimum amount of
for diffusion will be much smaller than that for the O2 , because of energy necessary for a successful jump, that is, barely passing over
this hydrogen bonding. Thus the use of the concentration gradient the central point. This location is the transition-site doorway D,
clearly does not represent the true driving force for the migration which has the nature of a three-dimensional saddle-point (Fig. 1B
process. and C).
In this example, where the structures and interactions for the
solid and the migrant molecule are well established, the molecular 2.5. Derivation of equation for permeability coefficient on the
basis is self-evident, but it must apply more generally where these TSM
forces may not be well understood.
These objections do not apply when we consider the pressure The present TSM is then a direct application of the well-
gradient as the driving force, since this is a true measure of the established standard theory of the transition state (also termed the
gradient of chemical activity (Fig. 1A). The present TSM uses this activated complex) for chemical reactions [26]. For an elementary
picture, and shows how this gradient may be linked to the migration chemical reaction, there is postulated to be equilibrium between
process. the starting reactants and those in the transition state, this being
The validity of the pressure gradient as the driving force the position of maximum potential energy on the lowest energy
is revealed in numerous cases where, when the conditions are path from the reactant to the products; this transition state then
changed, the values of D and S vary markedly but that of P remains converts into the products of the reaction (or may revert to the
essentially constant. Since the sorbed state enters in opposite ways reactants). For example, for the transfer of the atom B between the
in the sorption and diffusion processes, its effect cancels out in species A and BC, the two-step sequence involved is:
permeation.
For example, considering a case with the same permeant in the A + B − C  [A· · ·B· · ·C] ‡  [A − B + C]. (4)
Reactants Transition state Products
same solid at different pressure and concentration levels, Rundgren
et al. have shown recently [18] with H2 /vitreous silica at 550 ◦ C, that Although this picture is well accepted for reactions in the gas
when the upstream pressure was varied up to 0.12 MPa, the value phase, there is some hesitation in applying it to processes in con-
of P remained essentially constant (±10%) although the value of S densed phases. However, it is well established as the explanation of
decreased by a factor of approximately three while that of the dif- the primary salt effect for reactions in aqueous solution, that is, the
fusion coefficient D correspondingly increased by the same factor. effect on the rates of reactions between ions in solution of added
Much the same effect can apply to the same permeant in differ- “neutral salt” (an electrolyte with non-common ions); here it is the
ent solids, as with the case of H2 O in moderately polar polymers net charge on the transition state that is important, and this state is
previously discussed [7]. A striking example of this type of effect evidently sufficiently long-lived for the Debye–Hückel ionic atmo-
is shown in the work of McCall et al. [19] on the permeation of sphere from the added neutral salt to be established [26]. Moreover,
H2 O through poly(ethylene) (PE) films that had been air-oxidized the theory has been applied successfully in considering perme-
in processing to varying extents, leading to the incorporation of ation in physiological systems [27], while the concept of “transition
hydrophilic groups on the polymer chains; such oxidation up to state analogs” in now widely used in enzyme chemistry [28]. More
3% combined oxygen gave a 50-fold increase in the sorption coeffi- specifically, this type of picture was used some time ago to inter-
cient S compared with the starting PE, but an essentially equivalent pret quantum effects in the permeation of “light” gases (He, H2 ,
decrease in the diffusion coefficient D, so that the permeability D2 ) through organic polymers [10], and more recently to similar
coefficient P was essentially unchanged (±20%) by the oxidation. systems using molecular dynamics simulations [29].
The above arguments apply not just to organic polymers or vit- With permeation through an amorphous solid, the pressure dif-
reous silica from which the examples have been chosen, but to any ference p across the layer of solid also corresponds to a gradient of
system where the SDM analysis is currently used. partial pressure of the gas within the solid, representing a gradient
of chemical activity across the layer. For the ideal case under present
2.4. Analysis of the jump process discussion (dilute gases, and dilute solutions of the penetrant in
the solid, so that there is no “plasticization” or other modification
In the field of migration through organic polymers, little atten- of the host solid by sorbed permeant), this gradient will be linear
tion has been paid to the nature of the peak on the path of the (Fig. 1A). Concretely, the partial pressure at any level would be that
jump, with greater attention being paid to the “free volume” which within a microscopic cavity at that level; alternatively, it would be
relates to the environment of the sorption sites [2]; however, the the pressure realized if the solid were sliced through in the yz-plane
previously developed TSM [7] focuses particularly on the state of at that level, the two parts slightly separated, and the permeation
the permeant molecule at this peak. It turns out, that in the field flux resumed. This does not mean that there are actual free per-
of migration in inorganic glasses, these transition sites have been meant molecules in the solid, but simply that any depth x can be
recognized for over half a century, and were termed doorways [20], assigned its value of the partial pressure px .
a name that has been retained since [21–25]. It is therefore conve- Applying these concepts to the cubic lattice picture having lat-
nient to use term doorway (D) interchangeably with transition-site tice parameter  (Fig. 1), the occupancy of doorways at depth x is
(TS) in the rest of the paper. controlled by the insertion equilibrium:
A simple analysis of the jump process further highlights the
G(x) + D(x)  [GD(x)]‡ (5)
significance of the TS doorway. Firstly, if we apply the Principle
of Microscopic Reversibility to a jump, it follows that the reverse where D(x) is an unoccupied doorway at the depth x, GD(x) is an
process must also be acceptable as a jump; thus on the average occupied doorway at that depth (labeled with the double cross “‡ ”
the jump will be symmetrical about the midpoint. Secondly, the to show its analogy with the transition state in Eq. (4)), and G(x)
requirement for activation of the molecules to enable the jump to relates to the partial pressure at that level as already discussed.
take place indicates the presence of an energy barrier, which would The rate of loss of G from the occupied doorways (downstream or
46 P. Molyneux / Journal of Membrane Science 320 (2008) 42–56

upstream—Fig. 1B) is then taken to be controlled by the standard [(S(GD) − S(D)] in Eqs. (9), (10) and (13) is the equivalent entropy
transition-state rate coefficient k‡ ≡ kB T/h where kB is the Boltz- change for this same process.
mann constant and h is the Planck constant [26]. On the TSM, it It is useful to analyze the value of EP into three presumably
is this gradient of occupied sites [GD(x)]‡ that gives rise to the independent contributions to the insertion process of Eq. (5):
observed flux in the downstream direction.
EP = EMM + Eex − EMG (14)
This then leads [7] to an expression for the flux that correctly
incorporates its experimentally observed proportionality to the where EMM is the energy to overcome any noncovalent interactions
area, the pressure difference across the film, and the reciprocal of between the host matrix (M) across the unperturbed doorway aper-
the film thickness. The permeability coefficient P then becomes ture, Eex is the energy to expand the matrix by the insertion of
 k T    the (presumably) larger gas molecule G (taken to be a incompress-
B ‡ −U ‡ ible sphere), and EMG is the energy released from any interactions
P= exp(S ) exp (6)
ehNA RT (presumably, noncovalent) between the molecule of G and the sur-
rounding host matrix in the doorway.2 The first contribution, EMM ,
where e is the exponential number, NA the Avogadro constant, S‡
is neglected here, and this is supported by the fact that the door-
the entropy change for the insertion of a mole of the gas G into a
ways are later deduced to have a diameter greater than 100 pm
mole of the doorways D according to Eq. (5), and U‡ is the internal
(Section 3.5).
energy change for this same process. The Arrhenius prefactor and
In previous analyses of this process as applied to silica and
the activation energy then become
derived inorganic glasses, attention has been focused on the second
 k T   
S ‡ contribution Eex . In its earliest development the doorway was con-
B
P0 = exp (7) sidered as a spherical cavity, but this was later replaced [20] by the
ehNA R
somewhat more realistic picture of a cylindrical cavity; this would
EP = U ‡ . (8) then lead, using the mechanics of elasticity for the expansion of a
thick-walled pipe, to the relation
Considering Eq. (7), evidently the prefactor P0 is not expected
GD (G − D )2
to be truly temperature independent since the expression has the Eex = (15)
factor T in front, but over narrow temperature ranges the variation 4
is much less than that of the exponential factor, so that it is per- where G is the shear modulus of the bulk solid,  D the diameter of
missible to use the average temperature in applying this equation. the unperturbed doorway, and  G is the diameter of the permeant
In any case, there may also be a further complicating effect from molecule (see Fig. 1C). However, this commonly applied formula-
changes in S‡ with temperature. tion [18,20–25] suffers from two defects: firstly, the doorway would
The entropy of activation S‡ can be expressed in terms of the be saddle-shaped (Fig. 1C), rather having parallel sides as Eq. (15)
entropies of the “reactants” and “products” for the insertion process requires; and secondly, there is no reason to believe that the elastic
(5) as behavior of the solid at the molecular level (relating to the rotation
and flexion of individual bonds) follows that for the bulk solid as
S ‡ = S(GD) − S(D) − S 0 (G) (9) defined by the shear modulus G.
In many previous formulations [10,18,20–25,29], Eex has been
where S(GD) is the entropy of 1 mol of occupied doorways, S(D) that
equated to the activation energy for diffusion, which ignores the
of the occupied sites, and S0 (G) is the absolute molar entropy of the
part played by the sorption enthalpy in the diffusive jump process
gas. Substituting in Eq. (7) and rearranging gives
(Fig. 1B).
 k T   [(S(GD) − S(D)] 
B Summarizing, in contrast to the diffusive jump picture, in the
P0 exp[(S 0 (G)/R)] = exp . (10)
ehNA R TSM the permeation activation energy is referred to a well defined
initial state, that of the free gas, with reference energy of zero for
It is convenient to define a quantity Y, the entropic coefficient,
any gas (Fig. 1B); however, the prefactor P0 has to be adjusted for the
relating the original permeability coefficient P and the absolute
different absolute entropies of different gases to give the entropic
entropy by
prefactor Y0 according to Eq. (12).
Y ≡ P exp[(S 0 (G)/R)]. (11)
2.6. Application of the TSM to experimental data for organic
Correspondingly, for the limiting case of 1/T → 0, this defines the polymers
entropic prefactor Y0 as
In the previous application of the TSM to organic polymers [7], it
Y0 ≡ P0 exp[(S 0 (G))/R]. (12) was tested on the collected literature data [30] for the permeation
of a variety of gases (ranging from He to SF6 ) with 16 amorphous or
This last expression is used in evaluating Y0 from permeation
semi-crystalline polymers using three test plots, involving the two
experiments, as discussed below. In these expressions for Y and Y0 ,
parameters EP and log Y0 together with the permeant gas molecular
the units of pressure implicit in the absolute entropy and explicit in
diameter  G :
the permeability coefficient have to be brought into concordance,
as discussed in Appendix A. Plot A: EP versus  G
Substituting from Eq. (12) in Eq. (10) and taking logs gives Plot B: log Y0 versus  G
 k T   [S(GD) − S(D)]  Plot C: log Y0 versus EP .
B
log Y0 = log − log  + (log e) . (13)
ehNA R
Previously [7] there had been four test plots, with the first of
As discussed below, this relation is useful in making estimates these (“Case 1”) a direct plot of the original Arrhenius parame-
of the doorway spacing .
Reverting to Eq. (8), the permeation activation EP is revealed as
the internal energy change U‡ to insert the molecule of G into the 2
Kirchheim has shown the importance of this site expansion energy term Eex for
doorway D (Eq. (5) and Fig. 1C). Likewise, the entropy difference the sorption of gases both by organic polymers [17a] and by vitreous silica [17b].
P. Molyneux / Journal of Membrane Science 320 (2008) 42–56 47

ters, that is, log P0 versus EP , but in view of the better theoretical increase in the internal energy change U‡ , shown to be equal to
basis for the entropy-adjusted form log Y0 , this first plot is not now the activation energy EP (Eq. (8)), to insert the molecule of G into
considered to be useful. the doorway D (Fig. 1). Focusing on the expansion contribution Eex
With certain exceptions [7], these three test plots were found in Eq. (14), if we follow this linear plot back down to its intersec-
to be essentially linear, and may be fitted to the respective conven- tion point with the x-axis (EP = 0), this point then represents the
tional forms condition where there is no energy required, that is, the molecule
exactly fits into the doorway (see Fig. 3 below Section 3.4). This
Plot A : EP = m1 G + c1 (16)
indicates that for the set of permeant molecules considered, there
Plot B : log Y0 = m2 G + c2 (17) are a fixed number of pre-existing doorways of diameter  D that
control the migration of the permeant molecules through the host
Plot C : log Y0 = m3 EP + c3 . (18) medium. The energy dependence of Eq. (16) may then be put more
specifically in the form
In the context of the present paper, and to anticipate the
discussion on microporous silica in Section 3 below, these EP = (G − D ) (21)
plots are exemplified by Figs. 3–5, respectively, below (Section
3.4). where the coefficient  is the force constant for the expansion of the
Before continuing with the interpretation of these Eqs. (16)–(18), doorways for this host solid.
it should be emphasized that their linearity does not follow directly Likewise, the second linear relation Eq. (10) (Plot B) shows that
from the theory, neither is it in any sense an integral part of it. It there is correspondingly a linear increase in the entropy of the
should also be evident that these are not three independent equa- matrix at the doorway when the molecule is inserted, due to the
tions, since they involve only three independent variables; this expansion by the inserted molecule (see Fig. 4 below Section 3.4);
leads to the parameters for the third equation as this follows because log Y0 is directly related to the entropy change
m2 [(S(GD) − S(D)] by Eq. (13), and the first two terms on the right hand
m3 = (19) side of this equation are independent of  G . This may be put in the
m1
more specific form:
(c2 m1 − c1 m2 )
c3 = . (20) (log e)(G − D )
m1 log Y0 = log Yz + (22)
R
As noted above, with the organic polymers not all the sys-
tems from the literature conformed to these linear relations of where the coefficient  is the corresponding entropy increment for
Eqs. (16)–(18). In the case of the permeants G, the commonest the doorway expansion process for that solid, and Yz is the limiting
anomalies were seen [7] with the two permeants CO2 and H2 O. value of Y0 for  G =  D .
The anomalies were in most cases only seen in the Plot A and These two parameters  and  therefore define the thermody-
Plot C, and not in the Plot B (see Eqs. (16)–(18)), indicating that namics of the expansion process. They are unusual in that, in the
this is related to an anomaly specifically in the permeation activa- case of the energy change, because this increases linearly with the
tion energy EP , with this value being lower than expected from expansion, this represent a constant force being exerted by the door-
the behavior of other gases; this seems to be the cause of CO2 way, rather than an increasing force as might be expected for a
being a relatively “fast” gas in permeation [1,7]. At the time of bulk solid with constant elastic modulus. In particular, it is in con-
the previous publication [7] no clarification could be given of trast with the form of Eq. (15) that is derived from the mechanics
this effect with CO2 . However, re-evaluation of this data [31] of elasticity for a cylindrical cavity in the bulk solid; however, the
indicates that there an essentially constant energy anomaly (low defects of this approach have already been noted in the discus-
EP ) of 18(5) kJ mol−1 for the organic polymers studied with this sion of this equation. Speculatively, the linear behavior may arise
permeant, with little if any correlation of the individual values from the transition-site doorways being asymmetric (transversely
with such parameters as the cohesive energy density of the poly- elliptical or slot-shaped) rather than being symmetric (circular) as
mer, or the enthalpy change of sorption HS of CO2 into the depicted by Fig. 1C.
bulk polymer. It seems that this effect results from to an ener- This same Plot B also provides a method to obtain an estimate
getically favorable interaction between the quadrupole moment of the doorway-spacing parameter , which equivalent to the jump
of CO2 molecule at the transition site doorway and the polar length of the diffusion model (Fig. 1). Considering Eq. (13) for log Y0 ,
groups in the surrounding host matrix. This is thus an example the limit EP = 0 corresponds to  G =  D , with the gas molecule G just
of the matrix-permeant energy contribution EMG towards EP in fitting into the doorway, there will be no perturbation of the door-
Eq. (21). This interpretation is supported by the fact that CO2 is way by the penetrant; so that not only will there be zero energy
well known to interact in a specific way with the bulk matrix of change but there will also be zero entropy change, that is the
many polymers [32,33]. This effect is also seen with silica (Sec- entropy term on the right of Eq. (13) will be zero; this corresponds
tion 3 below) and is discussed further there. A similar energy to the value of log Yz from Eq. (22). Thus Eq. (13) gives
anomaly occurs with H2 O [7], where parallel interactions could k T 
B
occur due to polar interactions, or possibly hydrogen bonding, at log Yz = log − log  (23)
ehNA
the transition site doorway; this would also explain why H2 O is
also characterized as a “fast” gas (vapor) in membrane separation from which in principle the value of  may be estimated since all
processes [1]. the other quantities in this equation are known.
Considering the third linear relation, Eq. (18) (Plot C), as already
2.7. Interpretation of the linear plots noted its linear form follows directly from Eqs. (16) and (17) (see
Fig. 5 below Section 3.4).
The simplest explanation of these linear relations in Eqs. The goodness of fit of these linear Eqs. (16) and (17) depends
(16)–(18) seems to be as follows. necessarily on well-defined values of the molecular diameter  G ,
In the first place, the linear relation of Eq. (16) (Plot A) indi- both with these organic polymers and with the inorganic amor-
cates that with increase in molecular diameter  G there is a parallel phous solid, silica, considered in Section 3 of this paper. The sources
48 P. Molyneux / Journal of Membrane Science 320 (2008) 42–56

and suitability of these values are therefore discussed critically in character of the migration so as to give useful separation of different
Appendix B. gases.

2.8. Effects of finite uptake 3.2. Permeation data of Way et al. for microporous silica
hollow-fiber membranes
As noted initially in Section 2.2, the present treatment applies
specifically to the “ideal case” of low pressures and low concen- In applying the TSM to microporous silica, what is required is a
trations. In real systems there are often deviations at the higher set of data for a set of permeant gases having a range of molecular
pressures and sorbed concentrations, so that the sorption coeffi- diameters, studied over a sufficiently wide range of temperature
cient S depends upon p, and D upon c. that good Arrhenius data may be obtained and the test plots of Eqs.
In the TSM, these effects may be viewed in the first instance as (16)–(18) may be applied as already discussed in Section 2.6. This
the result of the sorbed G on the characteristics of the doorways; mirrors the requirements for the data sets for organic polymers
taking the spacing ␭ to be fixed for the material, this then involves used in the previous development of the TSM [7]. This requirement
the remaining three parameters  D , , and . To disentangle these is largely fulfilled by the literature data on the microporous sil-
effects, it would be necessary to measure EP and Y0 as before over ica hollow-fiber membranes that were produced and patented by
a range of uptakes c, with care being taken that the uptake is the PPG Industries [36,37], and that were used in the three permeation
same for each range of temperature—for example, that there is the studies published by Way and colleagues [38–40], where they were
same value of c at 25 ◦ C as there is at 200 ◦ C. referred to as “developmental products”. As detailed in the refer-
In some cases the sorbed molecules may be so strongly bound ences, the hollow-fiber membranes had been produced by melt
as to have little if any effect on the transition-site doorways, so extrusion of a borosilicate glass, followed by exhaustive acid leach-
that the value of P remains essentially constant. This seems to be ing to remove Na and B, producing a network of pores with diameter
the situation in the case of H2 O with medium-polarity polymers claimed to be less than 20 Å (2 nm) [36,37]. For the present pur-
[7], and likewise of H2 O/oxidised PE [19] considered in Section 2.3. poses, it is convenient to refer to this form of microporous silica as
This may also apply to the similar behavior for H2 with vitreous ␮SiO2 , to distinguish it from (for example) normal compact vitre-
SiO2 [18] also considered in that section. ous silica, vSiO2 . However, this is only one example from a diverse
At higher pressures there may be the purely hydrostatic effect range of these materials.
of the gas G upon the insertion process of Eq. (11), which may be Before considering these data further, it should be noted that
formalised as a volume change of activation V‡ ; this would lead there are parallel data, on another sample of the PPG product, pub-
to a contribution pV‡ to the free energy change G‡ and thence to lished by Shelekhin et al. [41,42]. Since there are marked differences
the activation energy EP , which would thus show up in the Arrhe- in the results from these two groups, the latter are considered sep-
nius plots. At such higher pressures, it may be necessary to use the arately below in Section 3.8.
gas fugacity as the activity-corrected form in place of the pressure. The fibers studied in two main papers from Way et al. [39,40] had
At higher uptakes still there may be effects from blocking of the inside and outside diameters of the fibers of 35 and 45 ␮m, giving a
sorption cavities to the release of the molecule from the TS doorway, nominal wall thickness of 5 ␮m, while the material itself was found
or plasticization of the whole solid matrix, but in this case the solid to be amorphous, with about 20% porosity and with polydisperse
must then be considered to be different in nature from the pure pores averaging about 10 nm. There was also physisorbed molecular
material. water (not quantitified) that was removed on heating above 450 K,
with exposure to the atmosphere at normal temperature leading to
its resorption.
3. Application of the TSM to permeation through The present paper focuses on the main studies in the two
microporous silica hollow-fiber membranes later publications [39,40], which used nine permeant gases: Ar,
H2 , He, N2 , O2 , CO, CO2 , CH4 and C2 H4 , whose permeation rate
3.1. Inorganic membranes for permeation was measured over 25–200 ◦ C. A pressure differential p = 21.4 atm
(2.17 MPa) was used for the main permeation studies.
Although organic polymers continue to play an important role as The studies also included measurements with gas mixtures;
membranes for separation processes, there is an increasing interest such studies are naturally important in relation to gas separation
in inorganic membranes, because of favorable properties such as [1,11,41,45,47]. However, only the single-gas measurements have
greater heat stability [11,34]. It is therefore useful to see to what been considered here.
extent the TSM model applies to this latter type of membrane. In For convenience of the future discussion, some important phys-
the present case, this is applied specifically to the inorganic glasses, ical properties of these nine permeant gases are listed in Table 1:
notably silica as the paradigm simplest case; some other types of (electrical) polarity factors (molar polarizability, dipole moment,
inorganic membranes are discussed briefly in Section 3.9. and quadrupole moment) [48,49], the absolute entropy S0 [50] (see
There is a great bulk of previous literature studies on permeation Appendix A), the critical constants [51], and various sets of values
through and diffusion in glasses [3,5,6]. Unfortunately, it is difficult of the molecular diameter  G [52–56] (see Appendix B).3
to link these studies into a coherent picture, not least because there The data from first of the three Way publications [38] show
are so many varieties of silica; for example, Doremus [35] has listed certain inconsistencies both internally (nonlinear Arrhenius plots)
four distinct commercial types of compact (nonporous) vitreous and with the later data [39,40]; they must therefore be viewed
silica. Furthermore, numerous novel types of microporous silica have as only preliminary, and (except for He) they are not considered
been developed recently, made for example, by the acid leaching further here.
of borosilicate glass in hollow-fiber form [36–42], or as a sol–gel
form by coating a polymeric silica sol onto a more porous substrate
such as such as alumina [34,43–46]. In this context, “microporous” 3
Data in this paper are given in the internationally standard concise form: “mean
implies a substance with pore diameter less than about 2 nm. Such value (limits of error in the last decimal place of the mean value)”, so that for
porosity is intended to confer a higher permeability than that for example, “1.23(4)” represents “1.23 ± 0.04” (see: http://physics.nist.gov/cuu/ Con-
the compact form, while at the same time retaining the activated stants/index.htm).
P. Molyneux / Journal of Membrane Science 320 (2008) 42–56 49

Table 1
Physical and molecular properties for the gases G that were studied as permeants with the microporous silica hollow-fiber membranes [38–40]

G Polarity factors S0 d Critical constantse Molecular diameters (pm)f


˛e a pe b Qe c pc Tc Vc Breck LJ CC

Ar 1.00 0 nd 251 4.87 151 75 340 343 (5) 295


H2 0.49 0 1.7 227 1.29 33 64 289 292 (4) 260
He 0.13 0 nd 222 0.23 5 57 260 258 (3) 210
N2 1.03 0 −4.7 288 3.39 126 90 364 371 (6) 313
O2 0.94 0 nd 256g 5.04 155 73 346 348 (8) 294
CO 1.18 0.4 −9.5 294 3.50 133 93 376 367 (7) 315
CO2 1.51 0 −14.3 310 7.38 304 94 330 398 (8) 324
CH4 1.47 0 nd 282 4.60 190 99 380 380 (6) 324
C2 H4 2.52 0 13.1 315 5.04 282 130 390 416 (7) 359
a
Molar electric polarisability (cm3 mol−1 ) [48,49].
b
Electric dipole moment (C m) × 1030 [49].
c
Electric quadrupole moment (C m2 ) × 1040 [49].
d
Absolute entropy at 298 K (J K−1 mol−1 ) [50], adjusted to 1 Pa as standard pressure—see Appendix A.
e
Critical constants: pressure pc (MPa), temperature Tc (K), molar volume Vc (cm3 mol−1 ) [51].
f
Molecular diameter values (pm) with sources: Breck [52], Lennard–Jones (LJ) [53–56], and “critical constants” (CC) [51]—see Appendix B.
g
Oxygen – special value of S0 – see Appendix A.

3.3. Permeability coefficients and the Arrhenius plots with the values of log Y0 (entropic prefactor—Eq. (12)) derived using
the values of the absolute molar entropy S0 (at 298 K, adjusted to
Considering the two main papers, the data in earlier paper [39] the standard pressure 1 Pa) as given in Table 1. These Arrhenius
were reported as permeability coefficients P, whereas those in later parameters may be taken to refer specifically to 298 K, since the
paper [40] were only quoted as permeances Q, that is, the ratio P/L Arrhenius plots are linear at least down to this temperature (Fig. 2)
where L is the membrane thickness. Correlation of the two sets and the entropy values S0 do refer specifically to this value. Table 2
of data indicated values of L between 4.4 and 4.9 ␮m; this is in also show, for comparison, the corresponding literature values of
accord with the direct measurements for the internal and exter- the Arrhenius parameters for permeation through normal vitreous
nal diameters [40] that gave L = 5 ␮m. The permeance values [40] silica, vSiO2 ; only with four gases – Ar, H2 , N2 and O2 – are com-
were accordingly converted into values of P using L calculated for parative data available for both types of silica [57], while there are
that sample. The Arrhenius plots for the combined data [39,40] are only from single literature sources for Ar [58] and for O2 [59]. The
shown in Fig. 2. The linear plots shown fit the experimental values comparison shows that the activation energies EP are much larger
of P with an average deviation of 6%. The anomalous behavior of He with normal vitreous silica than they are with the present microp-
is considered below in Section 3.7 (see also Fig. 7). orous type. The EP -ratio for H2 is about 1.2, whereas those for the
other three gases the ratios are very similar, averaging 2.14(5); this
3.4. Arrhenius data and derived parameters for the main gases latter fact suggests that the structure of the present ␮SiO2 differs
in a consistent way from that of the normal compact form.
Table 2 lists the Arrhenius parameters for the eight gases (that
Reverting to the ␮SiO2 data, they are plotted along with the
is, excepting He) derived from the linear plots in Fig. 2, together
molecular diameters  G to give the three plots as already used with
organic polymers [7] as discussed above (Section 2.4):

Fig. 3, Plot A: EP versus  G ;


Fig. 4, Plot B: log Y0 versus  G ;
Fig. 5, Plot C: log Y0 versus EP .

Table 2
Arrhenius parameters for the permeation of the gases G through microporous silica
(␮SiO2 ) hollow-fiber membranes (Fig. 3), derived in this paper from the data of Way
et al. [39,40], compared with literature data for compact vitreous silica (vSiO2 ) [57]

G ␮SiO2 vSiO2
9 + log P0 a EP b log Y0 c 15 + log P0 a EP b

Ar 0.87(4) 48.6(4) 4.98(5) 0.42d 106d


H2 0.38(5) 31.2(4) 3.18(5) 1.9(1) 37(1)
He nde nde nde 2.1(1) 22(1)
N2 0.12(13) 46.3(10) 6.11(13) 4.0(4) 100(8)
O2 0.64(6) 44.4(4) 5.01(6) 0.61f 93f
CO 0.70(5) 49.2(4) 7.06(5) nd nd
CO2 0.26(12) 38.9(10) 7.45(12) nd nd
CH4 0.89(10) 54.8(8) 6.62(10) nd nd
C2 H4 1.00(11) 57.8(8) 8.51(11) nd nd
a
P0 is the permeation prefactor (mol m m−2 s−1 Pa−1 ); note the different additive
factors for the ␮SiO2 and the vSiO2 results.
b
Permeation activation energy (kJ mol−1 ).
c
Y0 is the entropic prefactor (mol m m−2 s−1 )—see Eq. (12).
d
Fig. 2. Arrhenius plots for the permeation of gases as labeled through the micro- Single value [58].
e
porous silica hollow-fiber membranes. He: quadratic fit (but see also Fig. 6); other Nonlinear Arrhenius plot—see Figs. 2 and 7, and Section 3.7.
gases: best linear fits. Derived from the data of Way et al. [39,40]. f
Single value [59].
50 P. Molyneux / Journal of Membrane Science 320 (2008) 42–56

Fig. 3. Plot A—permeation activation energy EP versus molecular diameter  G for


Fig. 5. Plot C—log Y0 (entropic prefactor) versus permeation activation energy EP ,
gases through the PPG microporous silica hollow-fiber membranes. Key:  six gases
for permeant gases with the PPG microporous silica hollow-fiber membranes. Key:
as labeled – the straight line is least-squares fit for this set; ♦ CO2 ;  CO2 – inter-
 six gases as labeled – the straight line is least-squares fit for this set; ♦ CO2 ;  CO;
polated value, with the broken line indicating the energy anomaly EP ;  CO; 
: He – extrapolated (see Figs. 3 and 4). Derived from the data of Way et al. [39,40].
He—extrapolated EP value. The extrapolation of the linear fit to EP = 0 gives the esti-
mate of the (unperturbed) doorway diameter  D as labeled. Derived from the data
of Way et al. [39,40].
in that in the case of Plot B (Fig. 4) omitting this permeant as well as
CO2 gives a very good linear fit, with any deviations for the remain-
The molecular diameters values  G used here are the “criti- ing six permeants (H2 , Ar, O2 , N2 , CH4 and C2 H4 ) similar to those
cal constant” values  CC listed in the last column of Table 1 (see suggested by the Arrhenius plots (Table 2). This indicates that, at
Appendix B), as previously used with the organic polymers [7]. In least with this set, the log Y0 values are well defined; this indicates
each case the three plots are essentially linear, as previously seen in turn that the EP values are also well defined, and that any devia-
with the organic polymers [7]. The original papers [39,40] gave tions in the Plot A (Fig. 3) and Plot C (Fig. 5) from linear correlations
similar plots to the present Plot A, using the Breck values [52] for represent systematic effects. The good linear fit in Plot B (Fig. 4)
molecular diameter (see Table 1 and Appendix B), and also noted also indicates that the  G values for these six permeants are well
the essential linearity. defined, supporting the self-consistency of the “critical constant”
Again, as with the organic polymers [7], deviations are seen with values  CC used here and derived as discussed in Appendix B.4.
CO2 , as discussed in Section 3.6 below; however, closer examina- It is therefore convenient to divide the interpretation of the
tion of the data indicated that the behavior of CO is also anomalous, data into the primary effects, relating to the (average) straight lines
in the test plots, and the secondary effects, relating to systematic
deviations from these linear correlations.

3.5. Primary effects: H2 , Ar, O2 , N2 , CH4 , C2 H4

Considering the primary effects, the behavior of this group of


six gases relates mainly to the expansion of the SiO2 matrix, corre-
sponding to the energy contribution Eex of Eq. (14), the data from
the three linear Plots A–C results indicate, as discussed for the
organic polymers in Section 2.7, that the permeation of this set
of six gases involves a single fixed set of doorways with defined
diameter and defined spacing. The linear increase of EP and log Y0
results from the energy and entropy effects of inserting increas-
ingly large permeant molecules to expand the silica matrix at the
TS doorway (Fig. 1).4 Using Eqs. (21) and (22) gives the values of the
four characteristic parameters for the TS doorways listed in Table 3.
These values are similar to those seen with the organic poly-
mers [7], although in the latter case the thermodynamic parameters
in particular show a wide range of values because of the diver-
sity of polymers considered. As with the organic polymers, these
parameters must be considered average values.
Fig. 4. Plot B—log Y0 (entropic prefactor) versus molecular diameter  G for permeant It will be noted that these expansion effects are seen for perme-
gases with the PPG microporous silica hollow-fiber membranes. Key: ; six gases ants ranging in size and molecular complexity from H2 up to C2 H4 ,
as labeled – the straight line is least-squares fit for this set; ♦ CO2 ;  CO; : He –
extrapolated log Y0 . The extrapolation of the linear fit to EP = 0 is labeled with the
estimate of the (unperturbed) doorway diameter  D from Fig. 3. Derived from the
4
data of Way et al. [39,40]. See footnote 2.
P. Molyneux / Journal of Membrane Science 320 (2008) 42–56 51

Table 3 ically on the expansion contribution, Eex , as analyzed by Eqs. (21)


Parameters for the transition-site doorways in the permeation of gases through the
and (22). The secondary effects may be ascribed to the gas molecule
PPG samples of ␮SiO2 studied by Way et al. [39,40]a
in the transition-site doorway (Fig. 1C) interacting with the sur-
Name Symbol Value Units Figure Equation rounding silica matrix, presumably by polar interactions with the
Unperturbed diameter D 125(10) pm 3 (21) electrical distribution on the Si and O atoms of the matrix; this cor-
Force constant  0.43(7) nN 3 (21) responds to the contribution EMG of Eq. (14). Although no attempt
Entropy increment  1.7(2) pN K−1 4 (22) is made here to quantify these interactions, it is useful to note
Doorway spacing  30(4) nm 4 (22)
the degree of polarity for the various species involved. The com-
a
“Primary effect” data from the linear fits in the Figures specified for the six gases: mon factor here is the SiO2 matrix. There seems to have been little
Ar, H2 , N2 , O2 , CH4 , C2 H4 —see Section 3.5. mention on its charge distribution in discussions of its doorway
structure [20–25]; however, both Pauling electronegativity values
and including the intrinsically spherical (Ar) as well as the intrin- [60] and electron density measurements [61] give essentially +1.0
sically planar (C2 H4 ). This also shows the value of studying a wide unit charge on each Si atom and −0.5 unit charge on each O atom.
number and range of permeants, since it allows for a few to turn Carbon dioxide: The deviations with CO2 parallel those previ-
out to be markedly anomalous (here, He, CO and CO2 ) while still ously seen with organic polymers [7]; they correspond to the value
allowing sufficient remaining to define the main effects. of EP being lower than that expected by EP = 12 kJ mol−1 , but with
in this case also a marked entropy effect as shown by the apprecia-
3.6. Secondary effects ble (log Y0 ) value (Fig. 6). The effect seen so markedly with this
gas seems to be the result of the interaction of its large quadrupole
It is presumed in discussing these effects that the behavior all moment with the polar silica matrix as discussed; this quadrupole
the gases, including CO and CO2 as the most deviant cases, refers to moment has been modeled by a charge distribution with +0.66 unit
the same values of the average doorway spacing  and unperturbed charge on the C atom and −0.33 unit charge on each O atom [62].
doorway diameter  D that were estimated in the previous section Carbon monoxide: This gas is characterized by a marked
(Table 3). It is then convenient to quantify these secondary effects (log Y0 ) value (Fig. 6), essentially the same as that for CO2
in terms of the deviations in the two parameters EP and log Y0 from but a small if not zero value of EP . This anomaly also shows
the linear trends in Figs. 3 and 4, with the values of the respective itself in the Arrhenius plots (Fig. 2), where the plot for CO will
energy anomaly EP and entropy anomaly (log Y0 ) given by: clearly cross that for N2 at lower temperatures. The difference in
behavior of these two particular molecules is remarkable, since
EP = EP (cal) − EP (exp) (24) in their electronic structure and their critical constants they are
very similar and differ only in the presence of a small dipole
(log Y0 ) = (log Y0 )(cal) − (log Y0 )(exp) (25)
moment with CO (Table 1), which with the internuclear dis-
where “cal” indicates the value calculated from the trend Eqs. tance of 113 pm [49] corresponds to partial charges of only ±0.02
(21) and (22) for that value of  G , and “exp” indicates the exper- units on the respective atoms. Evidently the polarity of the SiO2
imental value (Table 2). These anomaly values are plotted against matrix again leads to a stronger effect than might have been
one another for display in Fig. 6. These effects may be discussed expected.
on the basis of the three contributions to EP shown in Eq. (14): Six gases: H2 , Ar, O2 , N2 , CH4 , C2 H4 . It will be seen from Fig. 6 that
EMM , Eex and EMG , but with the value of the first contribution these gases show appreciable and apparently significant values for
(matrix–matrix noncovalent interactions across the unoccupied the energy anomaly EP and with smaller and less significant val-
doorway) again neglected because of the width of the doorway ues of the entropy anomaly (log Y0 ). The values however seem to
(125 pm). The discussion of the “primary effects” focused specif- show no clear correlations with their molecular features (Table 1).
For example, C2 H4 has a similar quadrupole moment (albeit of the
opposite sign) to that of CO2 , but shows much lesser “secondary
effects” than the latter. Similarly, it is difficult to see why such three
different molecules as H2 , N2 and C2 H4 should be clustered together
(Fig. 6).
Summarizing, these secondary effects evidently arise because of
the much higher polarity of SiO2 than that of the organic polymers;
indeed, these effects may serve to probe the “fine structure” of the
TS doorways. However, although the deviations seen with CO2 are
understandable and parallel those seen with the organic polymers,
the picture for the other gases is much less clear. If the option exists
to extending these studies to other gases, the most useful would
be the other noble gases (Ne, Kr, Xe), since the simplicity of their
structure would greatly simplify the interpretation of the effects
observed.

3.7. Helium

As Fig. 2 shows, the Arrhenius test plot for He is markedly non-


linear (the curve drawn in the figure is an arbitrary quadratic best
fit). However, Fig. 7, which now includes the He data from the ear-
liest paper [38], shows that a simple plot of P (mol m m−2 s−1 Pa−1 )
Fig. 6. Entropy anomaly (log Y0 ) (Eq. (25)) versus energy anomaly EP , (Eq. (24)) versus T (K) gives an essentially linear dependence
for permeant gases with the PPG microporous silica hollow-fiber membranes. Key:
: six gases as labeled; ♦ CO2 ;  CO. Derived from the data of Way et al. [39,40]. 1013 P = 0.030(T − 295). (26)
52 P. Molyneux / Journal of Membrane Science 320 (2008) 42–56

type of type of PPG product [36,37], it is evident that there were


marked differences between the two materials used. The Shelekhin
fiber sample, for example, had a smaller outside diameter (32 ␮m)
and inside diameter (22 ␮m), although with the same nominal wall
thickness (L = 5 ␮m). Five gases were studied with this sample: He,
H2 , O2 , N2 , CH4 , CO2 ; the permeation of these was studied over
the temperature range 30–250 ◦ C, except for H2 that was studied
at only 30 ◦ C. Both sets of Arrhenius parameters were lower than
those for the Way samples (Table 2); the Shelekhin sample values
of log P0 were 3.9 units lower for O2 and N2 , and 4.6 units lower
for CH4 and CO2 ; for EP there were a fairly consistent lowering of
33(2) kJ mol−1 for all four of these gases. The lower values of EP
seem to be an extension of the reductions seen in going from the
compact vitreous silica vSiO2 to the microporous silica studied by
Way et al. (Table 2).
Their data were interpreted by the authors [42] on the basis of
the SDM. However, using the present TSM, the EP values gave a
rather scattered linear Type A plot (not shown) for the four gases:
He, O2 , N2 , and CH4 , where the downwards shift in EP corresponds
to a higher value of the doorway diameter  D = 240(20) pm (com-
pare with Fig. 4); the EP value for CO2 was again low, in this case by
about 11 kJ mol−1 .
It seems that the larger doorway size with the Shelekhin sample
Fig. 7. Permeability coefficient P versus temperature for He and H2 through the PPG compared the Way sample is the result of differences in the produc-
microporous silica hollow-fiber membranes—data of Way et al. [38–40].  He, Ref. tion histories of these samples. Comparative results of this kind on
[38]; ♦ He, Ref. [39]; +H2 [39,40]. The straight line is best fit to the He data; the curve suitably diverse sets of permeant molecules with membrane sam-
is the Arrhenius fit to the H2 data—compare with Fig. 2.
ples prepared under different controlled conditions promise to give
some clues as to the factors influencing the four TSM parameters
The same plot also shows the data for H2 as the next lightest gas, ( D , , , ) that determine permeation behavior.
with the best fit of the Arrhenius equation (Table 2).
This simple behavior of the temperature dependence of per-
3.9. Other microporous inorganic media
meation for He does not seem to have been observed previously
in the literature, where for compact vSiO2 the dependence is
The microporous silica discussed in Section 4 is only one type
essentially Arrhenius even down to these low temperatures [57].
in a wide range of such media whose gas permeability behav-
The parallel work of Shelekhin et al. [41] (see Section 3.8) also
ior is important in practice [1,11,14,34,52]. Two other important
gave an essentially linear Arrhenius plot for He with their sam-
types are the zeolites, with their open-network crystalline struc-
ple of the PPG ␮SiO2 , while their EP values correspond to an EP
ture, and where the crystallographically defined windows are the
value of about 30 kJ mol−1 for the present Way sample (compare
equivalent of the present doorways [1,9,11,12,14,52], and the micro-
with Figs. 3 and 5). If the He were escaping by Knudsen flow
porous carbon membranes, also referred to as carbon molecular
through microscopic pores, the temperature dependence would be
sieve membranes (CMSM) [11,13,34,47].
as (1/T)1/2 .
With these materials, insofar as the pore size and the perme-
At the same time, it is suspicious that the permeation flux
ant molecular diameter lead to activated migration and molecular
begins at a temperature (295 K) that is only a little below the
sieving, the TSM should again be applicable. As with the organic
starting experimental temperature (298 K = 25 ◦ C), as if heating the
polymers [7] and the present microporous silica (Table 1), this again
equipment were opening some apertures in the membrane sys-
requires experimental permeation data over a wide temperature
tem through thermal expansion. However, the effect seems to
range with a set of permeants having a spectrum of molecular diam-
occur both with the data in first paper [38] and those in the third
eters to probe the doorways and define their characteristics; the use
paper [40], where different samples of the hollow-fiber membrane
of nine permeants in the present case seems exemplary. Once again
were used (Fig. 6). It is also curious that with rise in temper-
the TSM obviates the need, inherent with the SDM in experimental
ature the permeation rate for H2 eventually overtakes that of
and in modelling studies, to deal separately with the sorption and
He (Figs. 2 and 6); any anomalous permeation mechanism for
the diffusion processes.
He would be expected to be additional to the normal Arrhe-
nius rate, which should be should be markedly higher than that
of H2 . 4. Conclusions
This simple but curious behavior of He therefore remains unex-
plained. • In this paper, the previously presented transition-site model
(TSM) as applied to organic polymers has been re-presented
3.8. Permeation data of Shelekhin et al. on the PPG microporous briefly with some changes in detail.
silica hollow-fiber membranes • The conventional and widely accepted sorption-diffusion model
(SDM) used to describe permeation through solids has now been
It has already been noted, in Section 3.2, that parallel permeation reviewed critically; one key feature of the present paper is the
studies on the PPG microporous silica hollow-fiber membranes observation that there is a fundamental point of weakness of the
have also been carried out by Shelekhin et al. [41,42]. Although SDM, in that the two analyzed factors, “sorption” and “diffusion”,
these data, and those of Way et al. [38–40], already considered are not independent, since both contain the characteristics of the
in some detail, were obtained on what was nominally the same sorption state. Other points noted are that the sorption sites must
P. Molyneux / Journal of Membrane Science 320 (2008) 42–56 53

be heterogeneous in character, and that the pressure differential cient or lattice parameter for this gas – Eqs. (16)–(18) – then the only
and not the concentration gradient is the true driving force for common factor is the gas itself. This was therefore dealt with prag-
the migration. matically by using the lower value of 256 J K−1 mol−1 for this gas,
• By contrast, the strength of the TSM lies in its focus on the which removed the anomaly for this gas in all the cases involved.
transition-sites that it turns out are already accepted as doorways This has also been done in the present case (Table 1), and leads to
for migration in nonporous inorganic glasses such as vitreous a good fit in the Plot B (Fig. 4) and Plot C (Fig. 5), as well as in Plot A
silica. (Fig. 3) where this factor is not involved.
• The TSM allows analysis of the linear test plots to estimate the
four characteristic quantities: the intrinsic diameter of the door- Appendix B. Molecular diameters of permeant gas
way,  D ; the thermodynamic parameters for the expansion of molecules,  G
the doorway,  (force constant) and  (entropy increment); and
the average spacing between neighboring doorways , which is B.1. Importance of molecular diameter values
equivalent to the “jump distance” of the SDM.
• The TSM has the advantage over the SDM that if decouples the In the evaluation and interpretation of the migration of guest
migration process from the sorption equilibrium. Strictly, this molecules in solids from the molecular viewpoint, it is evidently
only applies in the limit of low pressures and low amounts of necessary to have a self-consistent set of molecular diameters for
sorption; the actual amount of sorption only comes in as a per- the gases used; however, there seems to have been little specific
turbation effect, that is, of the sorbed molecules on the properties or critical examination of these diameter values used in this area
and behavior of the doorways. [2,3,7]. This requirement is particularly important for the accu-
• The present work has shown that the TSM applies to one particu- rate estimation of the diameter of the unperturbed doorway  D ,
lar type of microporous silica hollow-fiber membrane produced of the thermodynamic parameters  and  for the expansion of the
by PPG Industries, and that it gives values for the four character- transition-site doorway, and of the inter-site spacing . For each of
istic parameters that are similar to those obtained with organic these depends on the good linear definition of the three test plots
polymers. There are also some “secondary effects” that are inter- (see Figs. 3–5).
preted as host-guest interactions at the doorway. In quoting these values, it is found clearer to use picometers
• These results suggest that the TSM should be applicable more (1 pm = 0.01 Å = 0.001 nm) since these shows up better the effect of
widely to inorganic membranes showing activated permeation small changes on the permeation behavior.
by molecular sieving, including other types of microporous solids There are three main sets of values that have been applied in
as well as zeolites and other crystalline solids. the area of permeation through membranes, as considered below
in Sections B.3 and B.4. However, it is necessary firstly to consider
the effect of rotation on the effective molecular diameter.
Appendix A. Absolute molar entropy values, S0
B.2. Rotation of molecules in relation to molecular diameters
The values of the absolute molar entropy, S0 , for the gases play an
important part in the TSM, in correcting for their different entropy In considering the specific dimensions of molecules, and in
levels in the gas state on either side of the membrane, and putting particular how they pack together in the solid-state at low tem-
these on the same basis as the energy values which are all on the peratures, it is evidently necessary to consider the detailed shapes.
same (zero) level in the gas state. There are two points to be noted Commonly, this can be summarized as spherical, rod-shaped, pla-
in connection with these values listed in Table 1. nar, etc. Frequently, in considering the migration properties of
The first point relates to the matter of the pressure units that molecules in solids, these shapes are referred to in order to clar-
are used as the reference level in defining S0 , since this must be the ify or interpret the behavior; for example, a rod-shaped molecule
same as the pressure units used for the permeability coefficient; is commonly presumed to present its narrowest dimension to the
this reference level relates to the entropy change that takes place barrier to ease its passage. However, this ignores the significance of
when the solid (crystalline) form of the substance (with entropy rotation in the behavior of the molecule. At a high enough tem-
essentially independent of pressure) is converted to the gas (vapor) perature, a molecule will have an amount kB T/2 in each axis of
state. In a previous publication [63] this reference level has been rotation; in macroscopic terms, this energy amounts at 25 ◦ C to
considered as a form of “hidden” unit; an alternative method would about 1 kJ mol−1 . Although rotation is suppressed for all molecules
be to simply use permeability coefficients that are referred to the at sufficiently low temperatures, as the temperature is raised quan-
same standard pressure unit of the absolute entropy value, although tum effects lead to the onset of rotation. This is shown by the values
this is not such a transparent way of working. In either case, the con- of the heat capacity at constant pressure Cp , which in the general
version from the literature values of S0 [50] with the bar (=105 Pa) case rises with temperature as firstly translation, then rotation and
as the standard state pressure, to the pressure unit Pa used in the then vibration modes come into play [64]; the literature data for Cp
permeability coefficient, requires the addition of R ln(105 ), that is, show that this rotational freedom is essentially complete at 298 K
96 J K−1 mol−1 ; the values so adjusted are listed in Table 1, and used for all the molecules considered here [65]. The angular velocity ω
in the calculation of the entropic prefactor Y0 using Eq. (12). will be given by the standard mechanics formula relating it to the
The second point relates specifically to the value of S0 to be used moment of inertia I of the molecule about that axis
for O2 . In the case of the organic polymers [7], it was found that
using the value of 301 J K−1 mol−1 obtained by applying the above kB T Iω2
= . (B.1)
pressure-unit adjustment, with all the polymers where O2 had been 2 2
studied alongside a number of other permeant gases, in each case For molecules of the type used here as permeants
it gave the same anomaly with the Plot B and the Plot C, but no (Tables 1 and 2), this gives angular velocities ω in the range
anomaly with the Plot A; this indicates that the anomaly relates of 1012 to 1013 s−1 [64]. Thus if a specific orientation is envisaged
to a high value of the parameter log Y0 (entropic prefactor). Since for the permeant molecule in the doorway, this very rapid rotation
this occurred in a consistent way with all of the polymers involved, must be stopped. At the same time, the orientation involves an
and unless the anomaly relates to a different value of the rate coeffi- entropy penalty, related to the degree of restriction envisaged,
54 P. Molyneux / Journal of Membrane Science 320 (2008) 42–56

since there must still some finite degree of motion (oscillation) separation r
in the captured state. For example, an allowed oscillation of ±10◦  12   6
LJ LJ
about the long axis in all directions reduces the directional freedom u(r) = 4 εLJ − (B.2)
by a factor of about 100, that is, only about 1% of the molecules r r
would be oriented in this way ready for insertion; from the
where εLJ is the depth of the potential well and  LJ is the LJ
thermodynamic viewpoint, this would correspond to an entropy
molecular diameter [2]. More exactly, this equation only applies to
penalty S of about 40 J K mol−1 , leading at 298 K to an free energy
nonpolar molecules (especially the noble gases); with polar gases
penalty (TS) of about 12 kJ mol−1 . More stringent limitations on
there would be a third term for the polar (electrostatic) contribu-
the freedom on the molecule lead to a correspondingly smaller
tions, leading to the so-called Stockmayer equation. Table 1 shows
fraction of molecules with the correct orientation, and paralleled
the values of the LJ diameters from the literature [53–56]. The
by correspondingly greater thermodynamic penalties; the latter
values originate from one of three sources of experimental data:
would also be greater still at higher temperatures. These consider-
gas viscosity, thermal conductivity, or equation of state (second
ations strongly suggest that, in the first instance, molecules of all
virial coefficient). The problem with all these methods is that they
kind should be considered as rotating essentially freely, that is, to
may only give a combination of the two parameters εLJ and  LJ .
be spherically symmetrical, both in the gaseous state and when
This shows itself, when considering LJ parameters from different
inserted in the transition-site doorway, and that it is the diameter
sources, in a “high” value of εLJ being associated with a “low” value
of this sphere of rotation that determines the effective molecular
of  LJ , and vice versa. The data given in Table 1 have been evalu-
diameter.
ated in the light of these considerations, eliminating some values
that were evidently anomalous. Even with this procedure, as shown
B.3. Molecular diameter values of Breck
in Table 1 the uncertainty in these values is still about 6 pm. In
addition, any polarity of the molecules makes the application of
One widely used approach in permeation work is to use the set
the simple form of Eq. (B.2) less certain. In the case of CO2 in par-
of values listed by Breck [52] in connection with zeolite molecular
ticular, there are indications that the LJ parameters are markedly
sieves, which has been also adopted for use in the field of membrane
temperature dependent [53]. There is also the general limitation
processes in general [1], and more specifically in the publications
that these parameters require the availability of experimental data
drawn upon in this paper [38–40]; indeed, these values seem to be
from one of the three rather troublesome experimental methods
viewed as definitive. The original values [52] were drawn from a
already specified.
diversity of sources, including Lennard–Jones parameters (Section
B.4) in a rather inconsistent fashion; being put in Angstrom (Å)
B.5. Molecular diameters from critical constants via the van der
units, it is not clear what precision is being claimed: “3.9 Å” may
Waals equation
imply either ±1 or ±5 on the derived value of 390 pm.
In addition, in the cases of the isoelectronic pair, CO2 and N2 O,
A third approach is to use parameters derived from the co-
the values have been taken to be those of the smallest diame-
volume b of the van der Waals equation, as in the previous paper [7].
ter, leading to quoted values of 0.33 Å, that is 330 pm, giving the
The rationale here is that the parameter b is presumed to be equal
anomaly that they have a lower value than that for patently smaller
to four times the volume of the molecules, leading to the relation:
molecules such as CO, N2 and O2 ; this anomalous value has appar-
ently been accepted by the membrane community because it fits  3b 1/3
in with this being a “fast” gas in permeation, although the present = . (B.3)
2NA
treatment has shown that this is an energy effect (low EP ) rather
than a molecular diameter effect. Further evidence on this point With values of b available from the literature [66], this enables 
is given by the electric polarisability ˛e , which is another mea- to be determined for these same molecules. However, appearances
sure of molecular volume; the value for CO2 is some 50% greater are deceptive since it turns out that the listed values of b [66] are
than those for these same diatomic molecules (Table 1). Indeed, as not from a van der Waals fit, but are derived from critical constants
already noted (Section B.2), all these molecules must be taken to using the relation
be rotating freely in the gas state, and would tend to maintain this RTc
rotation even when held in the guest matrix, so that the diameter to b= (B.4)
8pc
be assigned to these triatomic molecules is the largest dimension.
Furthermore, it is not clear whether or not this “adjustment” for where Tc is the critical temperature and pc is the critical pressure;
CO2 and N2 O has also been applied to other asymmetric molecules it is somewhat odd that it is not the critical molar volume Vc that
(C2 H4 , C6 H6 , etc.). is used, but presumably the logic is that the value of Vc is in many
A similar anomaly is evident for the pair: H2 (289 pm) versus cases either lacking or not well defined in the literature [51]. How-
H2 O (265 pm) [52]. This anomaly is again highlighted by the fact ever, this has the advantage that the critical constants Tc and pc are
that the volume-related molar polarizability ˛ of H2 O is nearly known for a very wide range of substances [51], while unlike the LJ
twice that of H2 [48], whereas the converse situation should occur values (Section B.3) there are no ambiguities from any polarity of
if the quoted Breck values apply. the molecules. This leads finally to:
It is also not clear whether this set of parameters, which has  3RT 1/3
c
been selected specifically to apply to the zeolites, is necessarily CC = (B.5)
16NA pc
going to apply to such diverse other media as (for example) organic
polymers, silica, and microporous carbon. where the subscript on  CC is used to indicate specifically the source
of the values. Judging from the data for the present substances
B.4. Lennard–Jones diameters (Table 1) supplemented by that for the other noble gases, these  CC
values are fairly consistently 15% less the corresponding  LJ values.
An alternative approach is to use the parameters derived from Some ambiguities arise here with the “light” gases, notably He
the Lennard–Jones (LJ) 6–12 potential for the intermolecular poten- and H2 (and their isotopes) and to a much lesser extent Ne, where
tial energy u(r) of a pair of identical molecules at internuclear quantal (quantum mechanical) effects come into play. In the case
P. Molyneux / Journal of Membrane Science 320 (2008) 42–56 55

of the noble gases (He, Ne, Ar, Kr, Xe), this is evidently cause of
the anomaly, that Ne has a smaller derived value of  CC than He,
S‡ molar entropy change for insertion of G into a TS
although the values thereafter increase smoothly for Ne–Xe. How-
doorway (J K−1 mol−1 )
ever, these directly calculated values were found to be satisfactory
T absolute temperature (K)
in the previous application of the TSM model to organic polymers
Tc critical temperature (K)
[7], while quantal effects are themselves present in the perme-
TS transition-site (doorway)
ation of gases through amorphous media, for example with organic
TSM transition-site model for permeation
polymers [10]. In the present case the  CC values for He and H2
U‡ molar internal energy change for insertion of G into
given in Table 1 have been adjusted by using the linear relation
a TS doorway (J mol−1 )
that is observed between  CC and ˛1/3 , where ␣ is the volume-
Vc critical molar volume (m3 mol−1 )
dimensioned molar polarizability as also listed in Table 1.
x general membrane depth in the direction of perme-
In the present work with microporous silica, it is these  CC val-
ation (m)
ues, as listed in the last column of Table 1, that have been used as
Y entropic coefficient Eq. (18) (mol m m−2 s−1 )
the molecular diameter  G in the test plots of Figs. 3 and 4, and in
Y0 entropic prefactor Eq. (19) (mol m m−2 s−1 )
the calculation of the derived parameters for the doorways listed
(log Y0 ) entropy anomaly (Eq. (25))
in Table 3.
 times-or-divided-by

Greek letters
˛e molar electric polarisability (m3 mol−1 )
Nomenclature εLJ Lennard–Jones energy parameter (J)
 force constant for expansion of a TS doorway (N)
b van der Waals co-volume (m3 mol−1 )
 (average) spacing between neighboring TS door-
D diffusion coefficient (m2 s−1 )
ways (m)
D location of a doorway (transition-site)
 entropy-change coefficient for expansion of a TS
D0 diffusion prefactor (m2 s−1 )
doorway (N K−1 )
Eex energy to expand the TS doorway by the inserted G
 molecular (kinetic, collision) diameter (m)
molecule (J mol−1 )
D diameter of an unoccupied TS doorway (m)
ED activation energy for diffusion (J mol−1 )
G molecular diameter of the permeant gas G (m)
EP activation energy for permeation (J mol−1 )
 CC value of  estimated from critical constants
EMG energy of noncovalent interaction between the
(Appendix B) (m)
inserted G molecule and the TS doorway matrix
 LJ Lennard–Jones molecular diameter (Appendix B)
(J mol−1 )
(m)
EMM energy of noncovalent interactions across the unoc-
cupied TS doorway (J mol−1 )
EP energy anomaly Eq. (24) (kJ mol−1 )
G bulk shear modulus (Pa) References
G permeant gas molecule
h Planck constant (6.626 × 10−34 J s) [1] J.D. Henry, et al., Gas-separation membranes, in: R.H. Perry, D.W. Green (Eds.),
Perry’s Chemical Engineers’ Handbook, McGraw-Hill, New York, 1997, pp. 22-
HS isosteric sorption enthalpy change (kJ mol−1 ) 61–22-67.
kB Boltzmann constant (1.381 × 10−23 J K−1 ) [2] Yu. Yampolski, B.D. Freeman, I. Pinnau (Eds.), Materials Science of Membranes
k‡ transition-state rate constant (s−1 ) for Gas and Vapor Separation, Wiley, New York, 2006.
[3] R.H. Doremus, Diffusion of Reactive Molecules in Solids and Melts, Wiley, New
L membrane thickness (m) York, 2002.
nd no data/not determined [4] C.P. Flynn, Point Defects and Diffusion, Clarendon Press, Oxford, 1972.
NA Avogadro constant (6.022 × 1023 molecules mol−1 ) [5] J.E. Shelby, Handbook of Gas Diffusion in Solids and Melts, ASM International,
Materials Park, OH, 1996.
p pressure (Pa) [6] P. Shewmon, Diffusion in Solids, The Minerals, Metals & Materials Society (TMS),
pc critical pressure (Pa) Warrendale, PA, 1989.
pe electric dipole moment (C m) [7] P. Molyneux, “Transition-site” model for the permeation of gases and vapors
through compact films of polymers, J. Appl. Polym. Sci. 79 (2001) 981–1024.
P permeability coefficient (mol m m−2 s−1 Pa−1 ) [8] R.M. Barrer, Activated diffusion in membranes, Trans. Faraday Soc. 35 (1939)
P0 permeation prefactor (mol m m−2 s−1 Pa−1 ) 644–656.
PE poly(ethylene) [9] R.M. Barrer, Porous crystal membranes, J. Chem. Soc., Faraday Trans. 86 (1990)
1123–1130.
PEMA poly(ethyl methacrylate) [10] H.L. Frisch, C.E. Rogers, Quantum isotope effect in permeation, J. Chem. Phys.
p pressure difference across the membrane (Pa) 40 (1964) 2293–2298.
Q permeance (mol m−2 s−1 Pa−1 ) [11] A. Singh, W.J. Koros, Significance of entropic selectivity for advanced gas sepa-
ration membranes, Ind. Eng. Chem. Res. 35 (1996) 1231–1234.
Qe electric quadrupole moment (C m2 )
[12] A.J. Burggraaf, Single gas permeation of thin zeolite (MFI) membranes: theory
R gas constant (8.3145 J K−1 mol−1 ) and analysis of experimental observations, J. Membr. Sci. 155 (1999) 45–65.
S sorption (solubility) coefficient (mol m−3 Pa−1 ) [13] J. Gilron, A. Soffer, Knudsen diffusion in microporous carbon membranes with
SDM sorption-diffusion model for permeation molecular sieving character, J. Membr. Sci. 209 (2002) 339–352.
[14] P. Bahukudumbi, D.M. Ford, Molecular modeling study of the permeability-
S(D) entropy of 1 mol of unoccupied TS doorways selectivity trade-off in polymeric and microporous membranes, Ind. Eng. Chem.
(J K−1 mol−1 ) Res. 45 (2006) 5640–5648.
S(GD) entropy of 1 mol of TS doorways occupied by the gas [15] V. Stannett, J.L. Williams, The permeability of poly(ethyl methacrylate) to gases
and water vapor, J. Polym. Sci. C3 (1965) 45–59.
G (J K−1 mol−1 ) [16] B. Wilks, M.E. Rezac, Properties of rubbery polymers for the recovery of hydro-
S0 sorption prefactor (mol m−3 Pa−1 ) gen sulfide from gasification gases, J. Appl. Polym. Sci. 85 (2002) 2436–2444.
S0 (G) absolute molar entropy of gas G (reference pressure [17] (a) R. Kirchheim, Partial molar volume of small molecules in glassy polymers,
J. Polym. Sci. Part B: Polym. Phys. 31 (1993) 1373–1382;
1 Pa) (J K−1 mol−1 <Pa>) (b) R. Kirchheim, Solubility of gases in vitreous silica described by a distribution
of dissolution energies, J. Am. Ceram. Soc. 84 (2001) 2699–2701.
56 P. Molyneux / Journal of Membrane Science 320 (2008) 42–56

[18] J. Rundgren, Q. Dong, G. Hultquist, Concentration-dependent diffusion of [42] A.B. Shelekhin, A.G. Dixon, Y.H. Ma, Theory of gas diffusion and permeation in
hydrogen in vitreous silica, J. Appl. Phys. 100 (2006) 104902-1–104902-5. inorganic molecular-sieve membranes, AIChE J. 41 (1995) 58–67.
[19] D.W. McCall, D.C. Douglass, L.L. Blyler, G.E. Johnson, L.W. Jelinski, H.E. Blair, [43] R.M. de Vos, H. Verweij, Improved performance of silica membranes for gas
Solubility and diffusion of water in low-density polyethylene, Macromolecules separation, J. Membr. Sci. 143 (1998) 37–51.
17 (1984) 1644–1649. [44] R.M. de Vos, H. Verweij, High-selectivity, high-flux silica membranes for gas
[20] O.L. Anderson, D.A. Stuart, Calculation of activation energy of ionic conductivity separation, Science 279 (1998) 1710–1711.
in silica glasses by classical methods, J. Am. Ceram. Soc. 37 (1954) 573–580. [45] M.C. Duke, J.C. Diniz da Costa, G.Q. Lu, P.G. Gray, Modeling hydrogen separa-
[21] D.K. McElfresh, D.G. Howitt, Activation enthalpy for diffusion in glass, J. Am. tion in high temperature silica membranes systems, AIChE J. 52 (2006) 1729–
Ceram. Soc. 69 (1986) C-237–C-238. 1735.
[22] S.L. Chan, S.R. Elliott, Interstices, diffusion doorways and free volume percola- [46] J.M.D. MacElroy, S.P. Friedman, N.A. Seaton, On the origin of transport
tion studies of a dense random packed atomic structure, J. Non-Cryst. Solids resistances within carbon molecular sieves, Chem. Eng. Sci. 54 (1999)
124 (1990) 23–33. 1015–1027.
[23] D.K. McElfresh, D.G. Howitt, A structure based model for diffusion in glass and [47] T.A. Peters, J. Fontalvo, M.A.G. Vorstman, N.E. Benes, R.A. van Dam, Z.A.E.P.
the determination of diffusion constants in silica, J. Non-Cryst. Solids 124 (1990) Vroon, E.L.J. van Soest-Vercammen, J.T.F. Keurentjes, Hollow fibre microporous
174–180. silica membranes for gas separation and pervaporation: synthesis, perfor-
[24] S.L. Chan, S.R. Elliott, Theoretical study of the interstice statistics of the oxygen mance and stability, J. Membr. Sci. 248 (2005) 73–80.
sublattice in vitreous SiO2 , Phys. Rev. B 43 (1991) 4423–4432. [48] J.H. Hildebrand, R.L. Scott, Intermolecular forces, in: The Solubility of Nonelec-
[25] S.L. Chan, S.R. Elliott, Calculation of diffusion activation energies in cova- trolytes, Dover, New York, 1964, pp. 48–61.
lent solids: application to vitreous silica, J. Phys. Condens. Matter 3 (1991) [49] R.D. Johnson III (Ed.), NIST Computational Chemistry Comparison and Bench-
1269–1280. mark Database [CCCBDB], NIST Standard Reference Database Number 101,
[26] P.W. Atkins, Activated complex theory, in: Physical Chemistry, Oxford Univer- Release August 12, 2005, http://srdata.nist.gov/cccbdb.
sity Press, 1978, pp. 906–916. [50] Standard thermodynamic properties of chemical substances, in: D.R. Lide, H.P.R.
[27] P.C. Jordan, Ion permeation and kinetics, J. Gen. Physiol. 114 (1999) 601–604. Frederickse (Eds.), CRC Handbook of Chemistry and Physics, CRC Press, Boca
[28] V.L. Schramm, Enzymatic transition states and enzyme transition state analog Raton, FL, 1996–1997, pp. 5-4–5-60.
design, Annu. Rev. Biochem. 67 (1998) 693–720. [51] R.C. Reid, J.M. Prausnitz, B.E. Poling, Property data bank, in: The Properties of
[29] A.A. Gray-Wheale, R.H. Henchman, R.G. Gilbert, M.L. Greenfield, D.N. Gases and Liquids, McGraw-Hill, New York, 1986, pp. 656–732.
Theodorou, Transition-state theory model for the diffusion coefficients of small [52] D.W. Breck, The molecular sieve effect, in: Zeolite Molecular Sieves: Structure,
penetrants in glassy polymers, Macromolecules 30 (1997) 7296–7306. Chemistry, and Use, Wiley, New York, 1974, pp. 633–637.
[30] S. Pauly, Permeability and diffusion data, in: E.H. Brandrup, E.H. Immergut [53] S.V. Churakov, M. Gottschalk, Perturbation theory based equation of state for
(Eds.), Polymer Handbook, Wiley, New York, 1999, pp. VI/543–VI/569. polar molecular fluids. Part I. Pure fluids, Geochim. Cosmochim. Acta 67 (2003)
[31] P. Molyneux, unpublished work. 2397–2414.
[32] S.G. Kazarian, M.F. Vincent, F.V. Bright, C.L. Liotta, C.A. Eckert, Specific inter- [54] J.O. Hirschfelder, C.F. Curtiss, R.B. Bird, Force constants for the Lennard–Jones
molecular interactions of carbon dioxide with polymers, J. Am. Chem. Soc. 118 (6–12) potential, in: Molecular Theory of Gases and Liquids, Wiley, New York,
(1996) 1729–1736. 1954, pp. 1110–1113.
[33] S.P. Nalawade, F. Picchioni, J.H. Marsman, D.W. Grijpma, J. Feijen, L.P.B.M. [55] F.M. Mourits, F.H.A. Rummens, A critical evaluation of Lennard–Jones and Stock-
Janssen, Intermolecular interactions between carbon dioxide and the carbonyl mayer potential parameters and some correlation methods, Can. J. Chem. 55
groups of polylactides and poly(␧-caprolcatone), J. Control. Release 116 (2006) (1977) 3007–3020.
e38–e40. [56] R.A. Svehla, Estimated Viscosities and Thermal Conductivities of Gases at High
[34] R.S.A. de Lange, K. Keizer, A.J. Burggraaf, Analysis and theory of gas transport Temperatures, Technical Report R-132, National Aeronautics and Space Admin-
in microporous sol–gel derived ceramic membranes, J. Membr. Sci. 104 (1995) istration, Washington, DC, 1962.
81–100. [57] J.E. Shelby, Permeation and diffusion in inorganic glasses, in: Handbook of Gas
[35] R.H. Doremus, Various types of fused silica, in: Diffusion of Reactive Molecules Diffusion in Solids and Melts, ASM International, Materials Park, OH, 1996, pp.
in Solids and Melts, Wiley, New York, 2002, p. 75. 15–53.
[36] J.J. Hammel, W.J. Robertson, W.P. Marshall, H.W. Barch, B. Daas, M.A. Smoot, R.P. [58] W.G. Perkins, D.R. Begeal, Diffusion and permeation of He, Ne, Ar, Kr, and D2
Beaver/PPG Industries, Inc., Process of gas enrichment with porous siliceous- through silicon oxide thin films, J. Chem. Phys. 54 (1971) 1683–1694.
containing material, U.S. Patent 4,842,620 (June 27, 1989). [59] F.J. Norton, Permeation of gaseous oxygen through vitreous silica, Nature 191
[37] J.J. Hammel/PPG Industries, Inc., Porous inorganic siliceous-containing gas (1961) 701.
enriching material and process of manufacture and use, U.S. Patent 4,853,001 [60] L. Pauling, The nature of silicon–oxygen bonds, Am. Miner. 65 (1980) 321–323.
(August 1, 1989). [61] R.F. Stewart, M.A. Whitehead, G. Donnay, The ionicity of the Si–O bond in low-
[38] J.D. Way, D.L. Roberts, Hollow fiber inorganic membranes for gas separations, quartz, Am. Miner. 65 (1980) 324–326.
Sep. Sci. Technol. 27 (1992) 29–41. [62] J.G. Harris, K.H. Yung, Carbon dioxide’s liquid–vapor coexistence curve and crit-
[39] J.D. Way, A mechanistic study of molecular sieving inorganic membranes for ical properties as predicted by a simple molecular model, J. Phys. Chem. 99
gas separations, Final Report submitted to U.S. Department of Energy under (1995) 12021–12024.
contract DE-FG06-92-ER14290, Colorado School of Mines, Golden, CO, 1993, [63] P. Molyneux, The dimensions of logarithmic quantities, J. Chem. Educ. 68 (1991)
www.osti.gov/bridge/servlets/purl/10118702-ZAx4Au/native/1011872.pdf. 467–469.
[40] M.H. Hassan, J.D. Way, P.M. Thoen, A.C. Dillon, Single component and mixed gas [64] P.W. Atkins, Physical Chemistry, Oxford University Press, 1978.
transport in silica hollow fiber membrane, J. Membr. Sci. 104 (1995) 27–42. [65] G. Aylward, T. Findlay, SI Chemical Data, Wiley, Brisbane, 1998.
[41] A.B. Shelekhin, A.G. Dixon, Y.H. Ma, Adsorption, permeation, and diffusion of [66] van der Waals constants for gases, in: D.R. Lide, H.P.R. Frederickse (Eds.), CRC
gases in microporous membranes. Part II. Permeation of gases in microporous Handbook of Chemistry and Physics, CRC Press, Boca Raton, FL, 1996–1997, pp.
gas membranes, J. Membr. Sci. 75 (1992) 233–244. 6-47–6-51.

Potrebbero piacerti anche