Membrane Gas Separation

2.4
Gas Permeation 
It is well known that the gas permeation properties of glassy polymers tend to show 
‘ trade - off ’ behaviour, with highly selective polymers exhibiting low permeability and vice 
versa. This is commonly represented in terms of a double logarithmic Robeson [23,24]
plot of selectivity against permeability. Initial gas permeability measurements on PIM - 1 
membranes demonstrated that the polymer exhibits a high permeability (e.g. CO 
2
perme-
ability P (CO 
2
) = 2300 Barrer) [16] , exceeded only by ultra - high free volume polymers 
such as poly(1 - trimethylsilyl - 1 - propyne) (PTMSP) [25,26] . Of particular interest, however, 
is that this relatively high permeability is accompanied by a selectivity above Robeson ’ s 
1991 upper bound [23] for a number of important gas pairs, including O 
2
/N 
2
and CO 
2
/
CH 
4
. PIM - 1 helps to defi ne the recently revised upper bounds [24] for these gas pairs. 
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0
0.2
0.4
0.6
0.8
1
1.2
1.4
Pore width / nm
dV/dw / (cm
3
 g
-1
 nm
-1
)
Figure 2.3 Apparent distributions of pore size for PIM - 1 derived from N 
2
adsorption at 
77 K ( 
•
 ), CO 
2
adsorption at 273 K ( 
×
 ) and from atomistic computer simulation [22] (bars)

---

34
Membrane Gas Separation
Subsequent work has shown that even higher gas permeabilities can be achieved for 
PIM - 1, depending on the history of the sample, with little loss of selectivity [5,27] . 
Comparative gas permeation experiments have been carried out using two different tech-
niques in two different laboratories, using the same batch of PIM - 1 [5] . In Germany, at the 
GKSS Research Centre Geesthacht GmbH (GKSS), a pressure increase time - lag method 
was used, operating with typical feed pressures of 200 – 300 mbar. This enabled calculation 
of diffusion coeffi cient, D , as well as permeability coeffi cient, P . In Moscow, at the A. V. 
Topchiev Institute of Petrochemical Synthesis of the Russian Academy of Sciences (TIPS), 
a gas chromatographic (GC) method was used. Pure penetrant gas at a pressure of 1 atm 
was passed through the upstream part of a permeation cell, while a carrier gas (He or Ar) 
was passed through the downstream part. GC was used to analyse the permeate and the 
fl ow of permeate was measured with a soap bubble fl owmeter. In order to check that 
the two methods of measuring permeability gave comparable results, despite differences 
in the boundary conditions, a standard polymer, poly(vinyltrimethylsilane), was studied in 
both laboratories. Good agreement was obtained, as can be seen in Table 2.1 .
A study was undertaken of the effects of membrane preparation protocol on the perme-
ability of PIM - 1. Membranes were cast from solutions of the polymer in tetrahydrofuran 
(THF) or chloroform onto a cellophane, glass or Tefl on surface. Three ‘ states ’ of PIM - 1 
were identifi ed. 
1 ‘ Water - treated ’ PIM - 1, for which a fl ow of water was used to assist removal of the 
membrane from the surface onto which it was cast.
2 PIM - 1 for which no water was used in membrane preparation, or else that was exhaus-
tively dried at elevated temperature under vacuum.
3 ‘ Methanol - treated ’ PIM - 1, that was soaked in methanol for at least a day, to fl ush out 
any residual casting solvent and allow relaxation of chains in the swollen state, then 
dried under vacuum to constant weight.
Representative permeability data from both methods of measurement are given in Table 
2.2 
for PIM 
- 
1 in each of these three states. It can be seen that contact with water 
dur ing membrane preparation can lead to a signifi cant reduction in permeability. This is 
caused by trapped water inside the micropores and is hard to remove by vacuum and 
elevated temperature. Methanol treatment, however, substantially enhances the permea-
bility. Indeed, methanol - treated PIM - 1 is amongst the most permeable known polymers. 
Furthermore, the selectivities for some important gas pairs are signifi cantly higher than 
Table 2.1  Permeability coeffi cients for poly(vinyltrimethylsilane) determined by the 
pressure increase method (GKSS, measurements made by Sergey Shishatskiy) and by the 
gas chromatographic method (TIPS) 
Gas
P / Barrer
GKSS
TIPS
H 
2
215
220
He
178
160
O 
2
45
44
N 
2
11
11

---

Gas Permeation Parameters and Other Physicochemical Properties
35
Table 2.2  Permeability coeffi cients at 30 ° C for PIM - 1 in various states (1 = water - treated; 
2 = water - free; 3 = methanol - treated) determined by the pressure increase method (GKSS) 
and by the gas chromatographic method (TIPS) 
State
P (CO 
2
) / Barrer
P (O 
2
) / Barrer
P (N 
2
) / Barrer
GKSS
TIPS
GKSS
TIPS
GKSS
TIPS
1
950
1540
128
157
51
49
2
3700
4350
530
590
155
190
3
11 200
12 600 *
1530
1610 *
610
500 *
* Temperature = 23 ° C.
achieved for other polymers in the permeability range, as can be seen in the Robeson 
plots of Figure 2.4 .
Many high permeability polymers exhibit a reduction in permeability over time (ageing) 
[28] . This effect was investigated for a methanol - treated (state 3) PIM - 1 membrane. The 
sample was kept at ambient temperature in air and permeabilities measured at intervals 
by the gas chromatographic method. Results are shown in Figure 2.5 . A reduction in 
permeability is seen, although changes are small after the fi rst few days of measurement. 
Over 45 days of observation, P (O 
2
) decreased by about 23%, with an accompanying 
increase in permselectivity P (O 
2
)/ P (N 
2
) from 3.3 to 4.2. The effects of ageing are less 
marked than have been observed for other high free volume polymers. For example, 
uncross - linked poly(4 - methyl - 2 - pentyne) showed a decrease in N 
2
permeability from 
about 1000 Barrer to 650 Barrer within 45 days [29] .
The temperature dependence of the permeability of PIM - 1 membranes was investigated 
in order to evaluate activation energies of permeation, E
p
, for various gases. Values of E
p
derived from pressure increase (barometric) and gas chromatographic methods for a 
water - free (state 2) PIM - 1 membrane are given in Table 2.3 . The activation energies are 
small, and for CO 
2
are actually slightly negative.
The pressure increase time - lag method employed at GKSS enabled diffusion coeffi -
cient, D , as well as permeability coeffi cient, P , to be evaluated. Solubility coeffi cient, S , 
could then be calculated assuming a simple solution - diffusion model of permeability, 
P = SD . Values of D and S for PIM - 1 membranes in various states are given in Table 
2.4 . The D values are smaller than obtained for other high permeability polymers, whereas 
S values are signifi cantly higher (e.g. for PTMSP D (CO 
2
) has been reported as 250
×
10 
– 7
cm 
2
s 
– 1
and S (CO 
2
) as 76
×
10 
– 3
cm 
3
[STP] cm 
– 3
cmHg 
– 1
) [30] . It is the high S values that 
give PIM - 1 its remarkable combination of high permeability and good selectivity. The 
unusually high S values may be attributed to the microporous structure of PIM - 1, coupled 
with the presence of polar nitrile ( – C 
≡
N) groups. To obtain further information about 
solubility coeffi cients and deeper insight of sorption thermodynamics in this polymer 
inverse gas chromatography was used.