Polymeric Membranes for H₂S and CO₂ Removal from Natural Gas for Hydrogen Production: A Review

Abstract

Natural gas, an important source of hydrogen, is expected to be crucial in the transition to a hydrogen-based economy. The landscape of the gas processing industry is set to change in the near future with the development of highly acidic sour gas wells. Natural gas purification constitutes a major share of the gas separation membrane market, and the shift to low-quality sour gas wells has been mirrored in the trends of membrane material research. Purification also constitutes the major portion of the cost of natural gas, posing implications for the cost of hydrogen production. This review provides an update on the current state of research regarding polymeric membranes for H₂S removal, along with CO₂ separation, from natural gas that is used for hydrogen production via steam methane reforming. The challenges of adapting polymeric membranes to ternary H₂S/CO₂/CH₄ separations are discussed in detail. Key polymeric materials are highlighted, and the prospects for their application in H₂S removal from natural gas are evaluated. Finally, the growing interest in H₂ production from H₂S is discussed. Advances in the membrane industry and the emergence of new membrane materials may significantly improve the commercial viability of such processes.

1. Introduction

Natural gas, a relatively clean fossil fuel and an important chemical feedstock, is the primary source of industrial hydrogen [1,2]. As of 2020, about 60% of the world’s hydrogen was produced from natural gas through steam methane reforming (SMR) [1]. In this process, methane is reacted with steam to produce syngas through the steam reforming reaction (CH₄ + H₂O $\rightleftharpoons$ CO + 3 H₂). This is followed by the water gas shift reaction, wherein CO is reacted with steam to make more hydrogen (CO + H₂O $\rightleftharpoons$ CO₂ + H₂) [1,2]. SMR will continue to be an important route for hydrogen production in the foreseeable future, until renewable routes, such as water electrolysis, become scalable [2,3]. As a result, natural gas is seen as a key bridge fuel on the path to a hydrogen-based economy [3,4].

Raw natural gas contains mainly CH₄ (70–90%), along with contaminants such as CO₂ (0–20%), H₂S (0–20%), and higher hydrocarbons (0–5%) [5,6,7,8]. These contaminants must be removed before transportation to prevent corrosion and damage to equipment [6]. For instance, US pipeline standards mandate a CO₂ content below 2% and a H₂S content of below 4 ppm to prevent corrosion [6]. Desulfurization is also critical for H₂ production from natural gas. SMR typically uses nickel- or platinum-based catalysts, which are known to be extremely sensitive to sulfur [9,10,11]. Even trace amounts of H₂S can poison the catalyst and drastically lower the yield of H₂. CO₂ is a lesser concern; however, removing CO₂ from natural gas near the wellhead can improve the efficiency of H₂ production and ease the requirements on downstream carbon capture and sequestration [12].

Treatment of natural gas constitutes the major portion of its cost [13]. Reservoirs with high acid gas content (>15%) can be significantly expensive to develop since extensive treatments are required to meet pipeline standards [14]. Reservoirs with high H₂S content are especially difficult to deal with, due to the high toxicity and corrosiveness of H₂S [8]. Extraction of such gases in a safe and environmentally benign manner is fraught with technical and economic difficulties. Consequently, highly sour gas reserves have historically remained undeveloped [14].

In recent years, however, the availability of new technologies has made the processing of such sour gases feasible [15,16]. Moreover, with the increasing interest in phasing out coal, the global demand for natural gas is on the rise [4,17]. Technological advances and the growing demand for natural gas have made it possible to revisit sour gas reservoirs previously considered unviable [18,19,20,21,22]. The best known example of this is the Shah gas field in UAE, which has a H₂S content of 23% and CO₂ content of 10% [15,16]. Discovered in 1967, the reservoir remained undeveloped for over 30 years due to the technological and safety issues involved in the processing of such highly sour gases. The field eventually began production in 2014, making it the most sour operational gas field in the world.

Natural gas purification has long been one of the most important markets for industrial gas separation membranes, and one that still holds tremendous potential for growth [6]. The conventional technology for acid gas removal, amine scrubbing, still controls about 90% of the natural gas purification market [6]. Absorption towers are reliable, tried-and-tested technologies that can reduce the acid gas contents down to ppm levels. However, for the treatment of highly acidic gases, the capital costs associated with such systems can be prohibitively expensive [6]. Furthermore, issues such as corrosion and amine degradation are very common at high acid gas concentrations, necessitating constant maintenance and process monitoring [23].

Membranes were first introduced as an alternative to absorption in the 1980s, primarily for use in remote locations where frequent maintenance was difficult [6]. Despite initial hiccups, they quickly carved out a niche for themselves in the natural gas market. Membrane systems can treat highly acidic gases more efficiently and at lower costs than absorption [6]. The high acid gas partial pressures enable high fluxes across the membrane, and membrane systems for treating such gases can be fairly compact and inexpensive [6]. Moreover, since these systems bypass the absorption–stripping cycle, they are less susceptible to corrosion and require less maintenance and monitoring. Such intrinsic advantages allow membranes to establish themselves in the natural gas processing market. They have quickly made their way into other processes in the industry, such as the removal nitrogen and heavy hydrocarbons [6].

Over the past forty years, the market for natural gas membranes has grown significantly, valued at over USD 300 million/year globally [2,24,25,26,27]. While other applications have developed over time, the majority of this market still belongs to acid gas removal. Natural gas purification is among the most actively investigated separations in membrane research, and changes in the industry have influenced the trends of membrane material development [27]. Over the past decade, there has been a surge in the number of publications addressing sour gas separation, reflecting the growing interest in this area [21].

The shift to sour natural gases has also made waves in other areas of chemical research. Currently, the most common method for large-scale conversion of H₂S is the Claus process [28]. The Claus process converts H₂S into elemental sulfur, which is then sold as a by-product to mitigate the cost of H₂S disposal. This process, which has been the industry standard for over a century, is now becoming unsustainable [28]. With the growing number of sour wells being developed, global sulfur production has increased [29,30]. The market for sulfur is currently in oversupply, further penalizing the already marginal economics of the process [29,30,31].

A number of alternative sulfur disposal processes have been proposed in lieu of the Claus process [32,33]. Among these, the production of H₂ from H₂S is one that has attracted significant interest in recent years [33,34,35,36,37,38,39,40,41]. The idea itself has been popular in academic circles for over five decades [42]. Even so, it is only recently that the changing economics of sour gas extraction have made the large-scale implementation of these technologies a possibility. The production of high-value hydrogen to mitigate the cost of H₂S disposal is an attractive proposition, both from an economic and environmental perspective. Such strategies are being actively investigated in regions with sour gas deposits [34,35,41].

With the growing number of sour gas wells, the natural gas industry is expected to change significantly over the course of the next decade. Membrane research has reflected these changes, and materials previously developed for sweet gas separations are now being explored for sour gas processing. This review attempts to provide an updated overview of the state of research in the field and the potential avenues for future development. The theory behind gas separation in polymeric membranes is discussed, along with the implications for the design of polymers for ternary H₂S/CO₂/CH₄ separation. The trade-off between H₂S/CH₄ separation and CO₂/CH₄ separation is covered, and possible approaches to overcoming this trade-off are detailed. Key polymeric materials are examined, and promising research directions are highlighted. Finally, the review looks into the growing interest in hydrogen production from H₂S, along with the implications for membranes. The feasibility of a new membrane application, namely, H₂/H₂S or H₂S/H₂ separation, in the natural gas-adjacent industry is explored.

It should be noted that this review assumes some familiarity with concepts such as membrane gas separation, plasticization, physical aging, etc. Brief explanations are provided wherever necessary. For further details, readers are referred to recent publications [21,43,44].

2. Solution–Diffusion Theory

Gas transport in polymeric membranes can be described by the solution–diffusion mechanism [44]. The permeability is given by the equation:

$$P_i=D_i\times S_i$$

(1)

where $P_i$, $D_i$, and $S_i$ refer to the permeability, diffusivity and solubility of gas species $i$, respectively. The ideal selectivity ${\alpha }_{i/j}$ between two gas species $i$ and $j$ can be described in a similar fashion:

$${\alpha }_{i/j}\equiv \frac{P_i}{P_j}=\frac{S_i}{S_j}\times \frac{D_i}{D_j}$$

(2)

The ratio $D_i/D_j$ represents the diffusivity selectivity and is inversely proportional to the molecular sizes of the two gas species $i$ and $j$. For light gases, the kinetic diameter of the molecules is used as a measure of the molecular size [45]. The term $S_i/S_j$ refers to the solubility selectivity. In general, the condensability of the gas species is a good measure of its solubility [45]. Membrane researchers commonly use the Lennard–Jones temperature to estimate gas solubility [45]. The overall ideal selectivity $P_i/P_j$ depends on the relative magnitudes of these two terms.

Below the glass transition temperature, polymer chains have limited mobility; consequently, the polymer possesses a rigid structure that allows for strong size sieving. Accordingly, glassy polymers offer separation based on the diffusivity selectivity [46,47]. In contrast, rubbery polymers possess mobile chains which make for poor size sieving. For rubbery polymers, the selectivity typically depends upon the solubility selectivity [46,47].

Table 1. Relevant physical properties of H₂S, CO₂, and CH₄ [48,49].

Gas	Kinetic Diameter (Å)	Lennard–Jones Temperature (K)
H2S	3.62	301
CO2	3.32	195
CH4	3.82	149

shows the kinetic diameters and Lennard–Jones temperatures of the gases of interest, namely, H₂S, CO₂, and CH₄ [48,49]. Both the solubility and diffusivity favor the permeation of the acid gases, H₂S and CO₂ over CH₄, implying that both glassy and rubbery polymers can be suitable for this separation. However, membrane researchers usually prefer glassy polymers for this separation, for reasons which will be discussed later.

For CO₂/CH₄ separation, both the size selectivity and solubility selectivity strongly favor CO₂. Polymers with high CO₂/CH₄ selectivity are typically glassy polymers with strong size sieving ability and high gas solubilities [27]. However, due to the similar kinetic sizes of H₂S and CH₄, high selectivity cannot be achieved by size sieving. Accordingly, polymers with high H₂S/CH₄ selectivity are generally rubbery polymers with solubility-based separation [6].

Despite the different separation mechanisms of H₂S-selective and CO₂-selective polymers, some polymers still offer workable selectivity for both H₂S and CO₂ over CH₄. Even so, ternary H₂S/CO₂/CH₄ separations necessitate some concession on the individual H₂S/CH₄ and CO₂/CH₄ selectivities. Figure 1, which presents the mixed-gas H₂S/CH₄ selectivities and CO₂/CH₄ selectivities of selected polymers, highlights this compromise. Rubbery polymers can achieve H₂S/CH₄ selectivities in excess of 100, whereas glassy polymers offer more balanced separation with moderate H₂S/CH₄ and CO₂/CH₄ selectivities of around 25.

3. Upper Bounds for H₂S/CH₄ and CO₂/CH₄ Gas Pairs

A more systematic way to compare the separation performances of glassy and rubbery polymers is by examining their locations on the upper bounds of the CO₂/CH₄ and H₂S/CH₄ gas pairs. The upper bounds for membrane gas separations were developed by Robeson in 1991 as a metric for assessing the state-of-the-art of polymeric gas separation membranes [54]. It has long been known that polymeric membranes display a distinct trade-off between permeability and selectivity [27,45,54]. Recognizing that this alluded to an intrinsic limit in the separation performance of polymeric materials, Robeson analyzed the gas permeability data available at the time and proposed an upper bound on the performance [54]. The upper bound is represented on the log–log plot of the selectivity against the permeability [27,45,54]. For a gas pair $i$ and $j$, the upper bound is given by the equation:

$${\alpha }_{i/j}=\frac{{\beta }_{i/j}}{{P_i}^{{\lambda }_{i/j}}}$$

(3)

where ${\beta }_{i/j}$ is the intercept and ${\lambda }_{i/j}$ controls the slope of the upper bound. The upper bound relation forms a linear boundary on the log–log plot, beyond which few data points exist.

The upper bounds were initially presented as an empirical relation [54]. Freeman later developed a theoretical model to explain the existence of the upper bounds [45]. He used the Arrhenius equation to describe gas diffusivity $D_i$ within the polymer:

$$D_i=D_{0_i}exp\left(-\frac{E_{D_i}}{RT}\right)$$

(4)

where $D_{0_i}$ is the pre-exponential factor. $E_{D_i}$ is the activation energy of diffusion. $R$ and $T$ stand for the universal gas constant and the absolute temperature, respectively. The pre-exponential factor $D_{0_i}$ relates to the entropy of diffusion and is given by the following relation:

$$\ln D_{0_i}=a\frac{E_{D_i}}{RT}-b$$

(5)

where $a$ and $b$ are parameters independent of gas type. $a$ has a universal value of 0.64. $b$ takes a value of 9.5 for rubbery polymers and 11.5 for glassy polymers.

For light gases, the activation energy can be related to the kinetic diameter $d_i$ of the gas as follows:

$$E_{D_i}=c{d}_i^2-f$$

(6)

where $c$ and $f$ are parameters independent of gas type. $f$ serves as a measure of chain rigidity, varying from 0 cal/mol for rubbery polymers to 14,000 cal/mol for glassy polymers [45]. In this instance, it was used as an adjustable parameter, and a value of 12,600 cal/mol was obtained through fitting [45]. Combining Equations (4)–(6) gives us the equation for diffusivity:

$$\mathit{ln}{D}_i=-\left(\frac{1-a}{RT}\right)c{d}_i^2+f\left(\frac{1-a}{RT}\right)-b$$

(7)

Gas solubility is rather more variable than diffusivity and harder to correlate across different families of polymers [45]. For the glassy polymers that dominate the upper bound, the solubility is given by the dual-mode sorption model [55]:

$$S_i=k_{D_i}+\frac{C_{Hi}^{\text{'}}{b}_i}{1+b_i{p}_i}$$

(8)

$k_{D_i}$ and $b_i$ are the Henry’s law solubility coefficient and the affinity constant, respectively [50]. $p_i$ refers to the gas pressure of the gas species $i$. $C_{Hi}^{\text{'}}$ is the Langmuir sorption capacity and takes a value of 0 for rubbery polymers [56]. Both $k_{D_i}$ and $b_i$ correlate strongly with gas condensability [55]. $C_{Hi}^{\text{'}}$ is linked to the excess non-equilibrium free volume for glassy polymers and difficult to correlate with accuracy [55].

For simplicity, Freeman related gas solubility to its Lennard–Jones temperature $\frac{{\epsilon }_i}{k}$ using the correlation:

$$\ln S_i=M+N\frac{{\epsilon }_i}{k}$$

(9)

The value of $M$ is sensitive to the polymer and the gas species. Freeman initially recommended a value of −9.84 for $M$ [45]. $N$ takes a value of 0.023 K⁻¹ [45]. Robeson et al. later revised these to a value of −7.30 for $M$ and 0.0249 for $N$ in 2014 [46].

Combining Equations (1), (2), and (7) gives us the gas permeability and selectivity:

$$\mathit{ln}{P}_i=-\left(\frac{1-a}{RT}\right)c{d}_i^2+f\left(\frac{1-a}{RT}\right)-b+ln{S}_i$$

(10)

$$\mathit{ln}{\alpha }_{i/j}=-\left({\left(\frac{d_j}{d_i}\right)}^2-1\right)\mathit{ln}{P}_i+\left(ln\left(\frac{S_i}{S_j}\right)-\left({\left(\frac{d_j}{d_i}\right)}^2-1\right)\times \left(b-f\left(\frac{1-a}{RT}\right)-ln{S}_i\right)\right)$$

(11)

Comparing Equations (2), (3), and (7) with the definition of ${\lambda }_{i/j}$ in Equation (12), we have the upper bound parameter ${\beta }_{i/j}$:

$${\lambda }_{i/j}={\left(\frac{d_j}{d_i}\right)}^2-1$$

(12)

$${\beta }_{i/j}=\frac{S_i}{S_j}{S}_i^{{\lambda }_{i/j}}\times \mathrm{e}\mathrm{x}\mathrm{p}\left\{-{\lambda }_{i/j}\left(b-f\left(\frac{1-a}{RT}\right)\right)\right\}$$

(13)

The upper bound slope, ${\lambda }_{i/j}$, depends solely on the kinetic sizes of the gas pair. The intercept, ${\beta }_{i/j}$, incorporates the dependence on gas solubility and polymer properties. The upper bound polymers exhibit enhanced separation performance by virtue of their high chain stiffness (reflected in the value of $f$), improved solubility $S_i$, or solubility selectivity $\frac{S_i}{S_j}$ [45].

The upper bounds represent the state-of-the-art performances and serve as a benchmark to be surpassed in the quest for new, improved polymeric materials. Freeman’s model predicts that revisions in the upper bounds with the development of new, improved membrane materials would likely affect only the upper bound intercepts, leaving the slopes unchanged. This prediction has largely held true in the 25 years that have passed since the upper bounds were proposed.

Using Equations (10)–(13), the upper bounds can be predicted for the CO₂/CH₄ and H₂S/CH₄ gas pairs. Figure 2a,b present these upper bounds, along with the permeability data of selected polymers [25,48,50,53,57,58,59,60,61]. As shown in Figure 2, the upper bounds for glassy polymers are positioned above the rubbery polymer bounds. This is attributed to the higher gas solubilities in the micropores trapped in the glassy polymers [46,47].

Figure 2a shows the CO₂/CH₄ upper bound. The CO₂/CH₄ gas pair is one of the most investigated membrane separations, and a myriad of data points are available in literature [27]. In the original 1991 paper, polyimides featured prominently on the CO₂/CH₄ upper bound and defined it while rubbery polymers were located far below [49]. Since then, the upper bound has moved up with the development of two new families of super glassy polymers: thermally rearranged polymers and polymers of intrinsic microporosity (PIMs) [25,27,50,60,62]. The latest major revision in 2019 was based on a number of triptycene-based PIMs with exceptional gas permeability [60]. In short, the CO₂/CH₄ upper bound is dominated by glassy polymers, and few rubbery polymers display comparable separation performance.

On the other hand, the H₂S/CH₄ upper bound is defined by rubbery solubility-selective polymers. The H₂S/CH₄ gas pair has not been studied much over the intervening years, and no significant improvements have occurred in the upper bound. The data suggest a flatter upper bound shown in Figure 2b. This is common for gas pairs with similar kinetic sizes, since performance improvements can occur only through solubility selectivity [63]. Most of the upper bound polymers are those based on poly(ethylene oxide) (PEO) [50,52,53]. Glassy polymers are positioned well below the upper bound, with the exception of an amidoxime-functionalized PIM reported by Yi et al. and a highly crosslinked PEO reported by Harrigan et al. [52,64].

Figure 2c shows the combined acid gas upper bounds proposed by Kraftschik et al. in 2013 [61]. Kraftschik et al. observed that the disparate separation performance of glassy and rubbery polymers made it difficult to compare their performance on the binary upper bound plots. They proposed the use of a combined acid gas plot to allow for a more meaningful comparison of their overall separation efficiency [61]. The acid gas permeability $P_{AG}$ and acid gas selectivity ${\alpha }_{AG}$ are defined as follows:

$$P_{AG}=P_{H_{2}S}+P_{{CO}_2}$$

(14)

$${\alpha }_{AG}=\frac{P_{H_{2}S}+P_{{CO}_2}}{P_{{CH}_4}}$$

(15)

While no clear upper bound suggests itself, rubbery polymers are positioned above glassy polymers on the combined acid gas plots, due to their high H₂S/CH₄ selectivity and high overall permeabilities. No theoretical upper bound limit has been defined for the combined acid gas plots. Even so, the combined acid gas plot has become popular among researchers targeting ternary sour gas separations [49,61]. Natural gas separations require simultaneous removal of H₂S and CO₂, and the combined acid gas plot is an effective metric for evaluating the overall efficiency of acid gas removal.

4. Mixed-Gas CO₂/CH₄ Upper Bounds and Free Volume Model

The upper bound correlations and the activated diffusion model are derived based on pure gas permeability data under idealized conditions. Lin and Yavari observed the realistic natural gas conditions involved a high content of plasticizing gases [62]. Under practical operating conditions, the plasticization-induced increase in free volume results in a significant deviation from pure-gas performance. Gas permeability increases, with a corresponding drop in selectivity. Figure 2a also includes the sweet mixed-gas and sour mixed-gas performances of CO₂/CH₄-selective glassy polymers. It is clear that the mixed-gas performances are significantly lower than the pure-gas performances [49].

The activated diffusion model does not provide any framework to account for the plasticization-associated loss of selectivity [62]. To address this limitation, Lin and Yavari derived upper bound correlations based on the free volume theory. Per the free volume theory, gas diffusivity can be expressed as a function of fractional free volume $FFV$:

$$D_i=D_{0_i}exp\left(-\frac{B_i}{FFV}\right)=D_{0_i}exp\left(-\frac{B_0+\kappa d_i^2}{FFV}\right)$$

(16)

where $B_i$, $B_0$, and $\kappa$ are parameters dependent on the choice of polymer. Using Equation (16) in conjunction with Equations (7)–(10), Lin and Yavari derived a modified correlation for the upper bound. The slope term, ${\lambda }_{i/j}$, remains unchanged from the Freeman model. The intercept, ${\beta }_{i/j}$, contains the fractional free volume (FFV) term:

$${\beta }_{i/j}=\frac{P_{i_0}}{P_{j_0}}{P}_{i_0}^{{\lambda }_{i/j}}\times \mathrm{e}\mathrm{x}\mathrm{p}\left\{-{\lambda }_{i/j}\frac{B_0}{FFV}\right\}$$

(17)

where $P_{i_0}$ and $P_{j_0}$ are preexponential factors. For a gas species $i$, the preexponential factor is defined as $P_{i_0}=D_{i_0}\times S_i$.

For glassy polymers, the glass transition temperature is generally used as a measure of chain rigidity [55]. Here, the depression in glass transition temperature was used to estimate the extent of plasticization. Accordingly, the increase in free volume is given by the equation:

$${FFV}_m={FFV}_p+{\alpha }_g{T}_{g,p}\left(1-\frac{T_{g,m}}{T_{g,p}}\right)={FFV}_p+{\alpha }_g{T}_{g,p}\left(f\left(C_g\right)\right)$$

(18)

${FFV}_m$ and ${FFV}_p$ refer to the FFV of the polymer–gas mixture and that of the polymer, respectively. $T_{g,p}$ and $T_{g,m}$ refer to the glass transition temperatures of the neat polymer and the polymer–gas mixture, respectively. $C_g$ refers to the concentration of the sorbed gas. ${\alpha }_g$ is the thermal expansion coefficient, which serves here as a measure of polymer’s swelling under plasticization. Several correlations are available to calculate $T_{g,m}$ if $C_g$ is known [62]. Lin and Yavari used the Fox equation for simplicity [62].

Accordingly, the expressions for $P_i^m$, the permeability of species $i$ in the polymer-gas mixture, and ${\alpha }_{i/j}^m$, the selectivity of the polymer-gas mixture, are given by:

$$ln\frac{P_i^m}{P_i}=\frac{B_i{\alpha }_g{T}_{g,p}\left(1-\frac{T_{g,m}}{T_{g,p}}\right)}{{FFV}_m{FFV}_p}$$

(19)

$$ln\frac{{\alpha }_{i/j}^m}{{\alpha }_{i/j}}=\frac{\kappa \left(d_i^2-d_j^2\right){\alpha }_g{T}_{g,p}\left(1-\frac{T_{g,m}}{T_{g,p}}\right)}{{FFV}_m{FFV}_p}$$

(20)

Then, the expression for ${\beta }_{i/j}^m$, the intercept for the mixed gas upper bound, is given by the expression:

$$ln{\beta }_{i/j}=ln{\beta }_{i/j}^m-\frac{{\alpha }_g{T}_{g,p}\left(1-\raisebox{1ex}{$T_{g,m}$}\!\left/ \!\raisebox{-1ex}{$T_{g,p}$}\right.\right)\left({\lambda }_{i/j}{B}_j-\left(1-{\lambda }_{i/j}\right)B_i\right)}{{FFV}_m{FFV}_p}$$

(21)

If the values of ${\alpha }_g$ and $T_{g,p}$ are known for the upper bound polymers, the influence of plasticization on the upper bound can be predicted. The term ${\alpha }_g{T}_{g,p}$ controls the polymer’s plasticization response. Lin and Yavari suggested a general value of 0.051 for ${\alpha }_g{T}_{g,p}$, based on Simha and Boyer’s recommendation [62,65,66].

Figure 3a shows the mixed-gas CO₂/CH₄ upper bounds, calculated at two different CO₂ contents. The upper bound shifts downward with increasing feed CO₂ partial pressure. The model can also be extended to account for other plasticizing gases if an estimate of the plasticizer’s glass transition temperature is available.

It should be noted that the free volume model does not consider the impact of competitive sorption on the mixed-gas permeabilities. Equation (19), which describes the effect of feed CO₂ pressure on the permeability, will only ever predict a rise in the permeability. This is largely true for cellulose acetate, the model polymer used in Lin and Yavari’s work.

However, polyimides and PIMs, the polymers that currently define the upper bound, usually exhibit an initial decline in CO₂ permeability due to competitive sorption between CO₂ and CH₄. Figure 3b illustrates the effect of feed CO₂ pressure on the CO₂ permeability of a polyimide based on 4,4′-(hexafluoroisopropylidene) diphthalic anhydride (6FDA), a polymer with performance close to the 2008 upper bound [67]. The decrease in CO₂ solubility due to competitive sorption and Langmuir saturation counters the plasticization-associated increase in diffusivity. For polyimides, the gas permeability typically goes through a minimum with increasing feed CO₂ pressure. As a result, polyimides exhibit markedly different plasticization behavior than cellulose acetate, and generally retain higher selectivity under plasticizing conditions.

This shortcoming of the model limits its application to sour gas separations. For ternary H₂S/CO₂/CH₄ mixtures, the competitive sorption of H₂S is expected to suppress CO₂ permeation [61]. Consequently, the modified free volume model will likely underestimate the downward shift in the CO₂/CH₄ upper bound under sour gas conditions. Additionally, there are considerable uncertainties in the values of the model parameters, ${\alpha }_g$, $B_i$, and $P_{i0}$ [62]. Given these shortcomings, the free volume model and the mixed-gas upper bounds have limited applicability when used predictively for sour gas separations.

Despite this limitation, the modified free volume model represents one of the first efforts towards developing a generalized upper bound equation that relates membrane performance to the concentration of plasticizing gases. It allows us to formulate an expression describing the inherent trade-off at the crux of sour gas separations: Improvements in H₂S/CH₄ selectivity come at the cost of CO₂/CH₄ selectivity and vice versa. For high H₂S/CH₄ selectivity, polymers must possess high H₂S/CH₄ solubility selectivity. For glassy polymers, this translates to high H₂S solubility. However, under practical conditions, such polymers will inevitably undergo severe H₂S-induced plasticization, resulting in a drastic loss of CO₂/CH₄ diffusivity selectivity. Conversely, for high CO₂/CH₄ selectivity in the presence of high H₂S partial pressures, polymers must possess superior resistance to H₂S plasticization. Such polymers possess low H₂S solubility, consequently low H₂S/CH₄ selectivity.

Figure 4a,b provide an illustration of the trade-off between H₂S/CH₄ selectivity and CO₂/CH₄ selectivity due to plasticization, as calculated by Lin and Yavari’s model. The selectivities are calculated as:

$${\alpha }_{H_{2}S}=\frac{D_{H_{2}S}}{D_{C{H}_4}}\times \frac{S_{H_{2}S}}{S_{C{H}_4}}$$

(22)

$${\alpha }_{{CO}_2}=\frac{D_{{CO}_2}}{D_{C{H}_4}}\times \frac{S_{{CO}_2}}{S_{C{H}_4}}\times \frac{\kappa \left(d_{{CO}_2}^2-d_{C{H}_4}^2\right){\alpha }_g{T}_{g,p}\left(1-\frac{T_{g,m}}{T_{g,p}}\right)}{{FFV}_m{FFV}_p}$$

(23)

where $T_{g,m}$ is calculated using the Fox equation:

$$\frac{1}{T_{g,m}}=\frac{w_p}{T_{g,p}}+\frac{w_{H_{2}S}}{T_{g,H_{2}S}}=\frac{w_p}{T_{g,p}}+\frac{f\left(S_{H_{2}S}{p}_{H_{2}S}\right)}{T_{g,H_{2}S}}$$

(24)

where $p_{H_{2}S}$ refers to the feed H₂S pressure. $w_p$ and $w_{H_{2}S}$ are the weight fractions of the polymer and H₂S, respectively.

In Figure 4, the curves are computed by varying $S_{H_{2}S}$. The term ${\alpha }_g{T}_{g,p}$ has been treated as a polymer property, rather than an innate constant. It is expected that $S_{H_{2}S}$ and ${\alpha }_g{T}_{g,p}$ are interrelated; however, for simplicity, we ignore it here. No experimental value is available for $T_{g,H_{2}S}$. A value of 120 K is used based on Reference [68] and provides a satisfactory fit. It is assumed that the H₂S/CH₄ selectivity is controlled by solubility and is unaffected by plasticization. Competitive sorption can be neglected once plasticization has set in and is not considered here. It should be noted that, despite limitations in predicting permeability, the free volume model captures the trend in mixed-gas selectivity quite well. Therefore, the predicted variation in mixed-gas selectivity should be representative of the actual membrane performance.

As shown in Figure 4, ternary separations necessitate some compromise on the H₂S/CH₄ selectivity and the CO₂/CH₄ selectivity under practical operating conditions. Simultaneously achieving both high H₂S/CH₄ selectivity and high CO₂/CH₄ selectivity under ternary gas conditions is rather more challenging than maximizing the binary H₂S/CH₄ or CO₂/CH₄ selectivity. The trade-off prescribed by Lin and Yavari’s model presents two possible routes for improving acid gas selectivities:

(1)

Decreasing ${\alpha }_g$: As shown in Figure 4a, increasing the chain rigidity and improving plasticization resistance can allow for improvements in both H₂S/CH₄ selectivity and CO₂/CH₄ selectivity. This can be conducted through methods such as crosslinking, annealing, etc. [52,61,69]. Both these approaches generally result in decreased acid gas permeability.

(2)

Increasing $\frac{S_{{CO}_2}}{S_{C{H}_4}}$: As shown in Figure 4b, increasing the solubility selectivity for CO₂ over CH₄ can mitigate the selectivity loss due to plasticization and improve both H₂S/CH₄ selectivity and CO₂/CH₄ selectivity. The most straightforward route to this is by incorporating functional groups that provide favorable sorption interactions for the acid gases [64,70,71]. It should be noted that, for such polymers, uncontrolled swelling and plasticization may occur [72]. While some degree of plasticization is acceptable and can even be beneficial, severe plasticization can cause deterioration in mechanical strength and should be avoided [69].

In practice, both approaches are used together to obtain a balanced combination of acid gas permeability and plasticization resistance. The next section discusses the use of these strategies when designing polymeric membranes for natural gas separations.

5. Polymeric Membranes for Natural Gas Purification

Before reviewing trends in membrane research, it is necessary to discuss some process design-based considerations that have shaped the direction of research. A detailed discussion of the process design and cost-based considerations is available elsewhere [6]; a brief overview is provided here. Studies comparing the economics of amine absorption and membranes show that membranes are favored at high acid gas content (>15%) and small scale (flow rate < 100 MMscfd (MMscfd: million standard cubic feet per day)) [5,6,73]. Specifically, membranes are best suited to natural gases with high CO₂ content (>5%) and comparatively lower H₂S content (<1%) [5,73,74]. For natural gases with high H₂S content, absorption is preferred over membranes for the following reasons.

The H₂S pipeline standards of below 4 ppm cannot be achieved via membranes; this may change with a new, more effective membrane. Currently, to overcome this, hybrid systems using membranes for bulk removal followed by an amine polishing step have been proposed [6,73]. However, this approach has, so far, been restricted to large plants which can absorb the additional costs and complexity introduced by the hybrid process [6,73,75]. This may become more common in the future due to the increasing demand for natural gas [76,77].

H₂S is highly toxic and corrosive. In the event of shutdown, it is generally preferable to deal with large quantities of H₂S in the form of dilute solutions rather than as concentrated gas streams [74]. However, an improved shutdown procedure may alleviate this concern.

Figure 5 shows the application envelopes within which membranes are favored over absorption processes [5,6]. For sweet natural gas separations shown in Figure 5a, membranes are competitive with absorption over a wide range of concentrations. However, for sour natural gases illustrated in Figure 5b, the application of membranes is limited to a very narrow range of concentrations. When H₂S concentrations exceed 1–2%, gas treatment becomes prohibitively expensive at small scales due to the requirement of expensive, corrosion-resistant materials [78,79,80].

Consequently, membrane researchers mainly targeted sweet natural gas and natural gases with relatively low H₂S content. These conditions favor the use of glassy polymers over rubbery polymers. For such membranes, the stage cut required to reduce the CO₂ content to 2% may also be sufficient to reduce the H₂S content to a few hundred ppm, at which point the use of H₂S scavengers becomes an option [5,74,81].

Furthermore, as multiple techno-economic analysis studies have noted, the main factor disadvantaging membranes for large scale operation is the loss of CH₄ in the permeate [73,75,82]. Hence, a more selective membrane can minimize CH₄ losses and reduce costs. Since the focus is CO₂-rich natural gases, this again favors the use of CO₂-selective glassy polymers over H₂S-selective rubbery polymers. The impact of membrane area requirement on the cost is unimportant in comparison, and the higher permeabilities of rubbery polymers do not offer any meaningful advantage [73,75,82].

Given these factors, it is unsurprising that membrane researchers have focused mostly on glassy polymers. Rubbery polymers have only been explored in the context of hybrid processes or for two-stage processes in combination with a glassy polymer membrane [5,73,82]. Relatively few publications discuss the development of rubber polymeric membranes for natural gas purification [21]. However, the push to develop sour gas fields will likely renew interest in the development of rubbery polymer membranes.

This review discusses some of the most well-studied glassy polymers for natural gas separations, namely, cellulose acetate (CA), polyimides, and PIMs. PEO-based polymers, a class of rubbery polymers with high H₂S/CH₄ selectivity, are also covered. Mixed matrix membranes and inorganic membranes are beyond the scope of this work, but information can be found in a recent review [21]. Likewise, the process design aspects of natural gas purification are outside the scope of this work, but information can be found in other publications [6,24,73,74,75,76,77].

5.1. Cellulose Acetate (CA)

One of the first commercial CO₂ separation materials, CA membranes remain the industry standard for natural gas separations [83]. CA membranes occupy a distinguished position in the history of membrane separations—the first ever industrially scalable reverse osmosis membrane, developed by Loeb and Sourirajan in 1960, utilized a CA selective layer [24]. CA-based membranes and modules emerged in the market shortly thereafter [84]. Research in the 1970s led to them being adapted for gas separations, and the first commercial CA-based CO₂/CH₄ separation system was set up in 1983 [85]. Several CA-based hollow-fiber and spiral-wound modules are available commercially today: Schlumberger’s Cynara and Honeywell’s UOP Separex are the most well-known examples [83]. They continue to hold a major share of the market for natural gas separations in spite of the emergence of newer, more selective membrane materials [83].

CA is a semicrystalline polymer with a glass transition temperature of ca. 200 °C [59,86]. Figure 6 depicts the structure of a representative CA polymer. With chain spacings of ca. 5–6 Å and a FFV of around 0.15, CA membranes offer a low CO₂/CH₄ diffusivity selectivity ~4 [87,88]. The hydroxyl and acetate functional groups in CA can offer favorable solubility interactions [89], allowing for a CO₂/CH₄ solubility selectivity in the range of 8–10 [87,88]. Consequently, CA membranes can offer a modest CO₂/CH₄ selectivity of 30–40 and CO₂ permeability of ca. 10 Barrer at low pressure [90]. Plasticization can cause this selectivity to drop to ca. 20 under practical operating conditions, with a corresponding rise in permeability to around 20 Barrer [24,59]. It offers slightly better H₂S/CH₄ separation, with a diffusivity selectivity ~1.5 and solubility selectivity ~20 [91]. Accordingly, CA membranes offer H₂S/CH₄ selectivity of ca. 25–40, with H₂S permeabilities ca. 25 Barrer under practical conditions [59].

Despite the modest selectivity, the impressive resilience of CA membranes under harsh plasticizing conditions has allowed them to retain their edge in the market. Multiple studies report that CA-based membranes can maintain stable CO₂/CH₄ performance in the face of aromatic natural gas contaminants such as toluene, benzene, and xylene [85,93,94,95].

Lu et al. studied the performances of cellulose triacetate (CTA) membranes at low pressure (7 bar) [93]. They found that their CO₂/CH₄ selectivity remained constant at ca. 25–30 at toluene partial pressures of up to 6 kPa. A minimal selectivity loss of ~15% was observed in the presence of xylene at partial pressures of up to 1.5 kPa. Another recent study examined the performance of commercial 200 nm-thick CTA hollow fibers with a ternary H₂S/CO₂/CH₄ mixture at 31 bar, with significant hydrocarbon content (3% C₂H₆, 3% C₃H₈, and 100–300 ppm toluene) [95]. Despite the high hydrocarbon content, the membrane maintained decent acid gas selectivity (H₂S/CH₄ selectivity of 28 and CO₂/CH₄ selectivity of 22) with impressive acid gas permeance (H₂S permeance of 140 GPU and CO₂ permeance of 115 GPU (GPU = Gas permeance unit, 1 GPU = 1 × 10^–6 cm³(STP) cm (cm^–2 s^–1 cmHg^–1)) at 35 °C. In contrast to polyimide membranes, another class for natural gas separations, where the presence of hydrocarbon contaminants has been known to reduce CO₂ permeabilities by as much as 50–90% [88,96,97,98,99], it becomes apparent why CA membranes have remained relevant despite the emergence of more selective membranes.

CA membranes have been in use for almost 50 years now, and research has essentially plateaued. Limited publications have appeared in recent years, most of which target the crystallinity of CA in order to improve separation performance [59,100,101,102,103].

CA contains hydroxyl groups, and the hydrogen-bonding between these polar moieties can cause crystallinity and rigidify the polymer matrix, leading to low permeability. Early studies of CA membranes have shown that crystallinity can have a dramatic effect on permeability [104]. Increasing the degree of acetylation was found to enhance CO₂ permeability, since the bulky acetyl group can suppress crystallization [104]. Increasing the degree of acetylation from 1.75 to 2.85 was found to increase CO₂ permeability from 2 to 6.5 Barrer, due to the reduction in crystallinity and packing.

Strategies that have been pursued to improve permeability in CA membranes include (1) substituting the –OH groups with bulky moieties to reduce hydrogen bonding and packing and (2) the use of blend polymers with disrupted crystallinity. Achoundong et al. modified CA films by grafting vinyltrimethoxysilane (VTMS) on to the –OH groups of CA [59]. Substituting the hydroxyl groups with the bulky VTMS substituent led to lower crystallinity and reduced chain packing. The modified CA films retained their acid gas selectivity (H₂S/CH₄ selectivity ~27 and CO₂/CH₄ selectivity ~20), but exhibited 10-fold increases in acid gas permeability (H₂S permeability of 165 Barrer and CO₂ permeability of 140 Barrer) over the neat CA membranes, even at feed pressures of up to 48 bar. Blends of CA with polyethylene glycol (PEG) have been studied [101,105]. Li et al. studied blends of CA and PEG and noted a slight increase in CO₂ permeability (7.5 Barrer for the blend membrane versus 6 Barrer for the neat CA) with no change in the CO₂/CH₄ selectivity (ca. 30).

Lam et al. [102] prepared blends of CTA with ionic liquids to reduce crystallinity and improve permeability. They fabricated blend membranes of CTA with 1-ethyl-3-methylimidazolium tetrafluoroborate ([emim][BF₄]) and 1-ethyl-3-methylimidazolium dicyanamide ([emim][dca]). The CTA/[emim][BF₄] blend showed a CO₂ permeability of ca. 18 Barrer with a CO₂/CH₄ selectivity (~25). A recent study by the same group [103] suggests that the impact of crystallinity on the performance of CA membranes may have been exaggerated by the difference between industrial membranes, which typically have a thickness <1000 nm, and the bulk films preferred for lab scale testing. They studied blends of CDA and CTA at film thicknesses ranging from 500 nm to 20 μm. For a blend polymer of 75% CTA in CDA, the degree of crystallinity dropped from 23% to 13% as the thickness decreased from 20 μm to 1 μm. The reduction in crystallinity was accompanied by an increase in CO₂ permeability from 7.1 Barrer to 14 Barrer.

5.2. Polyimides

Widely considered as the frontrunner for natural gas separations, polyimides are currently the main focus of research into natural gas membranes. They are a family of highly rigid, glassy polymers synthesized through a condensation reaction between dianhydride and diamine monomers [106]. Their high plasticization resistance, combined with their excellent mechanical and chemical properties, makes polyimides a prime candidate for sour gas separations [61]. Figure 7 shows the structures of some representative 6FDA-based polyimides, one of the most commonly used polymers for natural gas separations [107].

Polyimides have been known to material researchers since 1905 [106]. However, their poor solubility and processability limited their commercial applications for several decades [100]. Breakthroughs in polymer processing in the 1950s enabled their industrial production, and several polyimides became commercially available in the decades that followed [106]. Initially developed for the electronics industry, they attracted the interest of membrane researchers when they were found to display upper-bound performance for several industrially important gas separations [84,108,109,110,111]. Their excellent size selectivity, combined with good mechanical properties, made them a prime candidate for CO₂/CH₄ separations. Research into polyimides took off in the 1980s [111] and made rapid progress, resulting in the installation of the first industrial polyimide-based CO₂/CH₄ separation unit in 1994 [24]. Today, polyimide-based hollow-fiber modules have found widespread use in the chemical and oil and gas industries [26]. They are now considered the main competitor against CA membranes for control of the natural gas membranes market [6].

With rigid chain structures enabled by a stiff aromatic backbone, polyimides are highly size selective, with CO₂/CH₄ diffusivity selectivity of ca. 15–20 [111]. They possess CO₂/CH₄ solubility selectivity ~5, allowing for an overall selectivity of ca. 60–80 at low pressure [83]. However, the planar ring structures also promote chain packing, resulting in low gas permeability; commercial polyimides such as Kapton^® and Matrimid^® usually show CO₂ permeability in the range of 1–10 Barrer [112,113]. Early membrane researchers investigating structure/permeability relationships in polymers noted that polyimides based on the 6FDA monomer showed 10-fold improvements in CO₂ permeability, attributed to the bulky 6FDA group disrupting chain packing [109,114,115,116,117,118]. Their CO₂/CH₄ selectivities remained unaffected, possibly due to the increased chain rigidity imparted by the sterically hindered -C(CF₃)₂- group [109,114,115,116,117,118]. The 6FDA polyimides also showed improved solubility and processability [100], allowing for facile fabrication into hollow-fiber modules [56,57,58]. In view of these advantages, the majority of research concerning polyimide membranes has since circled around the 6FDA polyimides.

Despite their excellent CO₂/CH₄ performance, polyimides are not quite as well-suited for the solubility-based H₂S/CH₄ separation. Their strong size sieving nature penalizes the permeation of the larger H₂S molecules. For instance, the 6FDA polyimides typically possess a H₂S/CH₄ diffusivity selectivity ~1 and a H₂S/CH₄ solubility selectivity $~$10 [61,64,67,119]. Accordingly, they usually exhibit an underwhelming overall H₂S/CH₄ selectivity of ca. 10–20 along with a H₂S permeability of around 30 Barrer at low pressure [61,67,119].

Moreover, polyimides derive their impressive selectivity from their rigid size sieving nature, and the selectivity loss associated with H₂S plasticization is usually far more severe than for CA membranes [95,119]. As a result, polyimides have mostly been marketed to sweet natural gases and natural gases with low H₂S content. Researchers have only recently started adapting them to sour gas separations [61]. Research strategies aim at increasing chain rigidity, thus allowing a modest solubility selectivity for H₂S while still maintaining size selectivity for CO₂. For the tightly packed polyimides, increases in chain rigidity can often cause a decrease in gas permeability [120,121]. Researchers have examined polymer structures with highly rigid backbones and disrupted chain packing to mitigate the loss of permeability.

The 6FDA-durene polyimides are a group of polyimides noted for their high gas permeability and good processability [122]. Figure 8 shows the structure of a 6FDA-durene polyimide synthesized by Yahaya et al. [123]. Steric repulsions between the methyl substituents of the durene ring and the acyl groups of the imide cause the durene and imide rings to be rotated perpendicular to each other [124]. This results in a highly rigid backbone with restricted torsional motion and with reduced packing efficiency.

Yahaya et al. synthesized block copolyimides containing the 6FDA-durene moiety and examined their H₂S/CO₂/CH₄ separation performance [125,126]. They obtained H₂S and CO₂ permeabilities of around 10 Barrer each, along with H₂S/CH₄ and CO₂/CH₄ selectivities of 23 and 27, respectively, at 35 bar. They later prepared copolymers incorporating 6FDA-durene and cardo moieties, shown in Figure 8 [123,127,128]. The tricylic fluorene ring further restricted chain rotation and lowered packing efficiency, leading to higher gas permeability. They obtained H₂S and CO₂ permeabilities of 94 and 150 Barrer, along with H₂S/CH₄ and CO₂/CH₄ selectivities of 12 and 19, respectively.

Liu et al. examined a series of 6FDA polyimides attempting to formulate structure/permeability relationships for H₂S/CO₂/CH₄ separation [67]. They synthesized a series of polyimides based on 6FDA-DAM-DABA by varying the DAM:DABA ratio. Incorporating the 6FDA-DAM monomer was found to increase the free volume by disrupting packing. Increasing the DAM:DABA ratio from 1:2 to 3:2 improved H₂S separation performance at the cost of the CO₂/CH₄ separation performance as illustrated in Figure 9. The H₂S/CH₄ selectivity rose from 21 to 24, while the CO₂/CH₄ selectivity decreased from 34 to 22. At 48 bar, the 3:2 6FDA-DAM:DABA membrane showed H₂S and CO₂ permeabilities of around 100 Barrer, with H₂S/CH₄ and CO₂/CH₄ selectivities of around 20. The same group also studied a 6FDA-DAM membrane at 46 bar under a 20/20/60 mixture, obtaining H₂S and CO₂ permeabilities of 500 and 300 Barrer, along with H₂S/CH₄ and CO₂/CH₄ selectivities of 31 and 19, respectively [119].

The main concern regarding polyimide membranes is their performance loss in the presence of aromatic contaminants. Studies have reported that polyimide membranes are more susceptible to hydrocarbons than CA [83,88]. The polyimides chosen for membrane fabrication must necessarily possess good solubility in organic solvents, which translates into a high sorption capacity for hydrocarbon contaminants [88]. Consequently, high concentrations of such contaminants can lower acid gas selectivity through competitive sorption and plasticization. Moreover, even trace amounts of aromatic contaminants such as toluene and naphthalene can cause a drastic drop in permeability [88,96]. This performance drop is attributed to antiplasticization, a phenomenon which occurs when highly sorbing penetrants with strong penetrant–polymer interactions restrict chain mobility and rigidify the polymer matrix [43].

Antiplasticization causes steep declines in gas permeability, with complicated effects on selectivity [43,129,130,131]. Omole et al. studied the CO₂/CH₄ performance of propane-diol cross-linked 6FDA-DABA:DAM membranes at 30–1000 ppm of toluene [96]. While the crosslinked membranes did not undergo plasticization, the CO₂ permeance dropped from 35 to 20 GPU under 100 ppm of toluene at 50 bar, with a CO₂/CH₄ selectivity of around 35, due to antiplasticization and competitive sorption.

The poor hydrocarbon resistance of polyimides necessitates expensive pretreatments to reduce hydrocarbon concentrations to ppm levels [43]. This negates any economic advantage over CA membranes, which possess low sorption capacities for aromatics and show fairly stable performance [93]. A number of strategies have been pursued to improve the hydrocarbon resistance of polyimides: thermal annealing [61,131], crosslinking [132], and the use of a caulking layer [132]. These strategies can lower the sorption of hydrocarbon penetrants and prevent uncontrolled swelling, allowing for stable performance in the presence of hydrocarbon contaminants.

Kraftschik et al. prepared crosslinked membranes of 6FDA-DAM:DABA and studied their hydrocarbon resistance [69]. Crosslinking usually results in decreased acid gas permeability. To mitigate this, they incorporated triethylene glycol (TEG) crosslinkers. The ether moieties in TEG can undergo favorable sorption interactions with H₂S and CO₂, thus increasing acid gas permeability [89]. A post-crosslinking approach was used to react the carboxylic acid groups in DABA with the hydroxyl groups of the TEG crosslinker, as shown in Figure 10a,b. Under ternary gas measurements at 62 bar, the crosslinked membranes showed H₂S and CO₂ permeabilities of 40 and 50 Barrer, with H₂S/CH₄ and CO₂/CH₄ selectivities of 22 and 27, respectively. In comparison, the uncrosslinked membranes showed higher H₂S and CO₂ permeabilities of 100 and 80 Barrer, with H₂S/CH₄ and CO₂/CH₄ selectivities of 22 and 20, respectively.

However, the subsequent work by the same group found that the crosslinked polyimides possessed superior resistance to hydrocarbon contaminants [72,132]. The crosslinked polyimides were fabricated into hollow-fiber membranes and coated with a polydimethylsiloxane (PDMS) sealing layer. The membranes were tested at 48 bar under 500 ppm of toluene [132]. For binary CO₂/CH₄ permeation tests with toluene, a 30% reduction in CO₂ permeance, from 30 to 20 GPU, was observed due to toluene-induced antiplasticization and competitive sorption. However, under ternary H₂S/CO₂/CH₄ permeation tests, the presence of toluene was found to slightly enhance acid gas permselectivity [72]. Under ternary H₂S/CO₂/CH₄ tests at 35 bar, the H₂S/CH₄ and CO₂/CH₄ selectivities increased slightly, from 20 to 25 and 50 to 60, respectively, when the toluene content was increased from 100 to 300 ppm. The H₂S and CO₂ permeances remained unchanged at around 22 and 10 GPU, respectively. Moreover, membranes that had been stored for over 7 years showed nearly no performance loss, suggesting exceptional aging resistance.

5.3. PIMs and Other Microporous Polymers

PIMs are a relatively new class of microporous polymers that have garnered interest for several key gas separations. First synthesized in 2004, PIMs attracted attention for having extremely high gas permeabilities, several times that of most glassy polymers, while still retaining a moderate selectivity [27,133,134]. The separation performances of PIMs were found to surpass upper bounds for several industrial gas separations; the CO₂/CH₄ upper bounds were updated in 2019 using a series of benzotriptycene-based PIMs [27,60].

The remarkable separation performances of PIMs come from their characteristic contorted molecular structure, usually derived from a rigid, bicyclic monomer. Figure 11 depicts the ladderlike backbone structure of PIM-1 [135].

The contorted chain structures have very poor packing efficiency, resulting in loosely packed chains with large free volume elements or “pores” [135]. Proper film formation protocols can generate a unique microporous structure in PIMs [134]. This microstructure enables unusually high gas solubilities, an order of magnitude higher than most other polymers [134]. The high permeability of PIMs arises primarily from the increased gas solubility, rather than diffusivity [134]. Consequently, they still retain a moderate diffusivity selectivity, which allows them to exhibit upper bound performances for several gas pairs [134].

Strategies aiming at designing PIMs for natural gas separations revolve around improving their solubility selectivity. Another issue that needs to be addressed simultaneously is their poor plasticization resistance. Like all glassy polymers, PIMs are vulnerable to plasticization at high acid gas pressures [71,136], but their high gas solubility hastens the onset of plasticization and further reduces the already modest selectivity. For instance, the spirobisindane-based PIM-1 exhibits a CO₂ permeability in the range of 3000–6000 Barrer, depending upon the method of preparation, with a CO₂/CH₄ selectivity of ca. 15 at low pressures [71,137]. The selectivity drops to 8 at CO₂ partial pressures of 10 bar [71,136]. Such severe loss of selectivity makes plasticization resistance crucial when targeting sour gas separations with high H₂S contents.

Researchers have attempted to improve the plasticization resistance of PIMs by replacing spirobisindane with more rigid, aromatic monomers, such as iptycene, spirobifluorene, etc. [60,138,139,140], to increase the backbone rigidity. Interestingly, though such modifications have been successful in improving selectivity [60,141], they have provided minimal, if any, improvements in plasticization resistance [138].

Swaidan et al. compared the plasticization behavior of PIM-1 and a triptycene-based PIM with increased backbone rigidity [136,138]. The triptycene-based PIM was found to plasticize more severely than PIM-1, despite increased chain rigidity. At a CO₂ partial pressure of 10 bar, the triptycene-based PIM showed a 93% increase in CH₄ permeability, as opposed to a 62% increase in CH₄ permeability for PIM-1. Swaidan et al. explained this contradiction creating a distinction between “interchain rigidity” and “intrachain rigidity”. Interchain rigidity, which arises from interactions or bonds between two polymer chains, can prevent uncontrolled swelling of the polymer matrix and limit plasticization. Intrachain rigidity, which results from restricted bond mobility in the backbone, cannot prevent plasticization. It may even accelerate plasticization by preventing interchain interactions, as in the case of the triptycene-based PIM. Such distinctions are less pronounced in non-microporous polymers, where the closer chain packings allow for strong interchain interactions regardless of backbone rigidity [136,138].

Though PIMs have been extensively studied for CO₂/CH₄ separation [25], only two publications report their application to sour gas separations [64,137,142]. Both involve the use of spirobisindane PIMs with functional groups that allow strong interchain interactions [64,142].

Yi et al. synthesized membranes using a hydroxyl-functionalized 6FDA-polyimide with intrinsic microporosity (PIM-6FDA-OH) [142]. The hydroxyl functionality was incorporated to improve plasticization resistance through interchain hydrogen bonding. Thermally annealed membranes of PIM-6FDA-OH were tested at 48 bar under a ternary 15/15/70 H₂S/CO₂/CH₄ feed. While the membrane showed good resistance to CO₂ plasticization, H₂S-induced plasticization was observed at high feed pressures. They obtained decent H₂S/CH₄ and CO₂/CH₄ selectivities of 30 and 25, with H₂S and CO₂ permeabilities of 63 and 53 Barrer, respectively. However, the low acid gas permeabilities, far below the values typical of PIMs, suggested that the PIM’s microporosity had been diminished.

To improve the permeability, Yi et al. investigated an amidoxime-functionalized PIM (AO-PIM-1) for H₂S/CH₄ separation, illustrated in Figure 12a [64]. The amidoxime functionality is known to exhibit strong interchain hydrogen bonding and improve plasticization resistance [71]. AO-PIM-1 showed good resistance to CO₂-induced plasticization, with binary CO₂/CH₄ permeation tests showing a fairly constant CO₂/CH₄ selectivity of around 30 at a CO₂ partial pressure of 10 bar. Although H₂S-induced plasticization was observed, the AO-PIM-1 membrane still showed remarkable separation performance. At 77 bar under a 20/20/60 ternary feed, AO-PIM-1 showed H₂S/CH₄ and CO₂/CH₄ selectivities of 74 and 13, along with H₂S and CO₂ permeabilities of 4300 and 800 Barrer, respectively. The high H₂S/CH₄ selectivity was attributed to an increase in the H₂S/CH₄ solubility selectivity, enabled by favorable sorption interactions between amidoxime and the H₂S moiety. AO-PIM-1 possesses a H₂S/CH₄ solubility selectivity of ca. 40, whereas PIM-1 possesses a H₂S/CH₄ solubility selectivity of ca. 12.

Interestingly, both the functionalized PIM materials, PIM-6FDA-OH and AO-PIM-1, exhibited a diffusivity selectivity $\frac{D_{H_{2}S}}{D_{C{H}_4}}~$ 0.7 [64,137,142]. Yi et al. hypothesized that strong sorption interactions between the functionalized PIMs and H₂S may have lowered the H₂S diffusivity relative to CH₄.

Poly(norbornenes) are another class of high free volume, super-glassy polymers usually targeted towards solubility-based hydrocarbon separations [56]. Figure 13a presents the structure of a representative alkoxysilyl functionalized poly(norbornene) reported by Maroon et al. [143,144]. The bulky bicyclic side group leads to poor packing efficiency. The loosely packed chains, coupled with the flexible alkyl backbone, results in very mild size sieving selectivity, too low for most light gas separations. Researchers targeting acid gas separation have been successful in improving solubility selectivity through the incorporation of ether functionalities [143,144]. Lawrence et al. studied a series of alkoxysilyl-substituted poly(norbornenes), shown in Figure 13a, for their H₂S/CO₂/CH₄ separation performance [144]. Incorporation of the ether moiety group was found to decrease FFV and lower CH₄ sorption in the functionalized poly(norbornenes). This resulted in a decrease in CH₄ permeability. The reduction in acid gas permeability was less pronounced, attributed to the ether moiety improving the sorption of H₂S and CO₂ as illustrated in Figure 13b. They obtained H₂S and CO₂ permeabilities of 6715 and 1500 Barrer, with H₂S/CH₄ and CO₂/CH₄ selectivity of 40.5 and 10, respectively.

Aging resistance remains the most critical concern for high free volume polymers such as PIMs and polynorbornenes. Aging refers to the loss of non-equilibrium free volume with time as the polymer chains relax to equilibrium. The chains pack more densely, leading to a loss of free volume. As a result, the membrane permeability drops drastically. Spirobisindane-based PIMs are known to be particularly susceptible to aging, with studies reporting 50–90% reductions in permeability over time [136,145,146,147].

Accurate studies of aging are difficult due to the long time scales required for such experiments. In addition, aging behavior is highly dependent on thickness and film formation protocol, which can lead to unreliable results [43]. Such effects are expected to be particularly complicated in natural gas separations, where aging competes with plasticization, another thickness-dependent phenomenon [43]. For instance, Pinnau et al. studied bulk membranes of poly(1-trimethylsilyl-1-propyne) (PTMSP), a high-free volume polymer prone to aging, for low pressure n-butane/hydrogen separation [148]. They reported stable performance with minimal aging over 48 days, and suggested that condensable gases may be acting as ‘fillers’ to prevent the aging-related loss of free volume [148]. However, later reports describe a drastic, irremediable loss of permeability in asymmetric thin-film PTMSP membranes when used for nitrogen removal from hydrocarbon-laden natural gas [147,149,150].

Aiming to shed some light on the competition between plasticization and physical aging, Tiwari et al. studied the thickness dependence of these phenomena in PIM-1 [147]. For thin films (<1000 nm), aging was found to be the dominant effect even under highly plasticizing conditions. For a 30 μm-thick film of PIM-1 tested at 32 bar CO₂ pressure, the CO₂ permeability increased by 37% over the course of 100 h due to plasticization. Conversely, for a 270 nm-thick membrane, the CO₂ permeability dropped by 95% over the same time period due to aging and densification.

Mitigating physical aging in highly permeable polymers is an ongoing research effort. Iptycene-based PIMs with ‘configurational’ free volume have displayed considerable improvements in aging behavior in bulk films [141,151,152]. Similarly, the use of mixed matrix membranes and post-crosslinking strategies have been reasonably successful in slowing down aging in bulk films [153,154,155]. It has been suggested that thermal annealing of thin films at 200–300 °C would allow for a reasonable approximation of aged behavior, which would speed up lab-scale studies of aging. [43,156].

5.4. PEO-Based Polymers

Polymers incorporating the ether moiety are one of the few reverse-selective rubbery polymers that consistently feature on the upper bounds for acid gas separation [27]. The polar ether moiety can provide favorable sorption interactions for both H₂S and CO₂, making ether-based polymers a natural choice for sour gas separations [89].

The investigation of ether-based polymers for solubility-based separations dates back to 1970s [157]. Selexol^™, a physical solvent composed of dimethyl ethers of tetraethylene glycol (PEG ethers), was commercialized in 1969 [157]. The solvent is highly selective for the acid components in natural gas and can be regenerated at low temperatures [158]. It was quickly adopted in acidic gas fields where corrosion at high temperatures made the use of amine absorbents problematic [158,159,160]. Membrane researchers in the 1980s attempted to harness the solubility-based separation offered by PEG by developing blend membranes and supported liquid membranes [100,161,162]. However, despite good separation performance, their relatively poor mechanical properties limited their applications [100].

Around the same time, Pebax^®, a polyether–polyamide block copolymer shown in Figure 14, was commercialized for the medical and electronics industries [163]. These polymers were noted to have good mechanical strength and film-forming properties, and they soon made their way into membrane separations for pervaporation and filtration applications [164,165,166]. Pebax^®-based spiral-wound membrane modules were tested at the bench and pilot scale for H₂S and CO₂ removal from natural gas in 1993 by researchers at Membrane Technology and Research (MTR) [5,167,168]. However, stability concerns, as well as the poor economics of sour gas extraction, made PEO-based membranes compare unfavorably against the CA-based membranes that controlled the market at the time.

In view of such disadvantages, researchers investigating PEO membranes shifted focus to CO₂ removal from syngas and flue gas. Research into PEO membranes gained traction in the late 1990s, when global warming concerns spurred research into carbon capture. A number of papers establishing fundamental structure/property relationships in PEO were published shortly thereafter [169,170,171]. PEO-based membranes have been successfully tested at the pilot scale for CO₂ capture from syngas and flue gas [162,172,173]. Commercial spiral-wound modules are available for flue gas and syngas carbon capture [162]. Their technological maturity and the push for developing sour gas wells has renewed interest in developing PEO-based membranes for natural gas separation.

The rubbery PEO membranes provide solubility-based separation, with exceptional H₂S/CH₄ separation performance, but with poor CO₂/CH₄ separation performance. Their size sieving ability is very low, as evinced by a CO₂/CH₄ diffusivity selectivity of ca. 1 [174,175,176,177,178]. They possess CO₂/CH₄ solubility selectivity of 15, which enables an overall CO₂/CH₄ selectivity of 10–15, along with CO₂ permeabilities of 100–500 Barrer [174,175]. Limited data exist for the contributions of solubility and diffusivity to H₂S/CH₄ selectivity, but estimates suggest a solubility selectivity of around 70 and a diffusivity selectivity of ca. 1 [45,179]. In liquid PEGs, the H₂S/CH₄ selectivity can be up to 120 [176,177,178]. This is consistent with the reported H₂S/CH₄ selectivities of PEO-based membranes, which range from 40 to 120, with H₂S permeabilities of 500–1000 Barrer [21].

It should be noted that pure PEO is highly crystalline and is not generally used for membrane studies [89]. Instead, researchers use block copolymers of PEO with PAs, polyurethanes (PU), polyimide, and other rigid segments [53,180,181]. For H₂S/CH₄ separations, PA- and PU-based block copolymers are preferred, since the amide and carbonyl groups can enhance H₂S solubility [53,89,180]. Additionally, these copolymers exhibit improved plasticization resistance due to hydrogen bonding between the amide and carbonyl groups [50,182]. Blends with short chain PEGs or other polymers have also been investigated, but blend membranes are not preferred for natural gas separations [105,183,184].

The main obstacle facing the extension of PEO-based membranes to sour gas separations is their poor CO₂/CH₄ selectivity. This restricts their application to a very narrow range of natural gas compositions. Even a mild increase in CO₂/CH₄ selectivity can substantially impact their economics and expand the range of natural gas compositions over which PEO-based membranes can be applied [5,73,82,185].

Such an improvement is unlikely to arise from solubility selectivity. An analysis conducted by Lin and Freeman has shown that the CO₂/CH₄ solubility selectivity of PEO polymers is already close to the theoretical maximum possible for rubbery polymers [89]. Lin and Freeman had collated and examined solubility data of H₂S, CO₂, and CH₄ to identify functional groups which could further improve solubility selectivity. Figure 15a,b present their findings with plots on the variations of CO₂/CH₄ and H₂S/CH₄ solubility selectivities with the Hansen solubility parameter, δ. As shown in Figure 15a, CO₂/CH₄ solubility selectivity reaches a maximum at a solubility parameter value of around 22.5 MPa^0.5. The solubility parameter of Pebax^® polymers is around 20 MPa^0.5, very close to the optimal solubility parameter for the CO₂/CH₄ gas pair. The design of new polymer structures may improve H₂S/CH₄ selectivity, which peaks at a solubility parameter value of 26 MPa^0.5 as depicted in Figure 15b. However, it is unlikely to result in any significant improvement of CO₂/CH₄ solubility selectivity.

Hence, tailoring PEO polymers for sour gas separation will necessitate a focus on improving chain rigidity and diffusivity selectivity. This can be conducted through the use of crosslinked PEOs with rigid chain structures that provide size sieving selectivity. Harrigan et al. reported crosslinked PEGs with exceptionally high CO₂/CH₄ selectivity [52]. They used a trifunctional isocyanate crosslinker to crosslink PEGs with molecular weight (MW) ranging from 200 to 2000 g mol⁻¹. The reaction scheme is depicted in Figure 16.

Figure 17 presents the results obtained by Harrigan et al [52]. At PEG MWs of below 500 g mol⁻¹, crosslinking increased chain rigidity to the point that the polymer was glassy at ambient temperature. This allowed for substantial improvements in size selectivity, with H₂S/CH₄ and CO₂/CH₄ selectivities in the range of 92–65 and 40–58, respectively. However, the permeability was far lower than typical rubbery PEO, at below 1 Barrer for both H₂S and CO₂. At PEG MWs in the range of 600–1000 g mol⁻¹, the polymers were rubbery, with H₂S/CH₄ and CO₂/CH₄ selectivities of 105 and 23, along with H₂S and CO₂ permeabilities of 9–25 and 2–6 Barrer, respectively. Over PEG MWs of 2000 g mol⁻¹, the polymer was semicrystalline, leading to a drop in permeability. The H₂S and CO₂ permeability dropped to 3 and 1 Barrer, respectively.

Given the low acid gas permeabilities, the permeation data obtained by Harrigan et al. lie well below the upper bounds [52]. However, their CO₂/CH₄ selectivity is the highest reported for PEO-based polymers, implying that there may yet be an ideal balance to be found between acid gas permeability and CO₂/CH₄ selectivity. Crosslinking appears to be a promising strategy for the design of PEO-based polymers for natural gas separations.

Another, rather different, approach to improving the CO₂/CH₄ selectivity of PEO-based membranes can be found in the process design aspect. For solubility-based membrane separations, the operating temperature has a strong impact on selectivity [48]. Figure 18a shows the CO₂/CH₄ upper bounds for rubbery polymers at 35 °C and −20 °C. The upper bounds shift upwards as temperature drops, suggesting higher selectivities can be obtained at lower temperature. This trend is consistent with reported results. Figure 18b exhibits the CO₂/CH₄ separation performance of a crosslinked PEG reported by Lin et al. at temperatures ranging from 45 °C to −20 °C [187]. The CO₂/CH₄ selectivity increased to over 70 at −20 °C. The CO₂ permeability also decreased correspondingly, to 4 Barrer. However, under practical operating conditions, plasticization occurring at high acid gas partial pressures may be sufficient to increase permeability to workable levels [188].

Sub-ambient operation has been instrumental in the successful application of PEO membranes for CO₂/H₂ separation [172,173,188]. If similar studies can be extended to natural gas, it would allow currently commercial PEOs to be applied to sour gas processing without the need for additional material development efforts.

From a process design point of view, natural gas separations are well-suited to sub-ambient operation. Many of the pretreatments needed to remove condensable contaminants require chilling [6,189]. As a result, the incoming feed to the membrane is often below ambient temperature [6,189]. Moreover, membrane processes for natural gas are subject to Joule–Thomson cooling, caused by the expansion of the acid gas as it moves from the high-pressure feed side to the low-pressure permeate side [190]. As a result, the membrane process operates at 10–20 °C lower than the feed temperature, without the need for additional cooling.

In the past, attempts to harness Joule–Thomson cooling to improve performance were generally unsuccessful [5,6,189,190]. The temperature drops proved difficult to control and occasionally caused condensation and fouling, raising stability concerns. However, membrane research and module design have come a long way since then. Sub-ambient natural gas treatment has since been demonstrated on the pilot scale using CA membranes [189]. PEO membranes have also been operated at −5 °C on the pilot scale for syngas carbon capture [188]. There are other reports of successful sub-ambient membrane operation [191,192]. Designing membranes for sub-ambient operation would require a focus on membrane fabrication and process and module designs [192]. Such efforts are usually far more successful than material development efforts [193]. If sub-ambient temperature operation proves economically viable and can be successfully accomplished, it would significantly increase the competitiveness of PEO membranes in the natural gas market [5,190].

5.5. H₂ Production from H₂S and Prospects for Polymeric Membranes

As the natural gas industry continues to shift towards sour gas extraction, the Claus process is gradually becoming uneconomical [29,32]. The need for alternative H₂S disposal methods is now a pressing issue. The feasibility of large-scale implementation of such treatments is being actively investigated in areas with high concentrations of sour gas deposits [29]. In particular, H₂ production via H₂S decomposition has attracted significant attention from researchers and policy-makers alike [33,194]. This method is especially advantageous in that the by-product is still elemental sulfur, a chemical essential to the fertilizer and agrochemical industries. High-value H₂ can then be sold to improve the overall process economics.

This approach has gained traction in recent years, and various processes have been investigated by researchers. While a detailed assessment of such technologies is outside the scope of this work, recent publications provide a comprehensive overview [33,194]. A number of thermal and thermo-catalytic methods for H₂S decomposition have shown good promise for large-scale sulfur treatment. These technologies are all still in the research and development phase, and there are yet several issues to be addressed [33,194]. Despite this, they show good potential to be scaled up in time to match the projected peak in natural gas production [21]. Should economic and regulatory factors prove favorable, the following decades may see the emergence of large-scale plants based on such processes.

One feature of these methods is that the H₂S conversions are in the range of 20–50%, depending upon the reactor configuration and conditions [33,194]. H₂S decomposition is an endothermic reaction, and the equilibrium conversion is controlled by the temperature. Even at temperatures as high as 1000–1200 °C, the conversions are relatively low, around 20–40%. Achieving complete decomposition of H₂S in a single pass would require prohibitively high temperatures and is not feasible. Instead, process designs usually incorporate a H₂/H₂S separation unit and recycle unreacted H₂S back to the reactor. This allows near-total conversion and circumvents the need for additional desulfurization processes downstream to treat unreacted H₂S [195]. Figure 19 shows an example of such a process design, wherein a two-stage membrane process is used to produce high-purity H₂ and recycle unreacted H₂S back to the reactor [196].

The lack of adequate technologies for separation of H₂ from H₂S has always been a major obstacle for the commercial implementation of such processes [201]. Technologies such as absorption and cryogenic separation have generally resulted in unfavorable economics [201,202]. Membrane systems would likely prove more economical, but previously, very few membranes could offer workable selectivity for this separation. A techno-economic analysis study of H₂ production from H₂S conducted by Cox et al. in 1998 highlighted the shortcomings of commercial membranes available at the time [201]. These membranes (likely based on polyimide) possessed poor H₂/H₂S selectivities ~4 and were unsuited for H₂S concentrations exceeding 10%, severely limiting their applicability. The poor separation performance proved detrimental to the overall process economics.

The membrane industry has changed significantly since then. Though very few research publications discuss H₂/H₂S separation, today’s commercial membranes will likely have much better options for the separation of H₂S and H₂ [203]. Figure 20 shows the theoretical upper bounds for the H₂/H₂S and H₂S/H₂ gas pairs. H₂ possesses a smaller kinetic diameter (2.89 Å) and lower condensability ($\raisebox{1ex}{$\epsilon $}\!\left/ \!\raisebox{-1ex}{$k$}\right.$ = 60 K) than H₂S [48]. Accordingly, glassy polymers can exhibit size-based H₂/H₂S selectivity. The upper bound in Figure 20a shows the typical trade-off characteristic to size-selective separations. High selectivity is unlikely to be achieved at ambient temperature without a compromise on permeability. Since H₂S decomposition requires temperatures in excess of 700 °C, the use of higher temperatures may be an option to enable higher selectivity [37,48]. Even so, given the rather niche application and the scarcity of data in the literature, this is not likely to be a worthwhile avenue of investigation.

On the other hand, H₂S/H₂ separation is a more forgiving separation. H₂S/H₂ selectivity can be achieved through rubbery polymers with high H₂S solubility. As depicted in Figure 20b, the solubility-based upper bound is not subject to the usual selectivity-permeability trade-off, and achieving high selectivity should be far easier. Reports suggest that PEO-based membranes can manage decent H₂S/H₂ selectivities of 20–40 at room temperature [172,203,204]. Such membranes have already been available for syngas and flue gas carbon capture. Detailed studies of separation performance, process design, and techno-economic analysis are still necessary to evaluate their feasibility. Even so, should economic factors prove favorable, PEO-based membranes, though limited in their applications to natural gas processing, may well find applications in the adjacent industries.

6. Concluding Remarks

Natural gas is a key source of hydrogen, and an important bridge fuel on the path to renewable energy [3,4]. This review provides an update on the current state of research regarding polymeric membranes for natural gas purification.

Natural gas processing is the largest worldwide gas separations market and presents huge untapped opportunities for membrane researchers [14,21,25]. With growing interest in the development of highly acidic sour gas fields, the market share of membranes in natural gas processing is poised to grow rapidly in the future [14,21,25]. The interest in sour gas separations has been reflected in membrane research, and we have seen a surge in the number of publications discussing the development of membrane materials for ternary H₂S/CO₂/CH₄ separation in the past decade, presented in Figure S1 in the SI. This review gives a brief overview of the research directions pursued by membrane scientists targeting simultaneous removal of H₂S and CO₂ from natural gas.

The development of membranes with high selectivity under plasticizing conditions can significantly improve the competitiveness of membranes in the natural gas market.

Table 2. Polymeric membranes for natural gas purification.

Polymer	H2S/CH4 Selectivity	CO2/CH4 Selectivity	Strategies for Improving Sour Gas Performance
CA	20–25	15–20	Limited scope for improvement. Best suited to sweet natural gases.
Polyimide	20–25	~30	Crosslinking to improve resistance to H2S plasticization.
Microporous polymers	>50	~10	Incorporation of functional groups to improve CO2 and H2S solubility. Limited scope for improving CO2/CH4 selectivity.
PEO	>50	~10	Crosslinking, sub-ambient operation may improve CO2/CH4 selectivity.

provides an overview of the acid gas selectivities of commonly investigated polymer materials under sour gas conditions, along with promising strategies to improve selectivity. CA and polyimide membranes hold the bulk of the natural gas membranes market [6]. While research into CA membranes has stalled, polyimides remain popular among membrane researchers. The development of new polyimide structures with improved rigidity and reduced packing efficiency for sour gas separations is an area of ongoing research [125,127,205]. Crosslinking strategies can improve the H₂S plasticization resistance of polyimides, thus enabling a moderate CO₂/CH₄ selectivity of ca. 30 under sour conditions [69,132]. Their good CO₂/CH₄ separation performance makes them a suitable candidate for CO₂-rich natural gases with low H₂S content.

Designing polymers with high H₂S/CH₄ selectivity for H₂S-rich sour gases is an area of active investigation. While PIMs and poly(norbornenes) have shown very promising H₂S/CH₄ separation performances, these materials are still at an early stage of their development [64,144]. PEO-based membranes are currently the most promising commercial candidate for highly sour gases, but their low CO₂/CH₄ selectivity limits their competitiveness. Improvements in CO₂/CH₄ selectivity, through strategies such as crosslinking or sub-ambient operation, can open up the natural gas market to these membranes. Additionally, PEO-based membranes have the potential to considerably improve the commercial viability of technologies for the production of H₂ from H₂S, a topic that has attracted significant attention in recent years. The superior H₂S/H₂ separation offered by these membranes may play a pivotal role in the commercial implementation of such processes.