Advances in membrane-based helium recovery from natural gas: materials, process integration, and techno-economic perspectives

1 Introduction

1.1 Strategic importance and applications of helium

Helium is a chemically inert noble gas with an exceptionally low boiling point (4 K), making it a critical enabling fluid for cryogenic science and engineering and for applications that require stable, contaminant-free operation. Its low molar mass (4 g mol^-1) and favorable thermophysical properties (high thermal conductivity and specific heat, low density) support efficient heat transfer, rapid thermal cycling, and controlled inert atmospheres in sensitive processes (Pu, 2025). Accordingly, helium is used across cryogenics, leak detection, welding, semiconductor manufacturing, optical fiber drawing, and medical imaging, among other high-value sectors (Berganza and Zhang, 2013; Serra Leal et al., 2023; Veligura et al., 2015). Figure 1 summarizes the major end-use sectors of helium. Rather than reiterating end-use descriptions in detail, the key point for separation and supply-chain discussions is that several major helium consumers including advanced manufacturing and low-temperature research have limited practical substitutes and depend on reliable, continuous helium availability (National Research Council, 2000). This dependency increases the strategic value of technologies that can stabilize supply by improving recovery efficiency from low-grade sources.

1.2 Helium recovery challenges and technological approaches

Despite helium’s cosmic abundance, economically recoverable terrestrial helium is rare and is typically present at only 0.3–1.0 vol% in natural gas (NG) reservoirs (Elvira and Laura Annamaria, 2023). Depletion of high-grade fields has shifted attention toward leaner resources, increasing technical complexity and tightening economics. In addition, reserves and production capacity are geographically concentrated, creating supply vulnerability. Figures 2, 3 illustrate this concentration and the resulting imbalance between limited producers and widely distributed demand, which motivates diversified sourcing and improved recovery from lean feeds (Qin et al., 2024; Zheng et al., 2023). Technology selection is strongly governed by feed composition. In CH4-rich streams, methane influences refrigeration levels and cryogenic cut points; in CO₂- and especially N₂-rich systems, helium–nitrogen behavior can dominate flowsheet constraints and determine the feasibility of pre-enrichment steps (Hamedi, 2021; He et al., 2024). Below 0.3% He, energy requirements rise sharply because large gas volumes must be compressed or cooled for a small product output, making process integration and pre-enrichment decisive for viability (Hamedi et al., 2019a). Cryogenic distillation remains the industrial benchmark for moderate-to-rich feeds and can deliver ≥99.99% helium purity, but with high capital and energy costs. Membranes can provide continuous pre-enrichment (often to 10%–20%), yet their single-stage performance is constrained by selectivity limits and by susceptibility to plasticization or fouling under real-gas conditions. PSA can be effective for pretreatment and small-scale recovery, but performance and economics often deteriorate as helium becomes leaner (Shafaei and Mehrpooya, 2018). Absorption can be relevant for specific CO₂-rich cases but may incur regeneration penalties (Sidhikku Kandath Valappil et al., 2021).

For lean feeds, hybrid flowsheets particularly membrane–cryogenic and PSA–cryogenic sequences are frequently proposed to combine the strengths of each unit operation, reduce total cost, and preserve recovery (Quader et al., 2021). In LNG chains, helium-bearing tail gases are often pre-enriched prior to cryogenic polishing by leveraging existing refrigeration and nitrogen-rejection assets, while careful selection of pretreatment level, operating pressure, reflux ratio, and staged enrichment can substantially reduce power demand and off-design losses (Chahartaghi and Sheykhi, 2019; Hamedi, 2021). Overall, the central challenge is not the availability of separation options, but the ability to sustain performance and economics under realistic compositions, pressures, and impurity burdens across widely varying resource grades (Qin et al., 2024).

1.3 Membrane-based helium recovery: technologies and research trends

The combination of declining high-grade reserves and increasing demand has intensified interest in efficient helium recovery from NG and industrial streams. While cryogenic distillation and PSA remain established, their energy intensity and reduced effectiveness for lean helium feeds have motivated increased attention to membrane-based approaches (Alders et al., 2017). Membranes exploit helium’s small kinetic diameter and low polarizability to enable continuous pre-enrichment under moderate conditions and can be integrated into LNG facilities that generate helium-bearing tail gases (Quader et al., 2021; Sunarso et al., 2017). At the materials level, polymeric membranes (notably polyimides and perfluoropolymers) remain prominent but are constrained by the Robeson upper bound and by performance degradation risks under realistic mixed-gas, high-pressure feeds. Mixed-matrix membranes (MMMs) incorporating zeolites, MOFs, or carbon molecular sieves seek to improve transport pathways and raise helium flux while maintaining processable mechanical integrity (Chen et al., 2016; Ghasemzadeh et al., 2016). At the process level, hybrid membrane–cryogenic flowsheets are increasingly studied because pre-concentrating helium from 0.1%–1% to 10%–20% can substantially reduce downstream refrigeration requirements and improve overall efficiency across differing feed compositions (Brown, 2019; Qin et al., 2024). Bibliometric indicators are consistent with this shift. Figure 4 is included to map the structure of the helium-recovery literature rather than to benchmark separation performance. The density distribution shows that research activity concentrates around three dominant themes: (i) cryogenic and adsorption-based routes aimed at high-purity helium production, (ii) membrane-focused studies centered on materials development and durability, and (iii) systems-level design and evaluation, including optimization and techno-economic assessment. Notably, these themes occupy partially separated high-density regions, suggesting that advances reported at the material level are not consistently linked to module-scale constraints, flowsheet integration, or cost and energy drivers. This observation motivates the integrative scope of the present review, which explicitly connects membrane material performance and stability to process configuration choices and techno-economic outcomes under natural-gas-relevant conditions. While Figure 5 suggests a temporal transition from early cryogenic and PSA-centered work toward recent studies emphasizing membrane systems, optimization, and sustainability. These trends support the growing view that membranes will be most competitive when assessed as part of integrated, impurity-aware process systems rather than as isolated material-performance demonstrations.

Despite progress, much of the literature still treats membrane material development, transport properties, process configurations, and economic performance in a fragmented manner. This fragmentation makes it difficult to interpret whether reported material-level advances are likely to translate into robust module-scale performance, viable hybrid process integration, and competitive techno-economics under realistic NG conditions (He et al., 2024; Scholes et al., 2017). In particular, systematic links between membrane transport behavior, module implementation constraints, hybrid flowsheet design, and cost/energy outcomes remain insufficiently consolidated (Hamedi et al., 2019a). This review addresses these gaps by examining membrane-based helium recovery across three connected levels. First, it synthesizes helium-relevant transport mechanisms and material properties with emphasis on performance under realistic conditions. Second, it evaluates module design and the role of membranes in standalone and hybrid flowsheets, highlighting integration constraints and performance trade-offs. Third, it compares techno-economic outcomes of membrane-assisted systems against cryogenic and adsorption-based options across a range of feed compositions. By integrating recent research trends and industrial insights, the review aims to clarify when membrane-assisted helium recovery can progress from promising pilot-scale concepts toward scalable, energy-efficient, and economically competitive solutions. Table 1 places the present review in context by comparing representative surveys published in recent years with respect to scope (materials, process, economics, and design), feedstock emphasis (direct NG, LNG/NRU-related streams, and crude-He upgrading), and the distinct contribution of each work. The comparison suggests that prior reviews have most often advanced either a materials-oriented perspective or a process and techno-economic viewpoint; however, fewer provide a unified treatment that consistently links membrane transport performance and stability to module- and flowsheet-level integration constraints and, ultimately, to cost and energy outcomes under NG-relevant conditions. Building on this need, the present review offers an integrated synthesis that connects material-level evidence with process integration choices and techno-economic implications, while also addressing deployment considerations and technology readiness for membrane-based helium recovery.

TABLE 1. Positioning of the present review against recent helium and membrane-separation reviews.

Scope: M = membrane materials/transport; P = process configurations and integration; E = techno-economics/energy; D = deployment/readiness evidence.

2 Helium recovery and extraction methods

2.1 Cryogenic separation processes

Cryogenic separation remains the most advanced and widely used method for large-scale helium extraction. These systems depend on helium’s non-condensable behavior at cryogenic temperatures and its extremely low boiling point (4.2 K). In LNG and nitrogen removal unit (NRU) facilities, helium recovery is typically achieved by separating the helium-enriched vapor from the end-flash gas after methane liquefaction. Conventional industrial processes—licensed by APCI, Linde, Technip, and ExxonMobil—involve multistage flash separation, distillation, or both, depending on feed composition and the level of integration. Feed gas is initially pretreated to remove CO₂, water, sulfur compounds, and heavy hydrocarbons. It is then cryogenically cooled to temperatures below −160 °C, which produces a helium-rich vapor stream (0.one to one mol% He) and a condensed methane-rich liquid. Further distillation or polishing with PSA units can raise helium purity to over 90%, with overall recoveries surpassing 95% (Ansarinasab et al., 2017; Mehrpooya and Shafaei, 2016; Van Sciver, 2012). A comparative summary of conventional helium extraction routes and their typical performance metrics is provided in Table 2.

TABLE 2. Summary of conventional helium extraction methods from natural gas, including separation principles, performance, and typical applications.

Energy demand is reported qualitatively due to strong dependence on feed composition, pressure levels, and process integration.

Kim’s comparative study of LNG-integrated cryogenic flowsheets showed that both flash- and distillation-based processes achieve high helium recovery without sacrificing LNG throughput. However, these systems require significant energy input. Exergy analyses by Mehrpooya and Shafaei (2016) identified compressors and multi stream cryogenic heat exchangers as primary sources of exergy loss, highlighting opportunities for efficiency improvements through better refrigerant cycles and heat integration. Building on this, Ansarinasab et al. (2017) used an advanced exergy costing method on a modified Linde-type process and found that avoidable exergy loss in compressors accounted for most inefficiency costs. Pakzad et al. (2021) also showed that optimized cryogenic flowsheets with improved refrigeration cycles and staged expansion could produce 196.5 kg h^-1 of liquid helium at 23.4% exergy efficiency.

Despite their unmatched ability to deliver ultra-high helium purity (>99.99%) and high recovery, cryogenic systems are capital-intensive, energy-demanding (1.5–2.0 MJ mol^-1 He), and operationally complex. Their economic viability is therefore highest in large-scale LNG facilities where helium concentrations exceed 0.1 mol%. For leaner feeds, hybrid systems that combine cryogenics with adsorption or membrane polishing have been developed to cut energy use and boost cost-effectiveness (Ansarinasab et al., 2017; Peng et al., 2024).

2.2 Adsorption-based processes

Adsorption technologies offer a flexible, non-cryogenic alternative for small-to medium-scale helium recovery. PSA, VSA, and TSA systems utilize differences in gas–solid interaction energies: helium weakly interacts with common adsorbents, allowing it to pass through the bed while heavier components (CH₄, N₂, CO₂) are trapped. Key adsorbents include zeolites, activated carbons, and MOFs, selected for their high surface area and adjustable pore structures.

Dual Reflux PSA (DR-PSA) is a significant recent advancement. In experiments by Weh et al. (2022a), a DR-PSA system using binder-free zeolite 13X achieved helium purities of up to 99.999 mol% with 95% recovery from feeds containing only 1 mol% He. When set up as a two-stage cascade with recycle, energy consumption decreased to about 1.5 MJ mol^-1 He, nearing cryogenic performance. Follow-up work by the same group extended these results to CH₄–N₂–He mixtures typical of marginal gas fields, showing 90%–98% recovery at feed concentrations as low as 0.1 mol% He.

Adsorption-based separation is a practical route for helium recovery from natural gas derived streams, not because helium is strongly adsorbed, but because suitably chosen porous solids can preferentially retain the dominant non-helium constituents, allowing helium to exit as the light product. In industrial practice, helium is often encountered as a dilute component in upstream reject or vent streams most notably nitrogen rejection unit (NRU) off-gas so separation behavior is governed by multicomponent competition and impurity partial pressures rather than by single-gas selectivity metrics alone (Rufford et al., 2014). Under such dilute conditions, the feasibility and performance of PSA-type schemes depend strongly on cycle configuration and operating windows, as shown for zeolite 13X–based cycles where recovery and purity respond sharply to step scheduling and pressure levels (Jahromi et al., 2018; Weh et al., 2022a).

This makes front-end conditioning a controlling design element: deep dehydration and tight control of strongly adsorbing species particularly H₂O and CO₂, and, where relevant, heavier hydrocarbons/trace contaminants are routinely treated as prerequisites to avoid rapid loss of effective capacity, displacement of mass transfer fronts, and slow to regenerate fouling that undermines cycle to cycle reproducibility (Long et al., 2025; Rufford et al., 2014). At the unit level, PSA/VPSA/PVSA outcomes are dictated by the coupled effects of pressure ratio (including the extent of vacuum assistance in VPSA/PVSA to enhance desorption for dilute feeds), competitive adsorption under realistic impurity envelopes, and mass transfer kinetics within pellets and along the packed bed. Consequently, even modest excursions in impurity loading such as increased N₂/CH₄ fractions, intermittent CO₂ spikes, or moisture slip can materially alter breakthrough behavior, reduce working capacity, and tighten the purity recovery productivity trade-off, especially under fluctuating feeds (Jahromi et al., 2018; Weh et al., 2022b). Accordingly, adsorption performance claims are most interpretable when reported alongside cycle-defining details (adsorption/desorption pressures, equalization strategy, purge composition and rate, and step times) and a transparent description of pretreatment robustness and feed variability, which together determine industrial operability. In this sense, adsorption is most persuasive when positioned as a well-bounded block preconcentration or polishing of a specified reject stream within a broader flowsheet where the impurity envelope is defined and can be engineered around (Jin et al., 2026; Rufford et al., 2014; Xiao et al., 2023). Finally, while zeolite 13X remains a practical benchmark for helium-service PSA studies, emerging tailored adsorbents) should be discussed in terms of deploy ability criteria that dominate natural-gas derived helium recovery: tolerance to moisture/acid-gas upsets, regeneration completeness under realistic purge conditions, and multi-year cyclic durability under compositionally variable operation (Cui et al., 2025; Weh et al., 2022b).

2.3 Membrane-based separation processes

Membrane-based helium separation has attracted significant interest as a compact and energy-efficient alternative to traditional cryogenic or adsorption methods. Helium’s small kinetic diameter (2.6 Å) and high diffusivity naturally favor solution–diffusion membranes. Polymeric options such as cellulose acetate, polysulfone, and polyimides are currently the most common. However, they often exhibit modest selectivities, such as He/N₂ or He/CH₄ (10–40), and require multistage cascades to attain high purity (Rafia et al., 2024).

Membrane processes are especially beneficial for bulk separation due to their low cost. Rufford et al. (2014) observed that helium-selective membranes could replace or support cryogenic units in NRU off-gas treatment. However, to achieve purities above 95 mol%, at least three membrane stages and inter-stage compression are typically needed, which increases both capital and operational expenses.

Hybrid membrane–PSA and membrane–cryogenic systems have demonstrated strong potential in overcoming these limitations. Hamedi et al. (2019a) showed that a two-stage membrane sequence combined with a compact PSA unit could reach 99.9% helium purity with energy requirements of only 0.4–0.7 kWh·m^-3. Peng et al. (2024) introduced a cryogenic–membrane scheme using post-expansion nitrogen refrigeration, which cut compressor energy use by 23% and achieved 99.6% helium purity while also recovering ethane.

Inorganic and mixed-matrix membranes (MMMs) incorporating zeolites, graphene oxide, or carbon molecular sieves represent a growing research focus due to their improved thermal stability and permeability–selectivity balance (Alinejad et al., 2022). The research landscape is illustrated in Figure 6, where dense keyword clusters around “molecular sieves,” “permeability,” and “selectivity” highlight the central role of material-level transport mechanisms. Clusters involving MMMs reflect increasing interest in hybrid polymer–inorganic architectures capable of surpassing the Robeson upper bound.

These developments highlight the central role of membrane materials and module design in determining the feasibility of membrane-assisted helium recovery.

2.4 Hybrid and emerging technologies

Hybrid systems are increasingly positioned as the most viable route for helium recovery from lean feeds and decentralized deployments, but the literature often blurs an important distinction between conceptual promise and demonstrable operability under realistic variability (Rufford et al., 2014; Scholes and Ghosh, 2017). Cryogenic and membrane hybrids, as well as PSA and membrane combinations, are repeatedly proposed because they distribute the separation burden across unit operations with complementary strengths, yet their practical advantage is strongly contingent on integration choices, compression and refrigeration duties, and the achievable intermediate concentration between stages (Dai et al., 2021; Rufford et al., 2014; Xiao et al., 2023). DR-PSA cascade concepts have notably expanded the design space for non-cryogenic upgrading and are frequently cited as enabling high recovery even for marginal resources, although their competitiveness remains sensitive to the pressure level available at the source and the resulting utility intensity (Diagne et al., 1995; Weh et al., 2022a). Exergy-based assessments of cryogenic and adsorption hybrid schemes report thermodynamic efficiencies exceeding 40 percent, but cross-study comparability is limited unless system boundaries and utility accounting are harmonized with LNG-integrated benchmarks (Ansarinasab et al., 2017; Mehrpooya and Shafaei, 2016; Pakzad et al., 2021). As shown in Figure 7, membrane-based options present the most favorable balance across energy use, cost, and environmental indicators within the compared scenarios, whereas cryogenic systems remain the reference for highest purity and recovery, and PSA retains its main value proposition in flexibility and scalability for small-to-medium capacity upgrading and polishing services (Ma et al., 2025; Xiao et al., 2023).

On the materials side, the expected gains from advanced MOFs, graphene-related membranes, and dynamically intensified adsorption concepts are credible but uneven across maturity levels. Comprehensive membrane reviews emphasize that materials advances must be paired with process configuration choices that reduce compression demand and concentration-polarization penalties if cost and energy advantages are to translate at scale (Akbari et al., 2020a; Zhu and Xu, 2014). Graphene-based and other two-dimensional membranes are frequently highlighted for their theoretical potential to push permeability and selectivity simultaneously, yet the literature also underscores the gap between simulation level performance and manufacturable, defect-controlled modules (Yoo et al., 2017; Zarabadi-Poor and Marek, 2019). For adsorption, MOF development continues to expand the design space for selectivity and regeneration performance, with both scalable MOF demonstrations and large-scale screening studies suggesting pathways beyond conventional zeolites, while still leaving durability and deployment constraints as the central test for practical adoption (Cui et al., 2025; Zarabadi-Poor and Marek, 2019). Finally, comparative assessments that include economic and sustainability lenses generally place membrane-centered options favorably for modular deployment, while cryogenic routes remain the reference for highest purity and recovery, and PSA retains a strong niche for flexible small-to-medium scale polishing and upgrading duties (Fiedler et al., 1994; Ma et al., 2025; Quader et al., 2021).

2.5 Single-gas vs. mixed-gas behavior at high pressure

Membrane performance for helium separation often differs markedly between idealized single gas tests and realistic mixed gas conditions. In pure gas experiments at low pressure, many membranes exhibit high helium permeabilities and attractive selectivities (He over N₂ or CH₄). For example, advanced polyimide membranes have achieved He permeabilities 50 Barrer with ideal selectivities up to 146 for He/N₂ (Gao B et al., 2025; Nezhad et al., 2020) and even higher He/CH₄ selectivity, 269 in one crosslinked case (Wang et al., 2025). However, under natural gas like feeds containing CO₂, CH₄, N₂, H₂S, and heavier hydrocarbons at elevated pressures, these performance metrics often deteriorate. Strongly sorbing components like CO₂ and H₂S can competitively adsorb in polymer free volume, reducing the effective helium permeability and selectivity (Al-Juaied and Koros, 2006a; Wang et al., 2025). In glassy polymers (such as polyimides), high CO₂ partial pressure causes plasticization, a swelling of the polymer matrix that raises overall gas flux but lowers He selectivity (i.e., the membrane becomes less discriminating) (Park and Lee, 2005; Wang et al., 2025). For instance, an asymmetric hollow-fiber polyimide showed a sharp drop in mixed gas selectivity when exposed to 40 bar CO₂/CH₄/He feeds, unless it was stabilizing crosslinked (Lei et al., 2020). In mixed feeds, helium also faces dilution and competition effects, e.g., the presence of 49.7% CO₂ alongside CH₄ (50%) and trace He (0.3%) at 40 bar caused a 24% drop in the CO₂/CH₄ selectivity of a baseline polyimide, whereas the He/(CO₂+CH₄) selectivity actually increased by 80% upon raising temperature with a crosslinked membrane (Wang et al., 2025). This indicates that proper material modifications can partly offset mixed gas performance losses. By contrast, microporous inorganic membranes (zeolites, CMS) often rely on molecular sieving and adsorption differences, which can either favor helium or the co-components. For example, silicate (MFI) zeolite membranes at cryogenic temperature preferentially permeate CH₄/N₂ over He (adsorbing the larger molecules and yielding a He-rich retentate) (Yu et al., 2022). Thus, depending on the mechanism, a membrane may produce a helium enriched permeate (as is common for polymers where He diffuses fastest) or a helium enriched residue (as with some adsorption selective sieves) under NG conditions. In all cases, systematic mixed gas testing is crucial. Pure gas selectivities can drastically overestimate performance: such as light hydrocarbons or acid gases in the feed can reduce the He/N₂ and He/CH₄ selectivities of polyimide and cellulose acetate membranes by competing for free volume and inducing swelling (Al-Juaied and Koros, 2006a; Catalano et al., 2012). Even minor amounts of heavy condensable vapors (such as toluene, C₇ ⁺) can lead to complex behavior at 55 °C, 300 ppm toluene added to a CO₂/CH₄ mixture increased CO₂/CH₄ selectivity (due to blocking of larger gas transport) whereas at 35 °C the same contaminant caused selectivity to drop, consistent with polymer swelling (plasticization) dominating at lower temperature (Al-Juaied and Koros, 2006a; Sazali et al., 2018). In practice, mixed gas performance at high pressure is often lower than ideal values, and membranes must be tailored or pretreated to retain performance under these aggressive conditions.

3 Permeation theory and governing parameters

Membrane-based helium recovery, discussed in Section 2 as a key alternative to cryogenic and adsorption technologies, mainly depends on gas transport through dense polymeric or inorganic media. The solution–diffusion model offers the standard framework for describing this transport and remains the foundation for evaluating membrane materials, predicting module performance, and optimizing large-scale separation processes (Ghasemzadeh et al., 2019).

3.1 Temperature dependence

Gas permeability typically follows an Arrhenius-type relationship (Equation 1):

P i T = − P i E R T exp ⁡ P i ′ 0 (1)

Where:

• P i ′ 0​ = pre-exponential factor.
• P i E = apparent activation energy for permeation (J·mol^-1).
• R = universal gas constant (8.314 J mol^-1·K⁻¹).
• T = absolute temperature (K).

Increasing temperature typically increases free volume and segmental mobility, thus enhancing diffusivity and permeability.

However:

• Each gas has a different activation energy.
• Selectivity can increase or decrease with temperature, depending on the relative values of P i E and P j E.

For helium:

• Permeability increases with temperature.
• Selectivity over CH₄ and N₂ generally decreases at higher temperatures because the mobility of larger gases increases more rapidly.

Thus, an optimal temperature balance:

• Higher flux (due to increased P i).
• Adequate selectivity to reduce recompression energy.

3.2 Pressure dependence

For dense, nonporous membranes operating in the Henry’s-law regime, the flux of component i scales with its transmembrane partial-pressure driving force (Equation 2):

J i = Q i Δ P i (2)

In helium separations, the operative driving force is therefore the helium partial-pressure difference, not the total pressure drop (Equation 3):where y f H e and y p H e ​are helium mole fractions on the feed and permeate sides, and P f​ and P p​ are the corresponding total pressures. This distinction is critical for helium recovery because helium is typically dilute in natural-gas–derived feeds and off-gas streams; consequently, y f H e​ is small and membrane performance is often driving-force limited. In such cases, treating flux as simply “linear with pressure” can be misleading for process design: raising P f​ or lowering P p​ ​ increases P i​, but it simultaneously increases the driving forces of dominant components such as CH₄ and N₂​. This can increase co-permeation, reduce permeate helium purity, and exacerbate methane losses—so pressure must be evaluated in terms of purity and recovery simultaneously, not helium flux alone.

Δ P i = P f y f H e − P p y p H e (3)

Because helium is weakly sorbing and does not typically induce polymer swelling, deviations from ideal linearity in practical helium service are more commonly driven by mixed-gas effects and mechanical compaction in thin selective layers. At elevated pressures, CO₂​ and other condensable components present in realistic feeds can plasticize susceptible polymer matrices, disproportionately increasing the permeability of larger gases and thereby reducing the effective He/(CH₄​, N₂​) selectivity under mixed-gas conditions. In addition, high transmembrane stress can compact thin-film layers and/or supports, lowering effective free volume and producing sub-linear flux responses even when pure-gas measurements suggest linear behavior. Accordingly, operating pressure selection for helium-containing feeds should be treated as an integrated process decision, coupled with pretreatment (especially for CO₂​ and heavy components), permeate-side vacuum or staged compression, and explicit trade-offs among helium recovery, product purity, methane loss, and specific energy use (Jafari et al., 2024; Torabi et al., 2025b).

3.3 Design implications and performance relevance

Permeability, permeance, and selectivity have duty-dependent relevance in helium separations, and their process impact depends on where the membrane is deployed within the helium recovery chain. Permeability is an intrinsic material property, whereas module sizing and capital intensity are governed more directly by permeance, which sets the attainable throughput per unit area under a given pressure ratio. Selectivity then determines how aggressively staging can be pursued without excessive recycle and recompression an especially decisive constraint for helium, because the effective driving force is dictated by the helium partial pressure difference rather than the total pressure drop. Consequently, the best membrane is not defined by a single transport metric but by the combined effect of permeance–selectivity on stage count, recycle flows, and compression duty at the system level (Baker, 2023; Jabbari et al., 2021; Scholes, 2018).

Published process analyses and design discussions further indicate that membrane feasibility and optimal operating windows vary substantially by feed type. For direct recovery from natural gas, viability is strongly conditioned by feed helium content and the required He/CH₄​ separation factor, with multistage schemes becoming realistic primarily in higher helium fields or at sufficiently high feed pressures. For nitrogen rejection unit (NRU) exit gas, He/N₂​ selectivity requirements for intermediate enrichment can be less stringent, yet the required pressure driving force and stage cut constraints can become prohibitive at low helium concentrations unless higher selectivity and carefully staged configurations are achieved. In contrast, upgrading crude helium streams prior to final purification is repeatedly identified as a comparatively favorable membrane duty, whereas final polishing is often limited by hydrogen co-presence and the intrinsically challenging He/H₂​ discrimination in many membrane classes (Jabbari et al., 2021; Shoji and Moriya, 2003).

At the materials level, helium transport in dense polymers is typically diffusion-dominated due to helium’s low condensability and low solubility; therefore, increases in free volume tend to enhance helium permeability more effectively than affinity-driven solubility tuning. This diffusion dominance explains why useful selectivities are commonly achieved for He over larger gases such as N₂​ and CH₄​, while strong separation from H₂ is substantially more difficult and often motivates hybridization or alternative membrane classes. Recent reports on carbon molecular sieve and fluorinated carbon membranes illustrate how sub-nanometer pore-architecture control can strengthen molecular sieving while maintaining competitive permeance, thereby expanding the attainable He/CH₄​ design space and enabling meaningful enrichment even for dilute feeds in selected cases (Ghasemzadeh et al., 2016; Sunarso et al., 2017; Wu et al., 2024).

These developments reinforce a central design implication for helium service: maximizing permeability alone does not minimize system cost. Insufficient selectivity inflates stage count, recycle ratios, and compression duty, and this process-level penalty can dominate overall performance even when the membrane material exhibits high intrinsic transport capacity. Accordingly, parameter characterization should be framed around the specific duty and mixture, prioritizing mixed-gas data and duty-specific selectivity targets, rather than extrapolating from single gas permeability measurements (Ghasemzadeh et al., 2020; Simplício et al., 2014; Torabi et al., 2025b).

4 Membrane process design for helium recovery

Membrane-based gas separation is now regarded as a key technology for helium recovery from natural gas and related process streams. In practice, membranes are seldom used as standalone purification units; instead, they are integrated into hybrid process trains where membrane stages provide bulk pre-enrichment, and cryogenic or adsorption units achieve final purification to ultra-high purity (Alinejad et al., 2022; Jin et al., 2026; Scholes, 2018). This section explains the role of membranes in integrated flowsheets, compares single- and multi-stage architectures, highlights selected industrial implementations, and summarizes their energy and operational advantages.

4.1 Role of membranes in integrated helium flowsheets

In industrial helium recovery from natural gas, membrane units are typically positioned either:

• Downstream of nitrogen removal units (NRUs), treating nitrogen-rich vent streams, or
• Upstream of cryogenic distillation columns, as pre-enrichment units that reduce refrigeration duty (Hu and Li, 2025; Scholes and Ghosh, 2017).

Figure 8 compares two typical process options for helium production from natural gas. In NRU-tail configurations, vent streams with about 1–3 mol% He are fed into a membrane module that enriches helium to around 40–70 mol%. The enriched permeate is routed to PSA or cryogenic polishing, while the retentate is vented or recycled, depending on the plant design. When placed before cryogenic columns, the membrane selectively permeates helium (and often nitrogen), keeping methane in the retentate. This decreases the volumetric flow and refrigeration load of the downstream cold box, which is especially beneficial in LNG-based plants. In both configurations, membrane units act as flexible “intensification blocks” connecting low-grade helium sources to high-purity product steps. Their modular size, absence of phase change, and compatibility with high feed pressures make them suitable for both new LNG trains and existing facilities retrofitted for helium enrichment (Alders et al., 2017).

4.2 Single-stage vs. multi-stage membrane cascades

Because helium concentrations in natural gas are typically well below 1 mol%, a single membrane stage cannot simultaneously achieve high recovery and high product purity. For realistic selectivities and stage cuts, a one-stage configuration rarely produces a permeate richer than about 5–15 mol% He, which is insufficient for direct bottling or liquefaction and often still too dilute for efficient PSA polishing (Choi et al., 2018; Ghasemzadeh et al., 2020).

Consequently, industrially relevant designs employ two- or three-stage cascades with inter-stage recompression and recycle loops. A typical two-stage arrangement is illustrated in Figure 9. The first stage operates at high feed pressure to maximize helium flux and raise helium to approximately 15%–35%. Part of the retentate is recycled to stabilize the composition and recovery. At the same time, the permeate is compressed and fed to the second stage, which further enriches helium for subsequent PSA or cryogenic treatment.

Three-stage cascades (Figure 10) extend this concept by combining staged pressure reduction and progressive enrichment. Instead of dropping pressure sharply, the feed pressure is stepped down between stages, reducing compressor load, mitigating membrane stress, and recovering part of the expansion energy. In such designs, the first stage lifts helium from ≤1 mol% to roughly 20%–40%, the second to 40%–70%, and the third to 60%–90%, with final polishing provided by PSA or cryogenic distillation (He et al., 2014; Shah et al., 2017; Tilebon et al., 2024).

Techno-economic analyses indicate that well-optimized two- or three-stage membrane networks can reduce specific energy consumption by approximately 30%–70% relative to purely cryogenic helium recovery, while maintaining similar overall recoveries, particularly for low-helium feeds where stand-alone cryogenic designs are uneconomical (Robeson, 2008; Zhou et al., 2024). Table 3 summarizes single-stage, multi-stage, and hybrid membrane configurations, including typical feed ranges, recovery levels, and specific energy requirements.

TABLE 3. Membrane-based and hybrid processes for helium recovery from natural gas.

Energy demand is reported qualitatively due to strong dependence on feed composition, pressure, and degree of integration.

4.3 Industrial implementations and hybrid flowsheets

Several industrial and pilot-scale plants have validated membrane-based pre-enrichment for helium recovery. The Darwin plant in Australia, operated by BOC, is a well-known example of a facility that uses membrane modules to concentrate helium before PSA purification. Air Products’ Noble Gas Purification (NGP) process combines membrane pre-enrichment with adsorption-based polishing to produce helium at 99.999 mol% purity, while maintaining high overall recovery. Linde has patented multiple hybrid flowsheets that combine multi-stage membrane networks with cryogenic and PSA sections, typically using inter-stage compression and recycle to increase helium recovery and reduce methane loss (Khosravi et al., 2025; Torabi et al., 2025b). Air Liquide has also implemented membrane-assisted cryogenic schemes, utilizing membrane pre-enrichment to decrease cold-box duty in helium–nitrogen mixtures (Thonnelier and Candela, 2004).

Regarding materials, Ube Industries (Japan) has developed aromatic polyimide hollow fiber membranes with high He/N₂ and He/CH₄ selectivity, which have been commercialized in several helium recovery units (Dai et al., 2021; Harris and Barnes, 2008). These hollow-fiber modules provide high packing density and durability under high feed pressures, making them suitable for integration in both large LNG trains and smaller gas processing plants. Collectively, these case studies show that membranes are not substitutes for cryogenic or PSA units, but complementary components that improve energy efficiency and recovery in hybrid process trains (Ghasemzadeh et al., 2025; Häussinger et al., 2001).

4.4 Energy efficiency and operational benefits

A key advantage of membrane-based helium recovery is the absence of phase change, which can reduce the refrigeration burden relative to deep cryogenic routes. In practical helium separation from natural gas and NRU vent streams, however, the dominant energy penalty is typically shifted to gas compression or vacuum generation because the effective driving force is the helium partial-pressure gradient, which becomes intrinsically weak at dilute helium levels. This concentration dependence is well documented, and reported specific-energy requirements increase sharply as helium content decreases (Grynia and Griffin, 2016; Scholes et al., 2017); see Table 4.

TABLE 4. Comparative performance and energy metrics of representative helium recovery routes from natural gas.

NR, not reported.

From a design standpoint, higher helium permeance reduces required membrane area and can lower capital intensity, whereas higher selectivity reduces recycle flows and the compression duty associated with recompression of non-helium gases. Process-level studies further indicate that hybridization can translate these material advantages into meaningful reductions in energy demand under favorable operating windows. Representative hybrid membrane–PSA schemes and advanced PSA cycle concepts have been reported to achieve high-purity helium products, while remaining sensitive to feed composition, staging choices, and recovery targets (Werner et al., 2018; Weh et al., 2022a).

Earlier sections compared the main helium recovery routes—cryogenic distillation, adsorption-based systems, and membrane separations and highlighted their inherent trade-offs among energy demand, capital cost, recovery, purity, and environmental impact. Cryogenic units remain the reference when ultra-high purity and near-complete recovery are required, although recent analyses continue to emphasize significant irreversibility’s and the importance of integration and optimization to improve efficiency (Li X et al., 2025).

When integrated into hybrid flowsheets, membranes can reduce refrigeration duty by pre-enriching helium upstream of deep cooling, lower the size and duty of PSA units by decreasing impurity loading, and enable recovery from lower-grade feeds that would otherwise be vented due to the steep concentration dependence of energy consumption. As summarized in Figure 7 and quantified in Table 4, membrane-centered options often provide a strong compromise in modularity and energy use at small-to-medium scale, while cryogenic systems dominate when maximum recovery and ultra-high purity are prioritized, and PSA-based configurations remain attractive as flexible polishing steps or as core technologies for decentralized purification when tailored cycles are used (Jahromi et al., 2018; Jin et al., 2026).

5 Membrane materials for helium separation

The membrane materials' transport properties ultimately limit the efficiency of membrane-based helium recovery. This section examines the primary classes of materials involved in helium separation: polymeric membranes, carbon molecular sieve membranes, porous inorganic membranes, and mixed-matrix membranes. Focus is on permeability, selectivity, stability, and how these factors affect process design (Iulianelli et al., 2023).

5.1 Polymeric membranes

5.1.1 Architectures: dense films vs. thin-film composites

Polymeric membranes used for helium separation are usually manufactured either as:

• Self-supporting dense films with thicknesses on the order of 50–100 μm, used mainly for fundamental property measurements, or
• Thin-film composite (TFC) membranes, where an ultrathin selective layer (often <1 μm) is deposited on a porous support.

Dense films offer reliable intrinsic permeability and selectivity data, but their low permeance does not reflect industrial module performance. TFC architectures, on the other hand, can achieve helium permeances in the tens to hundreds of GPU while maintaining high selectivity, which is crucial for competitive module sizes and energy efficiency. For example, a TFC membrane based on rod-like PBDI reached a helium permeance of approximately 50 GPU with ideal selectivities of about 300 (He/N₂) and 1000 (He/CH₄), demonstrating the potential of well-designed ultrathin layers (Dai et al., 2021).

5.1.2 Polymer families and transport performance

Polymers reported for helium separation do not form a single performance trajectory; instead, the permeability–selectivity combinations summarized in Table 5 (P_He ​ and α values under the stated test conditions) delineate three practically distinct design windows that map directly onto helium-specific process duties. First, highly selective but modestly permeable materials, exemplified by ionomeric and rigid glassy families, are best interpreted as polishing candidates rather than primary recovery membranes. Their selectivity can suppress non-helium co-permeation and reduce recycle compression, yet their limited throughput can translate into large membrane areas when helium is dilute and the separation is partial-pressure limited. This trade-off is particularly relevant for He/N₂​ service, where the size contrast is smaller than for He/CH₄​, so inadequate selectivity more rapidly propagates into additional stages and higher recompression burdens (Alders et al., 2017; Scholes and Ghosh, 2016).

TABLE 5. Representative polymeric membranes for helium separation.

Second, the high-permeability/low-selectivity regime represented by ultra–high free volume per fluorinated amorphous polymers and many PIM-type materials can be attractive for high-flux pre-enrichment, but it is also the regime most prone to increasing methane permeation and diluting permeate helium, especially under the high-pressure mixed-gas feeds typical of helium bearing natural gas streams. Accordingly, these materials are most defensible when the process objective is modest early-stage enrichment and when downstream polishing is explicitly incorporated into the flowsheet. Their apparent advantage should also be interpreted cautiously, because the reported rankings often reflect mild test conditions, whereas industrial operation involves higher pressures, mixed-gas competition, and long operating times where physical aging and related free volume relaxation can reduce permeability and erode the initial transport advantage (Alders et al., 2017; Bushell et al., 2013).

Third, the most strategically interesting space is the balanced window where permeability and selectivity are simultaneously high enough to reduce stage count without making module area prohibitive. Fluorinated dioxolane copolymers and certain copolyimides fall into this category and are therefore more credible candidates for compact multistage designs that target meaningful helium enrichment while controlling methane and nitrogen permeation. However, claims of approaching or exceeding Robeson-type upper bound behavior should be framed as a starting point rather than a conclusion, since helium separations are rarely governed by single-gas transport alone. Mixed-gas effects, contaminant exposure, and the mechanical stability of thin selective layers ultimately determine whether an apparently exceptional intrinsic selectivity translates into sustained module-level performance (Alders et al., 2017; Robeson, 1991).

Overall, the evidence indicates that no single polymer family is universally optimal for helium recovery. Instead, material selection should be duty-specific: high-permeability materials are more defensible for early-stage pre-enrichment of dilute helium, while high-selectivity materials are more defensible for polishing after helium has already been concentrated. This interpretation aligns the permeability–selectivity trade-off with the partial pressure limited nature of helium recovery and avoids over extrapolating laboratory single-condition rankings into field performance (Ghasemzadeh et al., 2016; Torabi et al., 2025c).

5.1.3 Transport trade-offs and operational factors

The permeability constraints that govern most polymeric membranes reflect a persistent selectivity–permeability trade-off: materials exhibiting very high helium permeability often deliver only moderate He/N₂or He/CH₄ selectivity, whereas highly selective polymers typically show lower permeability. Only a limited subset of fluorinated copolymers and ladder-type polyimides approaches current upper-bound performance for He/N₂ and He/CH₄ separations (Wang et al., 2014). In helium recovery from natural gas where helium is commonly lean and methane retention is a non-negotiable constraint this trade-off is not merely a materials observation but a process-level determinant. High-P_He membranes with insufficient α_He/CH4 (or α_He/N2) may increase flux yet still require larger cascades, higher recycle ratios, and greater compression work to meet product targets, while also risking unacceptable methane slip; conversely, higher-selectivity polymers can reduce permeation and cascade complexity, but may impose prohibitive membrane-area requirements when permeance is limited.

Operational conditions further modulate these intrinsic trade-offs in ways that are amplified in multistage helium enrichment. Temperature generally increases permeability, but its net effect on selectivity is system dependent; in staged upgrading, small selectivity shifts can accumulate across stages and alter the optimum operating window as the cascade approaches stringent purity targets. Pressure increases the partial pressure driving force and thus flux, but under NG relevant pressures it can also trigger compaction or plasticization in susceptible polymers effects that are particularly damaging in helium service because even modest selectivity erosion can translate into larger stage cut requirements, higher methane permeation, and increased compressor duty in cascade configurations. Feed impurities accentuate this risk: CO₂, H₂S, and water can plasticize many glassy polymers, increasing the mobility of larger penetrants and reducing effective helium selectivity under mixed gas operation (Kraftschik et al., 2013; Macchione et al., 2007a). Consequently, single gas selectivity values can materially overestimate performance unless mixed gas behavior and contaminant exposure are explicitly accounted for.

Aging imposes an additional time-dependent constraint, especially for high free volume polymers such as PIMs. Physical aging gradually densifies the matrix, reducing permeability and sometimes increasing selectivity, but for process deployment the decisive issue is the stability of both permeance and mixed gas selectivity over operationally relevant timescales, since permeance decay directly increases required membrane area and can shift the cascade economics (Comesaña-Gándara et al., 2019; Wu et al., 2024). Accordingly, mitigation strategies crosslinking, thermal annealing, and the adoption of plasticization resistant fluoropolymers (such as Hyflon AD60X) are most valuable when they demonstrably preserve mixed gas selectivity and permeance under NG like pressures and contaminant levels representative of helium recovery service (Macchione et al., 2007b; Yampolskii et al., 2006).

5.1.4 Process feasibility and outlook

Polymeric membranes are appealing for bulk helium upgrading because they are relatively inexpensive, scalable (using hollow fiber and spiral-wound shapes), and mechanically durable (Baker and Lokhandwala, 2008). Modeling indicates that two or three membrane stages can increase helium concentration from natural gas streams with more than 1 mol% He, as long as α_He/N₂ is at least 5–10; higher selectivities (≳25) significantly decrease membrane size and energy use (Khosravi et al., 2025; Scholes et al., 2017). Typically, polymeric membranes raise helium purity to 50%–70%, with PSA or cryogenic units finishing the purification.

Future polymer development is advancing toward high-T, high-free-volume architectures that combine molecular-sieving properties with long-term stability, as well as toward thin-film composites with defect-free submicron layers. These materials will be key to enabling membrane-dominated helium recovery cascades in hybrid flowsheets.

5.2 Carbon molecular sieve membranes

5.2.1 Molecular sieving and pyrolysis control

Carbon molecular sieve (CMS) membranes are created by pyrolyzing polymer precursors, forming rigid ultra-microporous networks with pore sizes in the Å range needed for true molecular sieving. Unlike dense polymers, which primarily operate via solution diffusion, CMS membranes separate gases by size exclusion: as the carbonization temperature increases, the pore-size distribution shifts toward smaller openings, reducing the transport of larger molecules such as N₂ and CH₄ while enabling high mobility for helium (Jones and Koros, 1994).

Pyrolysis temperature is the key lever. Studies on polyimide-derived CMS indicate that carbonization in the 600 °C–700 °C range yields a favorable balance: He/N₂ and He/CH₄ selectivity increase strongly, while helium permeability decreases only moderately. At higher temperatures, further pore narrowing and graphitic ordering reduce permeability without substantial gains in selectivity (Lei et al., 2021). The carbonization atmosphere (vacuum vs. inert gas) and the heating rate also affect the final microstructure and performance, underscoring the importance of carefully controlled thermal history during CMS fabrication (Dai et al., 2021).

5.2.2 Performance, architectures, and stability

CMS membranes for helium separation have been fabricated as flat films, asymmetric self-supported structures, supported layers on ceramics, thin-film composite (TFC) membranes, and carbon hollow fibers. Reported performance envelopes for asymmetric CMS membranes cluster around a helium permeance of approximately 1000 GPU, with He/N₂ and He/CH₄ selectivities approaching 1000, although most experimentally observed data lie below this idealized limit (Baker and Low, 2014).

The literature includes several notable examples:

• Cellulose-derived flat CMS films carbonized at 600 °C exhibiting He/N₂ ≈ 1993 with PHe ≈174 Barrer (He and Lei, 2021).
• Cellulose-based carbon hollow fibers with PHe ≈403 Barrer and He/CH₄ > 4000.
• PPO-derived CMS layers on ceramic supports with He/CH₄ ≈ 326 and He/N₂ ≈ 173 at helium permeance of 716 GPU (Lee et al., 2006).

Representative CMS performance is summarized in Table 6, and Table 7 lists stability-related metrics.

TABLE 6. Helium separation performance of representative CMS membranes.

TABLE 7. Stability and aging characteristics of CMS membranes derived from different precursors.

Physical aging can be particularly severe in ultrathin CMS layers. For example, a PIM-PI-based CMS TFC with an 82 nm selective layer carbonized at 600 °C showed an approximately 88% decrease in He permeance over 50 days. Mitigation strategies include nanoparticle infiltration (such as Al₂O₃) and protective PDMS coatings, which slow densification but may reduce selectivity. Humidity can also affect transport in hydrophilic CMS like cellulose-based carbons, so upstream drying is recommended for stable helium operation (Ogieglo et al., 2020).

5.2.3 Process integration

CMS hollow fiber membranes provide high packing density and compatibility for large-scale gas treatment modules. A field test of a CMS module with 86 tubes on coal- and biomass-derived syngas showed stable He and N₂ permeances over about 330 h, with field selectivities and stage cuts matching laboratory results (Scholes et al., 2017). Although meant for hydrogen purification, the similarity in how H₂ and He are transported indicates that CMS modules can be effectively repurposed for helium pre-enrichment.

In integrated helium flowsheets, CMS stages are promising for front-end enrichment in membrane–cryogenic or membrane–PSA hybrids, especially in scenarios requiring high He/CH₄ selectivity at moderate permeance. Their ability to exceed polymeric upper bounds makes them strong candidates for demanding separations from methane-rich feeds (Abudaqqa et al., 2025).

5.3 Porous inorganic membranes

Porous inorganic membranes, such as silica, zeolites, MOFs, COFs, and two-dimensional nanosheets, offer rigid or adjustable pore structures with remarkable potential for molecular sieving of helium (Barmaki et al., 2025).

5.3.1 Silica and zeolite membranes

Microporous silica membranes possess ultra-micropores (<0.3 nm) that distinguish between He, N₂, and CH₄ with sub-angstrom precision. Sol–gel-derived silica films have demonstrated He/N₂ selectivities up to 560, with helium permeance around 10⁻⁶–10^–7 mol m^-2 s^-1 Pa⁻¹ (Sunarso et al., 2017). Silica membranes are thermally and chemically stable, withstanding temperatures above 600 °C, but may densify and exhibit decreased permeance over time (Table 8).

TABLE 8. Performance of silica membranes for helium separation.

Zeolite membranes provide crystalline precision in controlling pore size and connectivity. DDR and CHA topologies display He/N₂ and He/CH₄ selectivities ranging from 4 to 80, depending on temperature, demonstrating classical kinetic sieving behavior (Barmaki et al., 2025; Dai et al., 2021). Ultrathin MFI membranes have shown remarkable performance with multiple components; for an equimolar CH4/N2/He mixture at 153 K, separation factors of (CH₄ + N₂)/He = 152, CH₄/He = 265, and N2/He = 38, with high flux, have been reported. Si-RHO frameworks, as suggested by molecular dynamics simulations, theoretically exhibit unlimited He/CH₄ selectivity (Abudaqqa et al., 2025). Quantitative comparisons are summarized in Table 9.

TABLE 9. Performance metrics of zeolite membranes for helium separation.

5.3.2 MOF, COF, and 2D nanosheet membranes

Metal–organic frameworks (MOFs) have highly adaptable pore networks. Classical ZIF-8 membranes have limited He/N₂ selectivity (∼5.1) because flexible linkers expand the openings during operation. However, rigidified versions such as ZIF-8 exhibit He/N₂ selectivity of 30.4, helium permeance above 200 GPU, and long-term stability (>780 h), demonstrating the potential of lattice stabilization (Ghasemzadeh et al., 2024; Hu et al., 2013). ZIF-62 glass membranes offer He/N₂ selectivities of 17–40, permeances of 40–51 GPU, and good CO₂/humidity tolerance (Han et al., 2023). Table 10 summarizes key performance metrics of MOF membranes.

TABLE 10. Performance metrics of MOF membranes for helium separation.

COF membranes have been less studied experimentally but show great potential for computational screening. Many COFs are predicted to surpass Robeson’s upper bounds for He/N₂ and He/CH₄, mainly due to very high permeabilities (10³–10⁴ Barrer) caused by ordered, wide channels (Aydin et al., 2022). Selectivities are generally moderate, making COFs appealing for high-flux initial stages in membrane cascades (Table 11).

TABLE 11. Performance metrics of COF membranes for helium separation.

Two-dimensional nanosheet membranes based on graphene, graphene oxide, and silicene represent the extreme limit of thickness reduction. Vacancy-engineered porous silicene can achieve projected He/Ne and He/Ar selectivities above 10³, with low-energy barriers for helium penetration (0.33–0.78 eV) (Hu et al., 2013). Nano porous graphene oxide with confined ionic liquids (IL-NPGO) has achieved experimental helium permeances of 2.01 × 10⁴ GPU and a He/CH₄ selectivity of 48 (Abudaqqa et al., 2025), far exceeding those of most conventional membranes (Table 12). These systems illustrate the upper-performance frontier for helium separation, although their large-scale fabrication and mechanical integration remain challenging.

TABLE 12. Performance metrics of 2D nanosheet membranes for helium separation.

5.3.3 Comparative outlook

Taken together, porous inorganic membranes span a broad performance envelope:

• Silica offers unmatched He/N₂ selectivity but modest permeance and manufacturing challenges.
• Zeolites provide reproducible crystalline pores and strong kinetic control, with promising multicomponent performance.
• MOFs and COFs bring structural tunability and high permeability, although selectivity and framework stability must be carefully managed.
• 2D nanosheets redefine the upper limit of permeance and projected selectivity, but are still at an early stage regarding scale-up.

In process terms, silica and zeolitic membranes are best suited for high-purity polishing steps. In contrast, MOF, COF, and 2D membranes are more natural candidates for early enrichment stages in advanced cascades.

5.4 Mixed-matrix membranes (MMMs)

Mixed-matrix membranes combine a continuous polymer phase with dispersed porous fillers (such as MOFs, COFs, zeolites, and related nanomaterials) to mitigate the permeability–selectivity trade off of neat polymers by introducing molecular sieving domains within a processable matrix (Li et al., 2017; Muthukumaraswamy Rangaraj et al., 2020). In helium recovery from natural gas, the decisive objective is not higher P_He in isolation, but sustained mixed gas selectivity and permeance under NG relevant pressures and impurity envelopes, because helium is typically dilute and separation is partial pressure limited; consequently, small losses in effective α_He/CH4 ​​ or α_He/N2 ​​ can propagate into higher stage count, increased recycle, methane slip, and greater compression duty in multistage enrichment trains.

5.4.1 Polymer–filler interactions and mechanisms

MMM performance is governed by the polymer filler interface and by how transport pathways behave under mixed gas operation. Interfacial rigidification and selective pore pathways can be beneficial only if they preferentially restrict CH₄​ and N₂​ relative to helium and remain stable under pressure. Conversely, nonselective interfacial voids are disproportionately damaging in helium service, because even minor leakage pathways can collapse the effective selectivity and rapidly undermine cascade enrichment while increasing methane permeation. Pore blockage (polymer intrusion or contaminant induced pore restriction) can likewise negate permeability gains and increase membrane area requirements an especially costly outcome when the helium driving force is weak. Accordingly, helium relevant MMM design should prioritize void-free, mechanically stable interfaces, and demonstrated mixed-gas performance, rather than relying on single gas screening metrics (Molavi et al., 2018; Soleimany et al., 2018).

5.4.2 Representative MMM systems

Among reported MMMs for helium separation, Matrimid based systems remain the most extensively studied. Cu-BDC/Matrimid membranes illustrate the potential of well-matched MMMs: at 30 wt% loading, helium permeability increases from 35.1 to 49.7 Barrer, while He/CH₄​ selectivity rises from 124 to 168 and He/N_2​ selectivity from 162 to 217 (Akbari et al., 2020a; Qu et al., 2025). Cu-BTC/Matrimid MMMs typically exhibit an optimum near 20 wt% loading, whereas higher loadings can promote void formation and reduce separation efficiency (Akbari et al., 2020a) a critical limitation for NG helium recovery, where modest selectivity erosion can translate into disproportionate methane slip and added recompression in multistage operation. More recent 6FDA-TFMB MMMs incorporating hierarchical HNTs@ZIF-67 fillers achieve He/CH₄ selectivity above 300 at elevated P_He​, indicating that hybrid filler architectures can outperform the neat polymer when interfacial defects are controlled (Zhang et al., 2025). While representative datasets (Table 13) suggest that optimized MMMs can approach or exceed Robeson-type upper-bound trends for helium-containing gas pairs, upper-bound proximity should be treated as a screening indicator; for NG duty, the decisive test remains whether the MMM preserves selectivity and permeance under mixed-gas, high-pressure, contaminant-containing conditions (Akbari et al., 2020a; Torabi et al., 2025a).

TABLE 13. Representative MMMs for helium separation and their permeability–selectivity enhancements.

5.4.3 Challenges and future directions

For helium recovery from natural gas, MMM development is constrained by three helium specific deployment requirements: suppressing nonselective interfacial voids that can disproportionately erode α_He/CH4 ​​ and increase methane slip across cascades, maintaining mechanical robustness and manufacturability at high filler loadings under pressure cycling, and demonstrating stability under multicomponent feeds including acid gases and moisture over operationally relevant timescales, since both polymer plasticization/aging and filler hydrothermal tolerance can shift mixed gas transport behavior. Emerging strategies—hierarchical fillers (e.g., HNT@ZIF-67), surface-functionalized MOFs/COFs to strengthen interfacial adhesion, and incorporation of 2D nanosheets to impose controlled transport pathways are most impactful when they measurably preserve mixed gas selectivity and permeance under NG representative conditions (Li et al., 2017; Qu et al., 2025; Rafia et al., 2024).

5.5 Stability challenges: plasticization, aging, and fouling of different materials

Real natural gas feeds pose several stability challenges to membranes over time: plasticization, physical aging, mechanical stresses, and fouling. Polymeric membranes (such as polyimides, per fluoropolymers) are most vulnerable to plasticization by CO₂ and H₂S (Yampolskii et al., 2006; Yavari et al., 2018). Conventional glassy polyimides (e.g., 6FDA- or Matrimid based) can suffer a collapse in He/CH₄ selectivity once a critical CO₂ partial pressure is exceeded, as polymer chains swell and diffusional selectivity is lost (Liao et al., 2026). Approaches such as thermal crosslinking or polymer architecture modification can mitigate this: crosslinked 4,4′-TB polyimide hollow fibers retained high He selectivity even with 50% CO₂ at 40 bar, whereas uncrosslinked fibers would plasticize severely (Kim et al., 2005). Mixed matrix membranes (MMMs) incorporating porous fillers (MOFs, COFs, zeolites, etc.) are also being explored to suppress polymer swelling. Rigid fillers can restrict polymer segment mobility and act as anchors, improving resistance to CO₂ induced plasticization (Li Z et al., 2025). However, the long-term stability of MMMs in aggressive gas feeds is still largely unproven filler/matrix debonding or degradation (e.g., MOFs decomposing in presence of H₂S or water) is a concern that has not been fully addressed in field conditions (Akbari et al., 2020b).

Glassy polymers and MMMs also undergo physical aging, a gradual densification of free volume that decreases permeability over weeks to months (Belov et al., 2011). High free volume materials (e.g., polymers of intrinsic microporosity like PIM-1) show significant aging, which could reduce helium flux in long-term operation (Frentrup et al., 2015). Even standard polyimides experience 10%–20% permeability loss over months as they relax. Promisingly, advanced crosslinked blends have demonstrated minimal aging: a recent thermally crosslinked polyimide blend lost <4% of its helium permeability over 50 days, a remarkable aging resistance compared to typical polymers (Zinoviev et al., 2016). Such improvements, achieved via network formation that prevents microstructural rearrangement, suggest newer materials can combine high initial performance with slower aging. Nonetheless, few studies report multi month stability data for helium membranes. Reporting of multi week performance is needed to identify aging rates; at present, most high-performance lab membranes are tested only for hours to days.

Fouling and condensable contamination pose another practical challenge under real NG conditions. Natural gas may contain C₃ ⁺ hydrocarbons, aromatics (BTEX), compressor oils, or particulates. These can irreversibly foul membranes by condensing or adsorbing onto pore surfaces (Al-Juaied and Koros, 2006a). Light hydrocarbons (C₁–C₄) tend to cause transient performance loss (e.g., reduced permeability and selectivity in polyimide and cellulose acetate membranes), which can often be reversed by heating or purging. In contrast, heavier hydrocarbons (waxes, heavy condensates) are less volatile and can deposit as films that permanently clog membrane pores or surface defects. Baker (2008) noted that such heavy hydrocarbons can foul polymer membranes and shorten their lifespan, despite upstream pretreatment (Baker and Lokhandwala, 2008). Indeed, most industrial membrane systems include filters, chillers, or scrubbers to remove condensable before gas enters the membranes. Still, hydrocarbon contaminants are believed to cause some membrane failures in the field, and no fully effective in-situ cleaning method exists once a membrane is heavily fouled. Different materials handle fouling differently: dense perfluoro polymers (e.g., Teflon AF, Hyflon) are highly hydrophobic and chemically inert, so they tend to resist absorption of polar contaminants and plasticizing gases better than aromatics rich polyimides (Macchione et al., 2007a). This is one reason fluorinated polymers have been studied for NG separations they show excellent CO₂/H₂S resistance (little plasticization) and lower affinity for organic vapors. However, even fluoropolymers will condense non-polar waxes given a sufficient partial pressure; periodic solvent washes have been used to remove heavy hydrocarbon deposits in some systems (Li et al., 2023).

Mechanical stability is also critical. Polymeric hollow fibers have the advantage of intrinsic flexibility and the option of fabric or mesh reinforcement, allowing them to withstand high feed pressures (up to 70 bar in some NG membrane units) with minimal risk of rupture (Scholes et al., 2017; Scholes and Ghosh, 2016). Newer braid reinforced hollow fiber designs can further improve burst strength and mitigate fiber swell or delamination under pressure cycling. In contrast, inorganic membranes like ceramic supported zeolites or carbon molecular sieve (CMS) sheets are brittle; thermal cycles or pressure shocks can create micro cracks that bypass the selective layer (Quader et al., 2020). While thin film composite formats (e.g., a CMS or MOF layer on a porous alumina tube) help, scale-up remains challenging. For example, amorphous silica membranes have shown superb helium sieving performance in the lab, but their fragility and sensitivity to moisture (silanol groups can hydrate and widen pores) limit their practical deployment. Similarly, pure CMS hollow fibers can achieve high selectivities but require careful handling; any physical damage creates non-selective leakage paths. Researchers are exploring ways to improve mechanical robustness, such as blending 2D fillers (graphene oxide, etc.) or using hybrid organic-inorganic matrices, but these solutions are still experimental (Nezhad et al., 2020).

6 Economic assessment of helium recovery technologies

The economics of helium recovery strongly depend on feed composition, process flow diagrams, separation technology, and energy consumption. Although cryogenic separation has traditionally dominated industrial helium production, recent techno-economic analyses show that membrane systems can become cost-competitive, especially for low-grade natural gas streams, when high-permeance, high-selectivity membranes are used. Insights from recent studies (He et al., 2024; Jafari et al., 2024; Quader et al., 2020; Scholes and Ghosh, 2017; Zheng et al., 2023) demonstrate a clear shift toward hybrid membrane-based flowsheets driven by lower capital costs, operational ease, and reduced energy use. This section summarizes economic trends across membranes, cryogenic units, PSA systems, and hybrid setups, followed by comparative metrics and cost benchmarks.

6.1 Cost structure of helium recovery processes

The total cost of helium recovery is shaped by three fundamental components that collectively determine the economic viability of any separation technology. Capital investment accounts for compressors, membrane modules, cryogenic cold boxes, PSA units, and auxiliary heat exchangers. Among these technologies, cryogenic systems exhibit the highest capital intensity because they require deep refrigeration hardware, specialized materials, and multiple low-temperature distillation stages. Operating expenditures are primarily determined by electricity consumption, with compression and refrigeration accounting for the bulk of energy use. Membrane systems generally achieve lower OPEX because they operate without phase change, avoid extremely low temperatures, and benefit from modular configurations with reduced mechanical complexity.

A third cost element arises from helium losses in the separation train. Since helium is a high-value commodity, unrecovered helium represents a significant economic penalty. Technologies with inherently high recovery, such as cryogenic distillation or membrane–PSA polishing configurations, often achieve lower overall costs even when their energy consumption is higher. Consequently, the interplay between recovery efficiency and operational cost forms a critical part of system-level optimization (Ghasemzadeh et al., 2018; Hamedi et al., 2019b).

Across all reviewed studies, feed helium concentration consistently emerges as the strongest determinant of economic feasibility. For membrane-based systems, the financial break-even point is typically reported at feed concentrations of approximately 0.08–0.1 mol%. Below this threshold, a stand-alone membrane cascade becomes economically unfavorable due to the high membrane area requirements and the high recycle compression duties. In such cases, hybrid flowsheets combining membranes with cryogenic or PSA polishing are necessary to achieve competitive recovery and product cost (Yang et al., 2025).

6.2 Economic performance of membrane-based helium recovery

Recent techno-economic studies, including Zheng et al. (2023) and subsequent assessments of circular economy and pilot-scale projects (He et al., 2024; Quader et al., 2020), reveal that the economics of membrane-based helium recovery are mainly driven by three factors: compression energy, membrane replacement costs, and feed helium concentration. As summarized in Table 14, compression duty accounts for 50%–70% of total operating costs, while membrane replacement accounts for 15%–30%, depending on the module lifespan, which is typically 3–5 years. Since the required membrane area increases almost linearly as helium concentration decreases, streams richer than 0.1% He offer the most favorable economics for standalone membrane processes.

TABLE 14. Summary of economic indicators for membrane-based helium recovery.

Break-even analyses show that for feeds containing at least 0.1% helium, membrane systems achieve production costs of roughly 25–40 USD per MSCF, equivalent to 0.9–1.4 USD per Nm³. For leaner feeds in the 0.05%–0.1% helium range, costs increase to 50–70 USD per MSCF unless a hybrid configuration is used. Optimization studies demonstrate that using high-permeance membranes in the first stage and high-selectivity membranes in the second stage can reduce the break-even price by 20%–35%, confirming that cascade design is just as important as material performance.

6.3 Comparative economic evaluation of membrane, cryogenic, and adsorption processes

The economic position of membrane-based helium recovery becomes clearer when compared with established alternatives such as cryogenic separation and PSA systems. As shown in Table 15, each technology has a unique cost profile driven by capital costs, energy usage, and recovery rates. Cryogenic distillation remains the standard for producing high-purity helium and is cost-effective when helium levels are above roughly 0.1%–0.2%, especially in facilities already using refrigeration systems. However, cryogenic methods require significant capital investments in cold boxes, turboexpanders, and heat exchangers, and extensive refrigeration requirements primarily drive their operating expenses. Reported costs for cryogenic recovery typically range from 40 to 70 USD per MSCF, depending on feed composition and plant size.

TABLE 15. Comparative techno-economic performance of major helium recovery technologies.

PSA systems require less capital and operate near ambient temperature, making them appealing for decentralized facilities and for polishing pre-enriched streams. Their economic viability depends heavily on the cost of adsorbent replacement and the achievable recovery rate. For feeds with 0.1% helium, PSA units typically cost between $30 and $60 per MSCF, though performance declines when helium levels fall below this point. Hybrid DR-PSA setups significantly improve recovery but require additional compression steps, increasing operating expenses (Weh et al., 2022b).

Membrane-based processes occupy an intermediate economic niche. They avoid the need for extensive refrigeration while providing modularity and relatively low installation costs. Their competitiveness improves when used for pre-enrichment or as part of a hybrid system. As summarized earlier in Section 6.2, standalone membrane units reach break-even points of 25–40 USD per MSCF for feeds with helium concentrations above 0.1%, making them the lowest-cost option for moderate-grade streams. For leaner feeds, hybrid membrane–cryogenic or membrane–PSA flowsheets become more cost-effective, especially when membranes reduce the volume entering energy-intensive downstream units (He et al., 2024).

Overall, the comparative assessment indicates that membranes provide the most significant economic advantage in three conditions: moderate-grade feeds with 0.05%–0.2% helium, facilities aiming to reduce cryogenic duty through pre-enrichment, and distributed small-to-medium installations where capital minimization is critical. Cryogenic systems remain essential for ultra-high purity and large-scale production, while PSA serves as an efficient polishing step in multi-stage designs.

The spatial distribution of the technologies in Figure 11 clearly shows how cost and specific energy use influence their economic operating ranges. Membrane processes are located in the lower-left area, reflecting their combined advantages of low energy consumption and low capital costs, making them especially suitable for moderate-grade feeds and decentralized setups. DR-PSA systems are found in the mid-range of the plot, indicating their ability to achieve high recovery from lean feeds but at a higher cost than membranes. Cryogenic systems are clustered in the upper-right area, consistent with their significant refrigeration requirements and higher capital investment. Yet, they remain essential for large-scale and ultra-high-purity production. Hybrid configurations lie between these extremes, demonstrating how combining membrane, PSA, or cryogenic stages can balance cost and energy use, especially when the goal is pre-enrichment or reducing cryogenic load.

6.4 Technology readiness levels and impact of feed helium concentration and knowledge gaps

The maturity of helium recovery technologies varies widely, and their practical deployment strongly depends on the helium content of the feed stream. As summarized in Table 16, cryogenic separation remains the only fully commercial option, operating at TRL nine and used extensively in LNG and NRU facilities. It is best suited for feed gases with a helium content of about 0.10%–0.20%, where its high capital cost is justified by superior recovery and product purity. PSA systems, positioned at TRL six to seven, are widely used as polishing units and perform reliably for feeds above roughly 0.10% helium.

TABLE 16. TRL levels and suitable feed-helium ranges of major recovery technologies (He et al., 2024; Hoseinzade and Adams, 2019; Jafari et al., 2024; Quader et al., 2021; Weh et al., 2022b).

For leaner streams, DR-PSA offers a promising alternative at TRL five to six, achieving high recovery in fields containing 0.05%–0.10% helium. Polymeric membranes occupy a similar readiness range (TRL 5–7) and provide the lowest cost solution for feeds above about 0.10% helium, although purity limitations often require pairing with PSA or cryogenic polishing. More advanced membrane materials—CMS, silica, zeolite, MOF, and 2D nanosheet systems—remain at early development stages (TRL 1–4). Despite their attractive intrinsic selectivities, challenges with stability, fabrication, and module-scale-up currently limit their use to hybrid configurations for dilute feeds.

Overall, technology selection is closely linked to helium concentration: as feed richness declines, reliance shifts from stand-alone membranes toward hybrid membrane–PSA or membrane–cryogenic configurations. Conversely, for richer streams, established cryogenic systems remain the preferred route to high-purity helium.

Membrane materials for helium recovery span a spectrum from commercially-ready polymers to conceptual nanomaterials, and it is essential to distinguish their technology readiness levels (TRL) at the material scale (Jafari et al., 2024). Polymeric membranes (e.g., cellulose acetate, polysulfone, polyimides) are at the highest TRL such materials are already deployed in natural gas processing (for CO₂ removal, H₂S removal, or H₂ recovery), and similar modules could be repurposed for bulk helium extraction (Baker, 2023). For instance, cellulose acetate hollow fibers (UOP’s Separex, etc.) offer He/N₂ selectivity 34 and He/CH₄ 31 in ideal tests, and they have been suggested as viable for helium recovery since they are proven in acidic gas service (Flory, 1992; Scholes et al., 2017). Major gas separation firms (Air Liquide, UBE, Generon, etc.) advertise membrane systems for helium recovery from NG, indicating that at least pilot scale polymer membrane skids exist (TRL 7–8) (Quader et al., 2020; Scholes and Ghosh, 2017). However, to date no polymer membrane unit has been fully commercialized specifically for helium production from natural gas, likely due to the economic and performance gap with cryogenic processes. Thus, polymeric membranes are near-deployment the manufacturing and module technology is mature, but field validation for helium is still limited (Ahmadi and Shariati Zarch, 2020; Dai et al., 2021). The next step is demonstrating competitive performance on actual helium-containing natural gas streams. Data under realistic conditions (multi-component feeds, high pressure) for polymers are partial: there are reports of short-term high-pressure tests (e.g., crosslinked polyimide fibers under 40 bar CO₂/CH₄/He) and simulations of multi-stage polymer membrane cascades for NG helium recovery, but very few long-term field trials in the public domain (Wang et al., 2025). This is a knowledge gap that needs to be closed by pilot studies, especially to assess how feed contaminants and aging affect polymer membrane performance over months.

For per fluorinated polymers and other specialty polymers (PIMs, thermally rearranged polymers, etc.), the TRL is slightly lower. These materials have shown superior lab performance (e.g., Hyflon AD perfluoro polymer achieves He permeability 476 Barrer with He/CH₄ selectivity 157) and intrinsic resistance to plasticization, but translating them into industrial hollow fibers or composite modules is not fully proven (Macchione et al., 2007b; Scholes and Ghosh, 2017). To date, such polymers have mostly been fabricated as small-scale flat films or coated on supports. There is some scale up evidence, for example, process simulations include Hyflon AD in multi-stage membrane schemes for helium recovery, implying researchers can obtain these membranes in prototype module form. We may assign an intermediate TRL (4–5) here: concept validated in lab, with limited pilot testing. The knowledge gap lies in module fabrication and lifetime can these high free volume polymers be made into defect free hollow fibers, and will they maintain performance (or age) similarly to conventional polymers? Additionally, while per fluoropolymers are chemically durable, we lack data on their performance in mixed feeds containing heavy hydrocarbons (which could induce swelling or plastically deform even a Teflon-like matrix over time).

Carbon molecular sieve (CMS) membranes and inorganic microporous membranes (zeolites, metal–organic frameworks, etc.) are at a mid TRL as well. Small-scale CMS modules (e.g., bundles of CMS hollow fibers) have been tested for gas separations, but mostly for O₂/N₂ or H₂/CH₄ rather than helium. These materials excel in selectivity: CMS membranes can exceed polymer upper-bound selectivities for many gas pairs due to sub-Å pore discrimination. For helium, CMS and ultra-microporous zeolites have achieved impressive ideal selectivities (e.g., a SAPO-34 zeolite membrane showed He/CH₄ 14 with He permeance 2.3 × 10⁻⁷ mol m^-2·s^-1·Pa-¹, and a dense MOF membrane in a recent study reached He/CH₄ selectivity 1190 with 4% He in CH₄) (Cui et al., 2025; Dai et al., 2021; Denning et al., 2019; Qu et al., 2025). Those numbers underscore the conceptual potential (in fact, a MIL-116(Ga) MOF–polymer composite achieved 37 Barrer He permeability and 1190 selectivity in a mixed feed, an extraordinary result breaking prior limits) (Komal et al., 2025). Yet, these inorganic based membranes are generally only at lab scale (TRL 3–4). The cited MOF composite, for example, was a few-inch² flat membrane tested for hours (Smith et al., 2014). Field data are lacking–no public reports exist of a CMS or zeolite membrane unit running on a natural gas well stream containing helium. The gaps include: tolerance to impurities (e.g., water or H₂S could poison certain MOFs or cause framework collapse), mechanical scale-up (how to produce large-area defect-free inorganic membranes, given their brittleness), and long-term stability. Notably, one study found a CMS hollow fiber could maintain performance over >300 h in a humid CO₂/Xe separation (Hui et al., 2026), indicating some robustness, but analogous long-duration data for He/CH₄ or He/N₂ are scarce. Overall, CMS/zeolite membranes are mostly in the concept to bench stage promising in principle, with a handful of short pilot demonstrations needed to advance them (Farnam et al., 2021).

At the far end, graphene-based membranes (including single-layer nano porous graphene and stacked graphene oxide laminates) remain largely conceptual (TRL 2–3). Atomically thin graphene with angstrom size pores could, in theory, provide exceptional helium flux and selectivity. Simulations predict that appropriately tuned graphene pores or functionalized graphene sheets can distinguish He from larger gases with selectivity in the hundreds to thousands (Yoon et al., 2016). In practice, however, creating a large-area monolayer with well controlled sub-nm pores is extremely challenging. Experimental graphene membranes for gas separation have been limited to cm-scale samples in controlled settings for instance, one study demonstrated monolayer graphene was effectively impermeable to helium until nanoscale defects were introduced. To date, no data exist for graphene membranes under realistic NG mixtures (with CO₂, H₂S, etc.), and even the effect of heavy hydrocarbons or long-term operation on such membranes is unknown (Luque-Alled et al., 2023; Smith et al., 2014). Graphene oxide (GO) multilayer membranes, which are easier to make, tend to have larger gallery spacings (0.4 nm) that allow He and even larger molecules to permeate, but controlling selectivity is an issue. They are also sensitive to humidity (GO layers can swell with water). Thus, graphene-based membranes are an exciting concept with potential for ultimate performance, but they are still in the early research phase. Key knowledge gaps include scalable fabrication and demonstration that these membranes can operate in harsh gas environments without degradation.

While membrane technology holds promise for more energy-efficient helium recovery, material-specific readiness varies widely. Table 17 provides a summary of major membrane classes, highlighting their representative performance and the extent of testing under real-world conditions.

TABLE 17. Performance and readiness of key membrane classes for helium recovery (Aydin et al., 2022; Dai et al., 2021; Denning et al., 2019; Mei et al., 2024; Nezhad et al., 2020; Scholes et al., 2017).

NR, not reported.

7 Conclusions and future directions

This review examined membrane-based helium recovery from natural gas, with emphasis on separation fundamentals, membrane materials, process integration, and techno-economic performance. Cryogenic separation remains the industrial benchmark for ultra-high purity and high recovery, particularly in LNG and NRU facilities that can benefit from existing refrigeration systems. However, high capital intensity and energy demand mean that cryogenic-only designs are typically most economical for moderate-to high helium feeds. PSA and advanced cycle concepts such as DR-PSA offer flexible non-cryogenic options and are highly effective as polishing steps, but their competitiveness generally decreases for very lean feeds (often below 0.1 mol% He) unless configurations and integration are carefully optimized.

Membrane technologies play a complementary role by enabling bulk pre-enrichment ahead of high-purity polishing units. Polymeric membranes, including advanced fluoropolymers, high free-volume glassy polymers, and thin-film composite architectures, already provide practical permeability–selectivity combinations for pre-enrichment. Carbon molecular sieve membranes, silica and zeolite membranes, MOF/COF structures, and emerging two-dimensional nanosheet concepts can approach molecular-sieving behavior in controlled tests and move performance closer to projected upper-bound trends. Mixed-matrix membranes further indicate that well-designed polymer–filler interfaces can improve permeability and selectivity simultaneously. Nonetheless, physical aging, plasticization, defect control, and long-term mechanical stability) especially under mixed-gas, high-pressure operation with CO₂, H₂S, water, and condensable hydrocarbons (remain major barriers to broader industrial deployment and to translating laboratory metrics into durable module performance.

From a process perspective, membranes are most effective as pre-enrichment units in hybrid flowsheets. Two and three stage cascades, together with recycle optimization and integration with downstream units, can reduce overall energy use and production costs for low-helium feeds. Techno-economic studies suggest production costs of roughly $25–40 per thousand standard cubic feet for feeds above 0.1 mol% He, while hybrid membrane–cryogenic or membrane–PSA schemes can extend economic viability toward 0.05 mol% depending on site-specific assumptions and available infrastructure. In practice, the optimal choice among cryogenic, PSA, DR-PSA, and hybrid membrane configurations depends on feed composition, target purity, plant scale, and the availability of compression and refrigeration utilities.

Future priorities include plasticization-resistant polymers and MMMs with controlled interfaces, CMS membranes with reduced aging, and scalable inorganic or 2D membranes that retain intrinsic selectivity at the module scale. More pilot-scale evidence is needed to confirm long-term stability under realistic natural-gas conditions and to establish credible TRL progression beyond short-duration demonstrations. Finally, combining membrane-assisted helium recovery with digital optimization, advanced control, and practical development pathways for marginal fields could make helium supply chains more robust. As demand grows across healthcare, semiconductor manufacturing, and scientific research, membrane-based and hybrid processes are likely to play a larger role in enabling efficient and cost-effective helium production.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. Generative AI tools (ChatGPT, OpenAI) were used solely for language editing and clarity improvement. The AI tools were not used to generate original scientific content, data, analyses, interpretations, or conclusions. All scientific content and responsibility for the manuscript remain with the authors.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Glossary

AF Amorphous Fluoropolymer AI Artificial Intelligence CAPEX Capital Expenditure CHFM Carbon Hollow Fiber Membrane CMS Carbon Molecular Sieve COF Covalent Organic Framework CTA Cellulose Triacetate DFT Density Functional Theory DR-PSA Dual Reflux Pressure Swing Adsorption GPU Gas Permeation Unit LNG Liquefied Natural Gas MD Molecular Dynamics MMM Mixed-Matrix Membrane MOF Metal–Organic Framework MRI Magnetic Resonance Imaging MSCF Thousand Standard Cubic Feet NG Natural Gas NMR Nuclear Magnetic Resonance NRU Nitrogen Rejection Unit Nm ^**3** Normal Cubic Meter OPEX Operating Expenditure PBO Polybenzoxazole PEI Polyetherimide PFMD Fluorinated Dioxolane Copolymer (Poly(PFMD)) PIM Polymer of Intrinsic Microporosity PPL Poly (phthalazinone) PSA Pressure Swing Adsorption SEM Scanning Electron Microscopy TFC Thin-Film Composite TRL Technology Readiness Level TSA Temperature Swing Adsorption VSA Vacuum Swing Adsorption Symbol Meaning Unit J i Permeation flux of component i mol·m^-2·s^-1 j i Molar flux (Fick form) mol·m^-2·s^-1 D i Diffusivity of component i m²·s^-1 S i Solubility coefficient mol·m^-3·Pa⁻¹ P i Permeability Barrer Q i Permeance mol·m^-2·s^-1·Pa⁻¹ or GPU p i / − p i Transmembrane partial-pressure difference Pa l Membrane selective-layer thickness m Δ P i Pressure difference across the membrane Pa P i E Activation energy for permeation J·mol^-1 P i , 0 Permeability pre-exponential factor — R Universal gas constant 8.314 J mol-1·K−¹ T Temperature K or °C αᵢ/_j Ideal selectivity (permeability ratio) — m ˙ Mass flow kg·h^-1 (from cryogenic case studies) E Specific energy use MJ·mol^-1 He CAPEX Capital cost USD OPEX Operating cost USD·yr^-1 P f e e d Feed pressure bar or MPa T f e e d Feed temperature °C PHe Helium permeability Barrer PHe (GPU form) Helium permeance GPU wt% Filler loading in MMMs % C o s t H e Helium production cost USD·MSCF^-1 Barrer Permeability unit GPU Gas Permeation Unit (10^–6 cm³(STP)/cm²·s·cmHg) MJ·mol ^**-1** He Specific energy consumption MSCF Thousand Standard Cubic Feet Nm ^**3** Normal cubic meter mol·m ^**-2** ·s ^**-1** ·Pa ^**−1** Permeance Å Ångström µm Micrometer

Footnotes

Edited by: Kamran Ghasemzadeh, University of Edinburgh, United Kingdom

Reviewed by: Adele Brunetti, National Research Council (CNR), Italy

Rashid Jamshidi, Manchester Metropolitan University, United Kingdom