The challenge the team set out to address lies in the chemistry of light naphtha itself. This petroleum fraction typically contains hydrocarbons with five to seven carbon atoms, spanning linear alkanes, branched alkanes, cyclic alkanes, and aromatics. Although these molecules are chemically similar, their industrial fates diverge sharply. Linear and mono-branched alkanes are prized as feedstocks for ethylene production, the gateway to countless plastics and chemicals; steam crackers that convert naphtha into ethylene prefer streams rich in these straighter molecules because they yield the desired light olefins efficiently. Aromatics serve as essential chemical intermediates for a wide range of products, from polymers to solvents to specialty chemicals. Highly branched and cyclic alkanes, by contrast, are valuable constituents of high-octane gasoline, since their compact, branched structures resist the uncontrolled ignition that causes engine knock. Because the molecules share nearly identical boiling points and physical properties, separating them at the molecular level has remained a formidable obstacle for refiners seeking to route each component to its most valuable use. In practice, much of this material is burned as blended gasoline or processed in bulk, with the potential value locked inside the individual molecular classes left unrealized.

Membranes have long been proposed as a lower-energy alternative to distillation. A membrane acts as a molecular gatekeeper, allowing selected molecules to pass through nanoscale channels while blocking others, all without the repeated vaporization and condensation that makes distillation so energy-intensive. The energy savings arise because the separation is driven mechanically or by partial pressure differences rather than by heating the entire mixture through a phase change. Membrane technology has already displaced distillation in select industrial niches, most famously in the recovery of small molecules from natural gas streams, but extending it to complex liquid hydrocarbon mixtures has proved far harder. The Dalian and Sun Yat-sen team recognized a fundamental limitation: a single membrane is seldom capable of efficiently fractionating a complex petroleum mixture at the molecular level. Light naphtha presents too many molecular species with too many overlapping properties for any one material to sort them all in a single pass. A membrane tuned finely enough to discriminate between closely related alkanes will typically sacrifice throughput or miss entirely different classes of molecules. The researchers' response was not to search for one perfect membrane, but to engineer several imperfect ones and arrange them in sequence, each assigned a distinct molecular sorting task.

The platform for this graded strategy was a metal–organic framework, or MOF, a class of porous crystalline materials whose ordered nanopores can be tuned to regulate molecular transport with remarkable precision. MOFs are built from metal nodes connected by organic linkers, producing crystals with internal surface areas and pore geometries that can be designed almost atom by atom, which has made them attractive candidates for gas storage, catalysis, and separation science. Specifically, the team used CuBTC, a copper-based MOF whose pore architecture provides the physical scaffold for separation. CuBTC is one of the most extensively studied MOFs, prized for its relative robustness and reproducible synthesis, which makes it a practical starting point for membrane fabrication. The key innovation, however, came in how the researchers modified this scaffold. They treated the CuBTC membranes with tannic acid, a naturally derived polyphenolic compound found abundantly in plant tissues, under mild conditions. This treatment progressively altered both the effective pore size and the chemical functionality of the membranes, and by adjusting the treatment conditions, the researchers could steer the membranes toward different separation behaviors. In other words, a single parent material could be nudged along different developmental paths, emerging with distinct molecular preferences depending on how the modification was carried out.

Out of this tuning process emerged two membrane types with complementary capabilities. One membrane type primarily separated hydrocarbons according to their molecular dimensions, preferentially permitting the passage of linear and mono-branched molecules while holding back bulkier species. The second membrane underwent a more pronounced change in chemical functionality: its modified surface chemistry enhanced interactions with aromatic molecules, allowing aromatics to be pulled apart from structurally similar aliphatic hydrocarbons that would defeat a purely size-based sieve. This distinction is important because size and shape alone cannot resolve every ambiguity in naphtha; an aromatic ring and a branched alkane of similar carbon count can present nearly identical molecular dimensions, so the membranes must discriminate by chemical affinity as well as by geometry. In effect, the tannic acid treatment converted a single membrane material into a family of molecular sorting devices, each specialized for a different discrimination task.

The researchers then integrated these two membranes into a cascade separation process that mirrors the logic of a refinery flowsheet, but at the molecular scale. Refineries have always worked this way at the macroscale, passing crude fractions through successive columns and conversion units, each stage performing one narrow job. In the first stage of the membrane cascade, the size-selective membrane extracted linear and mono-branched hydrocarbons from the complex mixture, producing a stream suitable as an ethylene feedstock. In the second stage, the aromatics-selective membrane preferentially separated aromatic hydrocarbons, yielding a stream enriched in chemical intermediates. What remained after these two extractions was a third stream dominated by multi-branched and cyclic hydrocarbons, precisely the components that constitute high-octane gasoline blending stock. Nothing in the mixture was left without a destination; every molecule class was routed toward the product category where it carries the greatest value.

To demonstrate the approach, the team applied the membrane cascade to a simulated light naphtha mixture containing 15 different hydrocarbons representative of the real petroleum fraction. The cascade successfully resolved this complex mixture into the three value-specific product streams, achieving recovery rates of approximately 85 to 90 percent for each stream. That recovery figure matters as much as the selectivity itself: a separation that captures only a small fraction of each target class would waste valuable material, whereas recoveries near 90 percent suggest that the graded membranes can harvest the bulk of each molecular family while still maintaining the discrimination needed to keep the streams distinct. In separation science, recovery and purity often trade against one another, so achieving both simultaneously across three streams is the notable feat.

The energy implications of the membrane cascade were evaluated through chemical process simulations. For a representative five-component mixture, the researchers estimated that the membrane process would reduce energy consumption by approximately 91 percent compared with conventional distillation while achieving the same target recovery. A reduction of that magnitude, if it can be realized at industrial scale, would represent a substantial step toward more sustainable refining, since separation operations account for a large share of the energy consumed in petroleum processing. The comparison was made against distillation achieving identical recovery targets, meaning the energy savings were not obtained by sacrificing product yield; the membrane route delivers the same molecular harvest for a fraction of the thermal input.

The study, titled "Molecular-Level Petroleum Refining by Graded Membranes," was led by first authors Yuecheng Wang of the Dalian Institute of Chemical Physics and Fangdi Dong of Sun Yat-sen University, with corresponding authors Weishen Yang and Yujie Ban of the Dalian Institute of Chemical Physics and Dongdong Zhou of Sun Yat-sen University. The work was published as a peer-reviewed contribution in National Science Review, a journal of Science China Press, and is indexed under DOI 10.1093/nsr/nwag488.

As with any laboratory-scale advance, several questions must be answered before graded membranes can reshape refinery operations. The separation was demonstrated on a simulated light naphtha mixture rather than raw refinery streams, which contain a broader and less predictable cast of contaminants, sulfur compounds, and trace species that can foul or degrade membrane materials over time. The energy savings figure derives from process simulations for a five-component mixture, and translating those estimates to full-scale units handling real naphtha will require detailed engineering analysis, including the pressure drivers, membrane areas, and module designs needed to process the enormous volumes that refineries handle daily. The durability of tannic acid-modified MOF membranes under prolonged industrial operating conditions, and the cost of fabricating them at scale, remain open engineering challenges that the published study does not resolve. Fabricating defect-free MOF membranes over large areas, in particular, is an acknowledged difficulty in the membrane field, since even tiny pinhole defects can erode selectivity.

Nevertheless, the conceptual contribution of the work extends beyond light naphtha. The graded membrane strategy assigns distinct molecular sorting tasks to membranes engineered with tailored pore structures and chemical functionalities, and then chains those tasks together so that a complex mixture is decomposed stepwise into its most valuable components. That concept, the researchers note, provides a foundation for the development of more energy-efficient separation processes applicable not only to complex petroleum fractions but also to other multicomponent chemical mixtures, a category that includes many of the separations on which the chemical industry's energy budget is spent.

The broader context makes the result timely. Refining is among the most energy-hungry corners of the chemical enterprise, and distillation, for all its reliability, separates molecules through brute-force phase changes repeated dozens of times in a single tower. Replacing even a portion of that thermal load with membrane-based molecular sorting could meaningfully lower the carbon footprint of fuel and chemical production. By showing that a single MOF material, reshaped through a mild treatment with a plant-derived polyphenol, can yield membranes with genuinely different sorting personalities, and that those membranes can work in concert to fractionate a 15-component hydrocarbon mixture into three useful streams, the Dalian and Sun Yat-sen team has offered a concrete demonstration that molecular-level refining is technically attainable.