Introduction

From the depths of the sea to the vastness of space, there are many large-scale and high-tech engineering masterpieces in the feat of human conquest and transformation of nature, including expansion of cultivated land to meet the growing demand for food. Increasingly, people realize that the rapid industrialization and urbanization over recent decades have led to severe environmental challenges, particularly the deterioration of air quality caused by the rising levels of particulate matters (PMs)1,2,3. Public tackles this issue with fibrous filters, which are widely used in air purification systems to capture airborne particles. However, most commercial filters are made from petroleum-derived synthetic materials, such as polyester (PET) and polypropylene (PP), or ceramic material such as glass fibers. Unfortunately, these materials are not environmentally friendly and cannot degrade naturally. The traditional treatment approaches, such as direct landfill and incineration, cause serious environmental pollution and resource waste, while the reasonable recycling of these commercial filters is fraught with technical difficulties and high costs4. In contrast, the development of fully biodegradable filters derived from renewable biomass presents a promising solution to these challenges5,6. Filters made from naturally abundant and degradable polymers can be safely decomposed in soil environments, returning to the natural carbon cycle without leaving persistent residues. Moreover, the filtering materials, which primarily rely on physical and size-based mechanisms, are only effective for capturing large-sized PMs, and usually exhibit limited filtration capability for the most penetrating particle size (PM0.3, ≈300 nm) due to insufficient interactions with pollutants. These limitations are closely associated with structural defects, including non-uniform pore size, limited porosity, and poor fiber connectivity. Therefore, developing sustainable biomass-based fibrous filters with high efficacy performance would significantly enhance their competitiveness against petroleum-based products.

Constructing a heterogeneous and continuous fiber network has been recognized as an effective strategy to improve material utilization and performance7,8,9,10. For example, our previous work fabricated a spider-web-inspired membrane by introducing cellulose nanocrystals, which promoted jet splitting during electrospinning and resulted in the formation of ultrafine fibers with characteristic diameters below 100 nm11. While these advances significantly improved the fiber network structure, the fabrication process still relied on toxic organic solvents, which are difficult to remove completely during electrospinning. Residual toxic solvents may lead to serious complications in air filtration, particularly in applications involving human breath filtration, where solvents residues could enter the body through physiological fluids exchange, which may differ from the original purposes of ecological sustainability12. In this context, environmentally friendly biomass resources, which can be used as the solute in spinning precursor systems and can be soluble in green solvents, are emerging as viable options.

Agricultural waste has already been widely utilized across various important industries to achieve a carbon-neutral sustainable society. Corn straw, as a common agricultural waste, can be converted into a valuable resource. Moreover, zein protein is also abundant, mainly derived from corn gluten meal, which is a protein feed produced as a by-product during the deep processing of corn starch13,14. Originating from these agricultural residues of corn processing, the resulting filters with continuous networks are therefore expected to serve as promising green air filters with high efficiency. Although extensive research has been conducted on PM filtration, few studies have successfully fabricated air filters that combined green and sustainable material components with structural designs capable of supporting high efficacy.

Here, we establish a solute-solvent system solely containing two kinds of agricultural residues (zein protein and straw cellulose) via electrospinning. Specifically tailoring to this system by fully utilizing the inherent ability of zein protein and cellulose, an incomplete nonsolvent-induced phase separation is proposed to self-assemble a dual-network (D-net) structure. This specific structure enables the effective interception of PM0.3 particles, which can embed within the isotropic grooves of the microscale fibers and be adequately captured through interactions within the nanoscale fiber network, thereby achieving full-structure filtration. As a result, the developed filter exhibits superior filtration capacity, comparable to or exceeding that of traditional commercial filters. Moreover, the biodegradable fabric neither relies on petroleum-based polymers and toxic solvents, nor requires additives such as metal nanoparticles or salts, thereby promoting greenness and sustainability during material preparation, and perpetuating a closed-loop cycle after use. Our work suggests that, in this post-petroleum era, the synthesis of such fully biodegradable filters may lead to a sweeping change, that is, the petroleum-based chemical filter can be substituted gradually by a full-biomass filter deriving from the agricultural residues.

Results

Design and preparation of corn-based dual-network filter

Utilizing sustainable materials derived from corn plant residues, we fabricate a fully biodegradable air filter via a green electrospinning strategy (Fig. 1a). All components originate from nature and ultimately return to nature through degradation, forming a closed-loop cycle that minimizes environmental impact. This corn-based filter is fully biodegradable and does not require recycling, thereby promoting safe disposal and offering a cost-free solution for post-use treatment, fully aligning with the philosophy of environmentally sustainability. Our design concept, from nature to nature, reflects ecological harmony and sustainability, which is the principal feature that differentiates it from other air filter media.

Fig. 1: Sustainable fabrication strategy, fiber formation mechanism, and dual-network (D-net) architecture of corn-based electrospun fibers.
figure 1

a Schematic diagram showing a design philosophy conducive to the green, circular, and low-carbon development. b Incomplete nonsolvent-induced phase separation driven by dynamic solvent exchange. c Schematic of solubility-controlled inward contraction based on a tubular fiber model. d Jet splitting driven by viscoelastic mismatch between CNFs and zein. e Schematic diagram of the D-net structure. f SEM image of the corn-based D-net filter consisted of microscale fibers and nanoscale fibers. g Comparison of filtration efficiency and pressure drop for PM0.3 between this work and previously reported filters fabricated by various technologies22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38.

To elaborate the structural evolution of self-assembled architecture, we focus on a unique protein type, zein protein, as a proof-of-concept model. As shown in Supplementary Fig. 1, zein is abundant with polar functional groups (e.g., −OH, −C = O, and −NH2). Zein cannot dissolve in pure water or pure ethanol but can dissolve and denature in 60%–90% ethanol solution (Supplementary Fig. 2a and Supplementary Discussion I). It would have meant combining those two solvents is important for dissolving zein, and removal or partial removal of either component from this solvent system can cause zein to precipitate. This characteristic was verified by a neat experiment, after evaporation in the atmosphere for about 2 h, a solid protein film is formed on the solution surface (Supplementary Fig. 3). Moreover, ambient relative humidity (RH) significantly influences fiber morphology during fiber formation, as it not only causes the liquid-liquid phase separation into polymer-rich and polymer-poor regions, but also affects the solidification rate of polymer from either single-phase or polymer-rich regions15,16. Hence, from the macroscopic point of view during electrospinning, since the rate of ethanol volatilization is faster than water, it is quite reasonable to suppose that the solidification of zein fibers will occur if we tailor the RH to let the concentration of ethanol solution out of this range. To this end, we chose ethanol and water (80 wt.%: 20 wt.%) as the solvent, which not only endowed the material with a rough surface structure through phase separation, but also improved the greenness of the electrospinning process.

Based on Flory–Huggins theory calculations of the ternary zein-ethanol-water system (Supplementary Fig. 2b), we propose an incomplete nonsolvent-induced phase separation (NIPS) process driven by dynamic solvent exchange. As illustrated in Fig. 1b, the initial spinning solvent consists of 80 wt.% ethanol and 20 wt.% water. As the jet travels through a humid atmosphere, a dynamic solvent exchange occurs, involving solvent (primarily ethanol) evaporation from the jet and simultaneous water vapor absorption from the environment, which gradually shifts the solvent composition across the jet. In conventional NIPS processes, water acts as an external nonsolvent and immediately triggers phase separation upon contact with the jet, resulting in the formation of electrospun fibers with porous or structured surfaces17,18,19. However, in our system, since water is already present as a component of the initial solvent, the water content does not immediately exceed the critical threshold (~40 wt.%) required for zein precipitation (Supplementary Fig. 2c). Instead, a radial solvent gradient gradually develops from the jet surface to the core layer, leading to delayed and asymmetric solidification, which ultimately leads to progressive fiber deformation during jet elongation and facilitates the formation of distinct surface morphologies.

To reveal how incomplete NIPS process governs the morphological evolution of zein fibers, we designed a tubular structural model to mimic the deformation of individual fiber during electrospinning (Fig. 1c). At the early stage of fiber formation under an ambient RH of 90%, the jet surface first experienced an incomplete NIPS process and solidified to form a skin layer, while the bulk remained solvent-rich and exhibited a lower degree of solidification (Fig. 1c-i). As the jet stretches, the radial solvent gradient continues under the combined effects of external solvent evaporation, internal solvent diffusion, and water vapor inward penetration, gradually giving rise to an external-to-internal pressure difference (ΔP) (Fig. 1c-ii, Supplementary Discussion II). This ΔP induces radial compressive stress (Fr), driving inward contraction of the bulk. As solidification progresses, the increasing Fr eventually exceeds the mechanical stability of the fiber surface, leading to the formation of longitudinal grooves (Fig. 1c-iii). The mechanical stresses acting on the tubular fiber are further illustrated in Supplementary 4. To illustrate the influence of ambient RH on this process, we also performed electrospinning under low humidity (≈30%) as a comparison. As shown in Supplementary Fig. 5 and Supplementary Discussion III, fibers electrospun under high ambient humidity (≈90%) exhibit a grooved rod-like morphology with an average diameter of 2.06 ± 0.38 µm, whereas those prepared under low ambient humidity (≈30%) exhibit a smooth ribbon-like morphology with an average diameter of 3.18 ± 0.27 µm.

A major challenge is that the fiber created from the protein always has a large diameter, which may increase the pore size, decrease the porosity, and lower the probability of particles capture. Nanocellulose, when used as a dispersed phase in precursor, has been shown to promote the formation of fine fibers through jet splitting during electrospinning11,20, and thus provide a sustainable alternative to partially replace traditional additives such as cationic surfactants. Inspired by this, TEMPO-oxidized cellulose nanofibers (CNFs) derived from corn straw were introduced into the zein-based spinning precursor to regulate fiber morphology under RH 90%. The jet splitting behavior in our system may result from the viscoelastic mismatch between zein matrix and dispersed CNFs, as illustrated in Fig. 1d. During jet elongation, the flexible zein, as a continuous phase, are easy to stretch, while the rigid CNFs, as an insoluble dispersed phase, are more difficult to stretch. This viscoelastic mismatch in jet mobility can lead to uneven stretching, and when this stretching heterogeneity becomes pronounced, jet splitting may happen21.

Generally, by combining the fiber splitting facilitated by the addition of CNFs with the incomplete NIPS resulting from the limited solubility of zein, a dual-network (D-net) structure composed of interwoven nanoscale smooth fibers and microscale grooved fibers was fabricated via one-step electrospinning (Fig. 1e, f). This corn-based filter with a D-net structure exhibits a high filtration performance for PM0.3, as evidenced by its superior filtration efficiency and low pressure drop compared to previously reported filters fabricated through conventional technologies (Fig. 1g and Supplementary Table 1), all of which are non-biodegradable and non-renewable materials22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38. Importantly, our corn-based filter is entirely derived from agricultural residues, and neither relies on petroleum-based polymers or toxic solvents, nor requires additives such as organic salts to construct the D-net structure. This strategy not only enhances greenness and sustainability during material preparation, but also promotes a closed-loop life cycle after use, fully aligning with the principles of eco-friendly material.

Structural performances of corn-based D-net filter

We are of the opinion that the high filtration efficacy of the corn-based filter may result from a D-net structure caused by the addition of CNFs into the spinning precursor, which significantly altered fiber morphology, as shown in Fig. 2a, b. This D-net structure consists of nanofibers and microfibers with average diameters of 0.29 ± 0.18 µm and 2.61 ± 1.11 µm, respectively (Fig. 2c). In addition to the emergence of fine fibers, the diameter of coarse fibers increased compared to that of zein fibers, likely because CNFs can block the continuity of protein matrix, leading to slightly enlarged coarse fibers39. Dynamic light scattering measurements revealed that zein/CNFs solutions exhibited larger average nanoparticle sizes compared to zein solutions, suggesting the formation of larger assemblies that may influence jet dynamics during electrospinning (Supplementary Fig. 6). Moreover, the introduction of CNFs further deepened the surface grooves on the surface of coarse fibers, which may enhance the physical entrapment of PMs. This can be attributed to the rigidity of CNFs within the zein matrix, which disrupts the uniform flow of the spinning jet and introduces local viscoelastic variations6. Such rheological heterogeneity may give rise to localized stress concentration during jet elongation. In the subsequent solidification process, these stresses can promote intensified surface contraction, thereby forming deeper grooves and wrinkles along the fiber surface. As revealed by the confocal laser scanning microscopy (CLSM) results, FITC-labeled zein appeared as scattered green spots, while CW-labeled CNFs exhibited narrow strip-like distributions along the fiber (Fig. 2d). This morphology likely arises from the limited deformability and high modulus of CNFs, which restrict their response to matrix deformation, particularly during surface shrinkage. The rheological difference was verified by time-dependent viscosity measurements, which showed a sharp increase in viscosity within seconds after CNFs addition to the zein precursor (Fig. 2e). This result further supports the significant impact of CNFs on jet behavior under the shear stress generated by the electric field during electrospinning. The Fourier transform infrared (FTIR) spectrum of D-net fibers shows no significant difference or new absorption peak compared to pure zein fibers, suggesting that the interactions between zein and CNFs are primarily physical in nature (Fig. 2f). These interactions may include hydrogen bonding between hydroxyl or amide groups in zein and hydroxyl groups in CNFs, as well as electrostatic attraction between negatively charged carboxylate groups on CNFs and positively charged amino groups in zein. Additionally, both zein and CNFs contain hydroxyl groups, which provide protons, while the carbonyl group in zein may act as a proton acceptor. Therefore, even a small amount of CNFs can strengthen the fiber network by creating additional interactions within the zein matrix39.

Fig. 2: Structural performances of corn-based filter.
figure 2

SEM images of (a) zein fibers and (b) zein/CNFs fibers under RH 90%. c Fiber diameter distribution of the D-net fibers. d CLSM images of a D-net fiber with fluorescently labeled zein (FITC-zein) and fluorescently labeled CNFs (CW-CNFs). e Viscosity of zein solution and zein/CNFs changes with evaporation time. f FTIR spectra of CNFs, D-net fibers, zein fibers, and pristine zein. g Porosity and h pore size distribution of corn-based filter with ribbon, rod, and D-net structures. i SEM images of fiber cross section of ribbon structure (RH 30%) and rod structure (RH 90%). Values are shown as mean ± SD, n = 5. Source data are provided as a Source Data file.

The synergistic effect of high RH and CNFs introduction makes the fibers have larger specific surface area, thereby establishing a D-net structure, and thus enhancing the adsorption and retention capacity for PMs. This structure tackles key technical bottlenecks in manufacturing the net structure with large gaps, critical empty space, and discontinuous distribution. The pore structure analysis, including pore size distribution and porosity, further validates the advantage of the D-net structure in membranes (Fig. 2g, h). The increase of RH (from 30% to 90%) promotes the transition of the filter structure from dense to fluffy, which leads to an increase in the porosity of the filter (from 56.62% to 73.08%). This structural change is accompanied with a higher occurrence of concave regions and the formation of fine nanofibers (~74 nm), further increasing porosity (from 73.08% to 94.62%). This increase in porosity contributes to a reduction in the dominant pore size range, which is primarily 1–4 μm, with a significant proportion below 1 μm, as confirmed by quantitative SEM image analysis (Supplementary Fig. 7). Moreover, the miscible ethanol and water solution system can promote the water in the environment into the jet, accelerating phase separation and thereby enhancing the fiber porosity. Therefore, the corn-based filter exhibits small pore sizes while maintaining high porosity. It should be noted that while water vapor entering the jet can promote the pore formation in the core layer of the fiber, it will make the fiber surface less vulnerable to the pore structure, and even inhibits the formation of surface porous structure. As shown in Fig. 2i, both smooth ribbon-like and grooved rod-like fibers lack surface pores, while the rod fiber has core pores. For the skin layer, the water vapor in the air will continuously reduce the solubility of the solvent to the zein, which accelerates the zein precipitates from 80% ethanol solution, and make the jet surface solidify into a uniform layer of soft shell in advance. For the core layer, the number of escaping ethanol molecules is greater than that of escaping water molecules due to the volatilization rate, which causes the vapor pressure of the solvent to be greater than the water vapor pressure in the air. In this case, the fiber-air interface is always full of saturated solvent vapor, making the water vapor unable to spread to the jet, and thus hindering the phase separation inside the jet. From Fig. 2i (right), the groove rod fiber has pores in core layer, the main reason is that under the RH 90%, the water vapor pressure in the air increases at the fiber-air interface, which makes the water vapor diffuse into the jet, and the water content in the solvent increases, thus accelerating the phase separation and forming the core pores inside the bulk. Furthermore, in contrast to the ribbon fibers prepared at relatively low-humidity and the rod fibers prepared without the addition of CNFs, the formation of D-net fibers led to a fluffy fibrous accumulation, which is in correspondence with the common wisdom that a fluffy structure tends to benefit the reduction of the pressure drop (Supplementary Fig. 8).

Air Filtration performance of corn-based D-net filter

PMs pollution, especially for the most penetrating particles (PM0.3), has become a significant burden on public health and global economies40,41. Here, in considering the integrated properties of fluffy structure, high porosity and small pore size, applications in air filtration by the corn-based D-net filter are possible. The D-net filter can serve as a degradable and disposable sanitary material with the performances of low airflow resistance and high filtration efficiency, which can protect the human body and indoor air quality through natural passive ventilation. As presented in Fig. 3a, the D-net filter with a superlight weight of 8.0 g m−2 shows a filtration efficiency of 99.4168% at pressure drop of 42 Pa. This can be attributed to fluffy D-net structure, which can simultaneously improve the PM capture capacity and reduce airflow resistance. A quantitative study was carried out concerning how the base weight affected the airflow resistance of the D-net filter (Fig. 3b). Benefited from the complex changes in hierarchical porous structure, the pressure drops of the filter increased nonlinearly, which might be related to the increasing numbers of layers of fibers. At an increased base weight of 10.2 g m−2, the corn-based D-net filter showed 99.9994% for PM0.3 removal, which qualified for the standard for ultralow penetration air filters of >99.999%. Meanwhile, the pressure drop under the airflow velocity of 5.33 cm s−1 was 45 Pa, which was negligible compared to the atmospheric pressure (only 0.04% of the atmospheric pressure). Considering that different testing conditions are proposed for air filters for various purposes, especially the airflow velocity, we evaluated its filtration performance under 5.33, 10, 14.1, and 16.6 cm s−1, which represent typical airflow velocity standards for industry equipment, ventilation and air conditioning, personal respirator, and extreme cases, respectively. As the airflow velocity increased from 5.33 to 16.6 cm s−1, the corn-based D-net filter maintained nearly unchanged filtration efficiency for PM0.3 and a slightly increased pressure drop (Fig. 3c). For instance, under a high velocity of 16.6 cm s−1 (standard for personal respirator), our filter exhibited a steady efficiency for PM0.3 (98.5257%) and a slight increase from 53 Pa to 109 Pa for pressure drop, which could be attributed to the sieving behavior of the D-net structures. Correspondingly, we specifically selected commercial filtration materials to compare their filtration efficacy with our corn-based D-net filters (Fig. 3d and Supplementary Table 2). To evaluate the overall performance of the D-net filter, the quality factor (QF), which is related to the ratio between the particulate removal efficiency (η) of the air filter and the pressure drop (ΔP) due to airflow across the filter, was calculated. The corn-based D-net filters, by contrast, presented good filtration capacity for PM0.3 due to their progressive fiber structure. Moreover, compared with commercial filters that use traditional filtering fibers, the corn-based D-net filters surpass the others in terms of comprehensive factors, including process simplicity, high filtration efficacy, and environmental friendliness, particularly their biodegradability after service.

Fig. 3: Functionality and mechanism of air filtration and formaldehyde adsorption.
figure 3

a PM0.3 filtration efficiency and pressure drop of the corn-based filter with different architectures. Airflow velocity, 5.33 cm s−1. b PM0.3 filtration efficiency and pressure drop of the D-net filter with various base weights. Airflow velocity, 5.33 cm s−1. c PM0.3 filtration efficiency and pressure drop of the D-net filter at various airflow velocity. Base weight, 11.6 g m−2. d Comparison of filtration efficiency and pressure drop between commercial filters and the D-net filter. e HCHO adsorption efficiency of the corn-based D-net filter compared with a commercial HEPA. f Time-dependent behavior of the relative weight gains from the HCHO of the corn-based D-net filter. g SEM images illustrating the filtering behaviors consisting of full-structure filtering and sieving, as well as the capture mechanisms, including embedded policy for PM0.3 by the corn-based D-net filter. Values are shown as mean ± SD, n = 5. Source data are provided as a Source Data file.

In addition to its robust removal efficacy for PM0.3, this corn-based filter demonstrated an excellent adsorption capacity for formaldehyde (HCHO). HCHO is a molecule with a size much smaller than the PM0.3, and we suggest that the HCHO removal may be governed by an interaction-based filtration mechanism provided by this D-net structure, as discussed below. The time-dependent HCHO adsorption efficiency of corn-based filters with different structures was measured (Fig. 3e). For comparison, the adsorption efficiency of the commercial HEPA filter for HCHO was also tested. The HCHO adsorption efficiency of the D-net filter dropped from 81.36% to 71.25% steadily after 120 min of testing time and further to 52.66% after 240 min, while the HCHO removal efficiency of the commercial HEPA filter was below 6.45% and decreased to less than 3.6% after 240 min, likely due to the lack of active sites on the surface of the materials. The weight of captured pollutants was also measured after testing (Fig. 3f). After 240 min of testing, the total weight of captured pollutants on the corn-based D-net filter increased from 2.4 to 8.3 mg, while that of the HEPA filter showed a modest increase from 0.8 to 2.8 mg. The ratio of the weight of captured pollutants (Wp) to the initial weight of the filter (Wf) was used to describe the ability to absorb pollutants. For the corn-based D-net filter, the Wp/Wf increased from 0.36 to 1.26 with an increase of the testing time from 40 to 240 min, while that of the HEPA filter increased slightly from 0.005 to 0.017. These results highlight the pollutant loading capacity of the corn-based D-net filter, which not only achieves efficient HCHO removal but also accumulates a pollutant mass exceeding its own weight, demonstrating a unique combination of fluffy structure and high adsorption efficiency.

The obtained D-net structure and rich functional groups could help the establishment of an all-structure filtration mechanism, with a range from integrated filters to dual networks to single fibers, including microfibers and nanofibers. Generally, the filtration efficiency for particles primarily relies upon fiber morphology due to four primary size-based filtration mechanisms. To present the filtration mechanism, we performed an additional experiment (Supplementary Fig. 9). In this setup, organic particulate pollutants were generated by burning sandalwood. As shown in Fig. 3g, many particulates were observed adhering to the fiber surface, which cannot fully be explained by the classical size-exclusion-based mechanism. This observation suggests that these groups in polluted air can strongly interact with the functional groups on the surface of the nanofibers. At the integrated filter level, sieving, as a main filtration mechanism, can partly block the penetration of PMs with sizes larger than the pore sizes of the filters. From a view of D-net structure, the existence of D-net can further prevent the PMs penetration via mechanisms of inertial impaction, interception, and diffusion. An embedded capture mechanism also exists, that is, the PMs can be trapped within the groove-like surfaces of coarse fibers. At the individual nanofiber level, due to the functional groups they carry, the fibers can capture more particles, particularly those with functional surfaces through functional groups interaction, which contributes to enhanced filtration efficacy, especially for smaller and charged/functionalized particles. In addition to the size-based physical mechanism, the corn-based fibers can interact with airborne pollutants (PMs and HCHO) through three interaction-based mechanisms, including charge-charge, polar-polar, and hydrogen-bonding interactions42. Supplementary Fig. 10 illustrates the potential interactions among zein, CNFs, PM particles, and HCHO, representing a reinforcement mechanism driven by strong fiber-pollutant interactions. Zein, as previously discussed, contains many active functional groups that can interact with toxic chemicals and even solid PMs in polluted air (Supplementary Discussion I). The aldehyde groups in HCHO may interact not only with the carboxylate groups of CNFs but also with carboxylic and amine groups of zein. In addition, the zein/CNFs precursor forms a two-phase system, with zein acting as the continuous matrix and CNFs as the dispersed phase, resulting in interfacial polymerization. CNFs have high crystallinity, which allows their molecular chains to be orderly arranged, resulting in larger dipole moments and higher dipole polarization43. Moreover, the crystalline and amorphous regions of cellulose usually undergo different degrees of interfacial polarization at the interface owing to the difference in electron cloud density, which is an inherent dielectric property of the cellulose44,45. The incorporation of CNFs into the zein matrix introduces additional dipole orientation polarization and Maxwell-Wagner interface polarization under the electric field, which results in a higher relative permittivity46. Therefore, the extra load of cellulose introduces interfacial regions within zein matrix, and the huge difference between the two components further intensifies the interfacial polarization, thereby increasing the adhesion of PMs on the surfaces of micro/nanofibers. Hence, in addition to the small pore size of the D-nets filters, this adhesion effect also contributes to the stable removal efficacy.

Considering the real environmental conditions, humidity was also evaluated. The water vapor transmittance rate (WVTR) of the corn-based D-net filter was 789.64 g·m2·day−1. The filtering performance of the corn-based filter under high humidity conditions was investigated by placing the membrane in a constant temperature and humidity chamber with a fixed relative humidity of 90%. The treatment durations were 2, 4, 6, 8, 10, and 12 h, respectively. Notably, the dual-network structure can still be retained in the filter after 10 h humidification treatment. As shown in Supplementary Fig. 11, even after 12 h of exposure to 90% RH, the filtration efficiency of the corn-based filter is roughly constant, while the pressure drop increases due to a reduction in porosity. Therefore, the corn-based filter can still guarantee effectiveness under high humidity levels according to the Department of Energy Technical Standard. The D-net structure with fluffiness, high porosity, and small pore size endows the corn-based filter with high PM0.3 removal efficacy, which in return requires a robust mechanical strength. The ribbon-like, rod-like, D-net filters exhibit the tensile strength of 1.59, 1.66, and 2.13 MPa, respectively (Supplementary Fig. 12). This enhancement of mechanical properties of corn-based filters is attributed to the addition of CNFs, which helps to form a D-net structure to enhance its mechanical properties. Although the tensile strength of the filter is not strong, it is sufficient for the filter to maintain the shape against external force or airflow under practical conditions. As shown in Supplementary Fig. 13, this free-standing corn-based D-net fabric (0.009 g) could support a plastic frame weighing 15 g with one finger, and could be bent or stretched without any damage, suggesting good robustness and flexibility. Moreover, the corn-based filter can withstand an airflow velocity of 16.6 cm s−1, confirming its mechanical properties to function without any support of substrates, thereby making it possible to be used as filters.

Environmental viability of corn-based filters

The key challenge in air filtration is maintaining high performance while ensuring biodegradability and environmental sustainability. Developing sustainable materials that minimize energy consumption in air pollution control is essential for long-term environmental benefits. Given that air filters are typically discarded after use, assessing the full environmental impact of these filters is important for advancing more sustainable alternatives. Therefore, a cradle-to-grave life cycle assessment was conducted to assess the environmental consequences of the corn-based filters compared to PET fabric, PP fabric, and glass fiber fabric (Supplementary Tables 35 and Supplementary Method I). The system boundary includes raw material acquisition, production, and waste disposal (Fig. 4a). Across all examined environmental impact categories, using agricultural residues rather than petroleum-based polymer considerably lowers environmental impacts on nearly all categories, especially in terms of fossil resource scarcity, mineral resource scarcity, marine eutrophication, terrestrial ecotoxicity, and human non-carcinogenic toxicity (Fig. 4b, Supplementary Fig. 14 and Supplementary Tables 69). These results highlight the reduced dependence on fossil resources and mineral resources during production, minimizing potential toxicity to terrestrial and marine ecosystems, and lower human health risks during both production and end-of-life stages. The biodegradation performance of corn-based filter was tested in different environments. The sample with a size of 15.5 cm × 7.5 cm was buried in soil to assess its biodegradability, and the morphological changes were recorded every three days (Fig. 4c). Since the components of the biomass-based filter, zein and cellulose, are biodegradable, this D-net filter could gradually degrade in soil after disposal, leaving no environmental burden. After 2 weeks, the corn-based filter was entirely decomposed. The degradation speed in the phosphate buffer solution (PBS) was slower than that in soil (Fig. 4d). These results demonstrate that the corn-based filter, derived from agricultural residues, meets the requirements of environmental protection and land degradation. Its advantages are further embodied in their fully biodegradable property and do not need to recycle, which promote the safe disposal and provide a cost-free solution for recycling, so thus in accordance with the environmentally sustainable philosophy.

Fig. 4: Environmental viability of corn-based filters.
figure 4

a System boundary of life cycle assessment. b Environmental impacts of corn-based filters in contrast to the polyester fabric, polypropylene fabric, and glass fiber fabric, respectively. Digital photographs of corn-based filters with a size of 15.5 cm × 7.5 cm degraded in different environments, with (c) showing soil and (d) showing PBS.

Discussion

In summary, we have demonstrated a synthesis methodology for the preparation of high-efficiency sustainable air filters via green electrospinning technique. The resulting fibrous structure features continuously welded dual networks consisted of alternating microfibers (2.61 ± 1.11 µm) with grooved surface formed by incomplete nonsolvent-induced phase separation, and nanofibers (0.29 ± 0.18 µm) resulting from the jet splitting behavior driven by viscoelastic mismatch between zein and CNFs. Based on this structural design, an all-structure filtration mechanism with a range from integrated filters to dual networks to single fibers was proposed. This corn-based D-net filter with a superlight base weight (10.2 g m−2) exhibits high filtration performance for PM0.3, achieving a high removal efficiency (>99.99%) with low pressure drop (45 Pa). Importantly, this fully biodegradable filter eliminates the need for post-use recycling and offers a safe, low-cost end-of-life solution aligned with sustainable development principles. We envision that such intriguing dual-network filters, which can operate as a stand-alone filter or in combination with respirators, window screenings, and filter canisters, will herald vast potential for biomass-based high-performance filters toward personal protection, engine intakes, ventilation systems, and medical devices.

Methods

Materials and chemicals

The mature corn plants were harvested and collected in October 2022 at a corn field in Acheng District, Harbin, Heilongjiang Province, China (N45°32′29″, E126°58′30′′). The corn stalks were cut at the bottom, air-dried, and shredded into small pieces to obtain the corn straw. Zein protein, sodium chlorite (NaClO2, 80%), 2,2,6,6-tetramethylethylpiperidine-1-oxyl (TEMPO), acetic acid, ethanol, and potassium hydroxide (KOH) were purchased from Aladdin Chemistry Co. Ltd. (Shanghai, China). Deionized water was used throughout the experiments.

Fabrication of corn-based D-net filter

Cellulose was first extracted from corn straw and then subjected to TEMPO-mediated oxidation to obtain TEMPO-oxidized cellulose nanofibers. For simplicity, we refer to the TEMPO-oxidized cellulose nanofibrils as CNFs. The detailed fabrication process is provided in Supplementary Methods II. The preparation and the corresponding characterization of CNFs are illustrated in Supplementary Discussion IV and Supplementary Fig. 15. The CNFs suspensions were ultrasonicated in 80 wt.% ethanol solution at 200 W for 15 min in an ice-water bath. The dispersion state of CNFs was evaluated (Supplementary Fig. 16). Zeta potential of zein and CNFs in 80 wt.% ethanol was measured, respectively (Supplementary Fig. 17). The spinning solutions were prepared by dissolving 25 wt. % zein in 80% ethanol solution, with CNFs added to achieve final concentrations of 0 and 0.075 wt.% relative to the total solution weight, respectively. The spinning precursor was subjected to additional magnetic stirring at 400 rpm for 30 min to achieve a homogeneous mixture (Supplementary Fig. 18). Spinning solution properties, including viscosity, electrical conductivity, and surface tension were shown in Supplementary Table 10. The electrospinning process was performed by a conventional electrospinning device (Yong Kang Le Ye Co., Beijing, China), details are presented in Supplementary Methods III. The base weight of the corn-based filter was controlled by carefully adjusting the spinning time.

Characterization

The morphological structure and crystallinity of CNFs were characterized using transmission electron microscopy (TEM, Hitachi-7650, Japan) and X-ray diffraction (D/max-2200VPC, Japan). The viscosity of the zein solution and zein/CNFs solutions was determined by a hybrid rheometer (HAAKE MARS, Thermo Scientific, USA). The viscosity of the spinning precursor was measured using a digital rotational viscometer (SNB-1, Heng Ping Co., Shanghai, China) at 25 °C under 35% RH. During the test, the sample was placed in a beaker and tested with a No. 2 rotor at a rotation speed of 12 rpm. The electrical conductivity and surface tension were measured using a conductivity meter (DDSJ-318, Lei Ci Co., Shanghai, China) and a surface tension meter (JK99B, Zhong Chen Co., Shanghai, China) at 25 °C at 30% RH, respectively. The zeta potentials of zein and CNFs dispersions, as well as the particle sizes of zein and zein/CNFs solutions, were measured at 25 °C using a Zetasizer (Zetasizer Nano, Malvern Instruments, UK). The morphologies and structures of the micro- and nano-structured architectures were characterized using scanning electron microscope (SEM, Apreo S HiVac, Thermo Scientific, USA, precoated with gold for 120 s) and TEM. The particulate sizes of CNFs and diameter distribution of fibers were measured and were determined by the Nanomeasure software. The thickness of the corn-based D-net filter was measured by a thickness gauge (AIPLI0-10mm, readability 1 μm. The distribution of zein and CNFs in electrospun zein/CNFs fibers was examined with an Olympus FV1200 confocal laser scanning microscope. FITC and CW have excitation and emission spectrum peak wavelengths of approximately 488 nm/500–550 nm, and 405 nm/410–480 nm, respectively. FTIR spectra was conducted using a NICOLET 6700 spectrophotometer (Thermo Fisher Scientific Inc., USA) in the range of 4000–600 cm−1. The pore structure of the corn-based D-net filter was characterized by CFP-1100AI capillary flow porometer (Porous Materials Inc., USA). The porosity (ε) of the corn-based D-net filter was analyzed by a gravimetric method and calculated using Eq. (1).

$${{{\rm{\varepsilon }}}}\left(\%\right)=\frac{\left({M}_{{wet}}-{M}_{{dry}}\right)/{\rho }_{w}}{\left({M}_{{wet}}-{M}_{{dry}}\right)/{\rho }_{w}+{M}_{{dry}}/{\rho }_{M}}$$
(1)

Where Mwet and Mdry were the wet and dry weight of the corn-based filter; ρw and ρM were the densities of silicone oil and the corn-based filter, respectively. The WVTR of filters was measured using the inverted cup water method with a fabric moisture permeability meter (C360M, Labthink, China) according to the standard of ASTM E96, at a constant humidity (80%). PM0.3 filtration efficiency and pressure drop of the corn-based filter were measured by a filter equipment (TSI 8130A, USA). The process was detailed in Supplementary Method IV. The concentrations of HCHO were diluted in a glass bottle to the level that was measured by a particle counter (PPM-400 ST) with chemical sensors for HCHO. The experimental procedure is detailed in Supplementary Method V. The tensile property of the corn-based filter was tested by a microelectronics universal testing machine (SSANS, Shanghai Enterprise Development Co., Ltd).