A silk fibroin based green nano-filter for air filtration

Fibrous air filters fabricated by electrospinning have proved to be an effective approach among the various strategies for PM2.5 removal. However, in the electrospinning process, the large amounts of toxic organic solvents usually evaporate into the atmosphere and disposing of these used polymer-based air filters would leave further pollution in the environment. Here, we report on the fabrication of a silk fibroin based nanofiber air filter with robust filtration performance via a green electrospinning process. Silk worm cocoons were degummed and dialyzed against water to form the silk fibroin solution and then the silk fibroin nanofiber membranes were fabricated by electrospinning with the help of polyethylene oxide. Moreover, special attention was paid to the morphological evolution of the pollutants captured by the nanofiber nets during the filtration process. It was discovered that the inherent properties of silk fibroin play a key role in improving the filtration performance. Benefiting from the richness of functional groups, the resultant silk fibroin fibrous membranes exhibited a high filtration efficiency of 99.99% with a relatively low air resistance of only 75 Pa, leading to an obvious higher quality factor. Due to the biodegradability of silk fibroin, the membranes are disposable after use. We believe that the methodology and results presented here will not only provide a novel perspective for air filtration, but also pave the way for producing a safe and clean air filtration system.


Introduction
Air pollution is growing worse in certain parts of the world, especially in Asia, Africa and Latin America, hitting the developing countries hardest and resulting in a wide range of potentially life-shortening health problems. Now, the devastating air pollution poses a great threat to human beings, climates and ecosystems and this trend is expected to continue in the future. [1][2][3][4][5][6][7] A photo taken of the same place during clear and hazy days indicates that visibility decreases greatly due to air pollution (Fig. 1a).
Airborne particulate matter (PM), a major type of anthropogenic air pollutant, is a complex mixture of inorganic matters (including sulfur dioxide, oxides of carbon, oxides of nitrogen) and organic matters (including volatile organic compounds and elemental carbon). 1 Based on its aerodynamic diameter, PM can be classied into ne particulate having an aerodynamic diameter < 2.5 mm (PM2.5) and coarse particulate having an aerodynamic diameter < 10 mm (PM10). They are both composed of a variety of components, but differ vastly in physical properties, health risk and behavior in the atmosphere. PM2.5 is of particular concern as some components of it can penetrate deeply into lung alveoli and enter into bloodstream. 8 Moreover, PM2.5 has the longest atmosphere lifetime (days to weeks), and thus can accumulate and spread by air ow to extended area. Extensive studies have reported that cardiovascular problems can be induced within hours of inhalation of PM2.5, and chronic exposure leads to a reduction of several years in life expectancy. PM is believed to be the most deleterious air pollution in affecting human health and it has been proved to be responsible for 3-7 million deaths worldwide per year. 9-13 PM is mainly generated in the process of urbanization and industrialization, including incomplete combustion of fossil fuel, biomass burning, and vehicle emissions. 14,15 In current situation, many methods have been developed and applied to relieve people from air pollution. Individuals can take a commercial breathing mask to prevent the inhalation of PM2.5 when taking outdoor activities and in indoor spaces, air quality is ensured by employing the electrostatic air cleaners or modern ventilation systems. However, the function part of conventional air lter is mostly made of several layers of randomly oriented melt blown micron-scale bers, which is inherently incapable of capturing ne particles due to the ltration mechanism. 16 Fortunately, if the diameter of the ltration media decreases to some degree, the formed sinuous channels will become narrow enough to trap PM2.5 particles, let alone the PM10 particles. Therefore, beneting from the relatively small diameter, nanobers are attracting great attention in the eld of air ltration. 16 Electrospinning, as a promising technology to fabricate nanometer-scale bers in large scale efficiently, has stand out compared with various routes such as template synthesis, 17 phase-separation, 18 hydro-thermal reaction, 19 etc. Based on the electrospinning technology, a variety of polymeric nanobers have been prepared for air ltration, including polyurethane, 20 polyacrylonitrile, 21 polysulfone, 22 poly(lactic acid), 23 polycarbonate, 24 poly(vinyl alcohol), 25 polyamide, 26 etc. However, to make these synthetic polymers electrospinnable, large amount of organic solvents is used to dissolve them. The decent concentration of an electrospinning solution is usually less than 30%, indicating that most of the toxic, harmful organic solvents will evaporate into the atmosphere. Besides, once the air lter reaches the point where they can no longer be used, it will be disposed as non-recycled plastic waste, and then burned in waste-to-energy power plants or ends up in landlls where it may take up to 1000 years to decompose. Considering the source of PM pollution and the two aforementioned points, the demand for advanced ltration media research that looks toward the use of green technologies and alternative to synthetic materials increases.
Silk is a semi-crystalline biopolymer produced by silkworms. Silk broin (SF), a natural protein extracted from the Bombyx mori silk worm cocoon, 27 has been used for clothing in ancient China and been researched in various elds in recent years due to its several distinctive properties including mechanically robust strength, 28 good biocompatibility, lack of toxicity 29 and aqueous processibility. 30 Up to now, only a few investigation reports the use of silk in the eld of air ltration. Zhang et al. 31 demonstrated a human-friendly silk nanober air lter fabricated by electrospinning from SF/formic acid precursor solution. T. Scheibel et al. 32 produced silk-based ne dust lters for air ltration by using hexauoroiso-2-propanol (HFIP) as solvent, which is more toxic and dangerous. Clearly, the above studies have several drawbacks: (i) the inescapable use of harmful organic solvents that determined by the electrospinning mechanism, (ii) not elucidating the mechanism resulting in the difference between synthetic polymers and SF in ltration performance.
Here, we present the fabrication of SF-based brous ltration media with robust ltration performance via green electrospinning process (Fig. 1b), which is a signicant step toward safe and clean fabrication. To avoid the use of toxic and harmful organic solvents, the SF-based nanober membranes were fabricated by electrospinning from a non-toxic precursor solution (mixture of water, silk broin, and PEO). More signicantly, the effect of ber surface structure and property on the time evolution of lter clogging were investigated. Lastly, the biodegradability of SF-based membranes was studied. It has been found that SF-based nanober membranes possess a high ltration efficiency of more than 99.99% and a low pressure drop of 75 Pa. As an approach that is cost-effective, environment-friendly and scalable, the SF-based brous ltration media not only provides a novel perspective for air ltration, but also paves the way for producing a safe and clean air ltration system. The main text of the article should appear here with headings as appropriate.

Materials
Silk cocoons were purchased from Huzhou, Zhejiang province, China. Polyethylene oxide (PEO), polyacrylonitrile (PAN) and lithium bromide (LiBr) were purchased from Sigma-Aldrich. N,N-dimethylformamide (DMF) were obtained from Shanghai Chemical Reagents Co., Ltd, China. All chemicals were of analytical grade and used without further purication.

Preparation of silk broin solution
Silk broin was extracted from Bombyx mori silk worm cocoons according to previously published methods. 30 Cocoons were removed of the insect and immersed into boiling Na 2 CO 3 solution (0.02 M) for 60 min. The degummed bers were rinsed with distilled water to remove residual Na 2 CO 3 solution and air dried overnight. The dried bers were solubilized in LiBr (9.3 M) at 60 C for 4 h. The obtained solution was dialyzed against distilled water with a regenerated cellulose tube (3500 g mol À1 molecular weight cut off). The water was changed every 6 hours for 12 times. The solubilized silk broin protein solution was dialyzed against pure water for 36 h to remove LiBr, and then centrifuged to remove insoluble particulates and stored at 4 C. The nal silk broin concentration was about 8% (w/v) aer purication.

Preparation of electrospinning precursor solutions
The 8 wt% silk broin electrospinning precursor solution was prepared by adding PEO (1 000 000 g mol À1 ) directly into the silk aqueous solutions to generate SF/PEO solutions with the weight ratio of 1 : 4. PAN was dissolved in DMF to a concentration of 10 wt% for electrospinning.

Preparation of brous membranes
Different kinds of electrospinning precursor solutions were loaded in a 5 ml syringe, and a positive DC voltage (10-20 kV) was applied at the stainless steel needle tip and collector (connecting earth). The syringe pump pushed the solution through the needle tip slowly. The applied potential, feeding rate, electrospinning duration and needle-collector distance were carefully adjusted according to different kinds of precursor solutions. Lastly, SF/PEO, PAN brous membranes were collected on the aluminum foil collectors aer 10-50 min of electrospinning. For SF/PEO nonwoven fabrics, they were immersed into an ethanol solution for 10 min at room temperature and then incubated in water for 36 h at 37 C to remove PEO.

Characterization
The morphological features of SF/PEO, PAN, and commercial brous membranes were characterized by Hitachi S-4800 coldcathode eld-emission scanning electron microscope (FE-SEM). Fourier transform infrared spectroscopy (FTIR) (Nicolet 5700) was conducted in the transmittance mode in the spectral range of 800-4000 cm À1 with a resolution of 0.1 cm À1 . Water contact angle test was carried out using HARKE-SPCAX3 contact angle measuring instrument.

Filtration test
The samples were detached from the aluminium foils and then attached on the hole of the supporter to block the diffusion of smoke. The PM2.5 was generated by burning of cigarette and collected in the reactor under the presence of moisture. A particle counter (Lighthouse 3016) was used to measure the PM particle number concentration with and without lters. The pressure drop was measured by a differential pressure gauge (Summit digital Manometer 635).

Results and discussion
Electrospinning has been intensively studied to prepare nano-bers from different polymer solutions. We added PEO into silk broin solution to improve the electrospinnability because viscosity plays an important role in the electrospinning process (Fig. 1b). The viscosity of pure silk broin solution is relatively low that the balance between the viscoelastic force, surface tension and electrostatic repulsion is easy to destroy, so the jet would break into droplets instead of forming a continuous ber. The feeding rate, applied voltage at the tip and tip-tocollector distance were nely adjusted to generate uniform nanobers. The transmittance curves with the electrospinning time ranging from 10 min to 50 min are shown in Fig. 2a. The nanober membranes have a transmittance over 50% at most visible region when the corresponding electrospinning time is less than 30 min. Aer the nanober membranes were collected on the nylon ber nonwoven frame (possesses negligible ltration efficiency and pressure drop), they were immersed in an ethanol solution for 10 min and then incubated in distilled water for 36 h at room temperature to remove PEO. The ethanol annealing process was carried to induce the transition of bsheets, preventing the silk broin phase from dissolving in water and contributing to the superior strength of silk broin nanober. 33 Aer the treatment with water, the membranes were dried in a vacuum at room temperature.
FTIR spectra of SF/PEO blend nanober, PEO extracted SF nanober, and pure PEO nanober are shown in Fig. 2b. For pure PEO, the main peaks are at 2874 cm À1 and 1104 cm À1 due to C-H stretching and C-O-C stretching vibrations. For samples containing SF, the characteristic b-sheet peak was observed at 1516 cm À1 (Amide I) and 1628 cm À1 (Amide II) while the N-H stretching vibration was observed at 3282 cm À1 respectively, indicating the existence of amide bond. The FTIR spectra of PEO extracted SF did not show the feature peaks at 2874 cm À1 and 1104 cm À1 , conrming that PEO are extracted from the SF/ PEO blend ber aer incubating in water and the peaks at 1516 cm À1 and 1628 cm À1 indicated that SF was reserved. These results demonstrated that aer incubating in water for 36 h, PEO phase was successfully removed from the SF/PEO blend nanober.
Besides the FTIR analysis, morphological changes of SF/PEO blend nanober aer incubating in water were further examined by FE-SEM ( Fig. 3a and b). The diameter of SF/PEO blend nanober produced from SF solution was about 400 nm in average. When the SF/PEO blend jet was stretched from the Taylor cone, a thin skin layer formed at the liquid-gas interface, which is enriched with PEO. 34 Aer water incubation, the PEO phase rst dissolved then the surface of ber became rough and the diameter contracted a little (Fig. S1, ESI †). The size of these randomly spread globules on the surface were about 20-80 nm. The cactus-like structure provides the ber with higher surface energy and improve the wettability of the nanober membrane, which will be conrmed in the water contact angle test. When the water incubation time increases to 60 h that permits sufficient time for the hydrophilic broin chain block dissolve, the remaining crystalline part will self-assembly into a lamellar-like structure (Fig. 3c), nally transforming into a hydrophobic surface.
To compare the morphology, surface property and ltration performance of various bers, the mostly studied PAN nano-bers were fabricated via electrospinning as well. By changing the applied voltage, electrospinning time, solution concentration, and the distance between the tip and collector, the diameter and pore size distribution of PAN nanobers were nely tuned according to the SF-based one. The diameter of the electrospinning ber increased with a higher feeding rate or higher concentration. More details could be found in the Methods part. PAN nanobers membranes were fabricated by a traditional electrospinning process using dimethylformamide as solvents and then transported to a nylon nonwoven frame to form the air lter.
For all these brous air lters with a same diameter and pore size distribution, the wettability was examined by water contact angle measurement. Water contact angle indicates the wettability of a membrane surface, and an ideal ltration membrane is one that produce a high ux without clogging or fouling. The water contact angle of different kinds of nanober membranes were summarized in Fig. 3d. The pristine PAN brous membrane showed a hydrophobic surface whereas all silk broin membranes showed a good hydrophilicity. Fig. 3d clearly shows that the water contact angle of PEO extracted silk broin nanober membranes is almost constant up to 8 degrees, which indicates excellent hydrophilicity. The effect of surface wettability on the ltration performance will be discussed in the next section.

Filtration performance
To better understand this section, the reader is recommended to read this article, 35 which has thoroughly characterized the physical and chemical properties of the most universal and representative PM pollutants. In this study, the PM was generated by burning cigarette since the ingredient of smoke generated by the cigarette is similar to that of PM pollution in the air. 36 The particle size distribution of the smoke generated by burning cigarette was measured by a particle counter, which shows a broad distribution from 0.3 mm to 10 mm and a large proportion of PM2.5 (Fig. S3b, ESI †). The concentration of the smoke in the le cabin was diluted with clean air to a severe level, where the air quality index (AQI) reading was at 500. The nanober membranes were pasted between the cabins, blocking the diffusion of smoke from the le cabin to the right one (Fig. S3c, ESI †). We also cut a piece of commercial air lter to evaluate the ltration performance.
The ltration results were shown in Fig. 4a. All air lters share a similar thickness and pore size distribution (pore distribution of SF and PAN nanober membranes are shown in Fig. S2, ESI †). The removal efficiencies are calculated by measuring the PM particle number concentration with and without air lters. It could be clearly seen that for PM10-2.5, silk broin and PAN nanober air lter exhibited a good removal efficiency of 99.99% and 99.89%, respectively. While the removal efficiency of the commercial lter was only 88.02% for PM10-2.5. Data in the PM10-2.5 and PM10 columns indicates that no matter what material the nanometer-scale brous lters is made from, they all showed a higher removal efficiency (>99%) for coarse particles than the commercial air lter. This is because of the size-based ltration mechanism of the brous ltration system. 16 Not taking the gravity effect into account, there are four main ltration mechanisms of ber material: interception, Brownian diffusion, inertial effect and electrostatic effect. For coarse airborne pollutants, they will be trapped easily by the interception and inertial impaction from the nanober air lter. However, they can still penetrate the sinuous channels formed by micron-scale ber membrane. The surface morphologies of different lters before and aer ltration were studied by using a FE-SEM (Fig. S3c, ESI †). Aer the ltration, PM particles were obstructed by the nanober net and the ber diameter increased. However, the commercial air lter showed a much larger ber diameter and the capture of PM was much less than those we fabricated by electrospinning. For particles having an aerodynamic diameter < 2.5 mm, which does not have enough inertia to be captured by interception impaction and is mainly captured by the interaction between ber and particles, the surface property of different lters counts. It is noted that silk broin nanober air lter showed an enhanced ltration efficiency for PM2.5, increasing from 96.11% to 99.84% compared with the synthetic polymer lter.
In addition to ltration efficiency, pressure drop is another important parameter for air lters, as the pressure drop is related to the maintenance and remaining use time of the lter. Here, we investigated the pressure drop of silk broin nanober lter and the PAN one under different air ow velocities. Fig. 4b shows the relationship between the pressure drop and the velocity of the air ow. As the lter face velocity increased from 0.2 to 1 m s À1 , the pressure drop of cactus-like SF nanober lter increased linearly from 75 to 374 Pa and the pressure drop of PAN one increased from 133 to 650 Pa, respectively. The pressure drop of SF-based membranes is only half of that of PAN. The result is consistent with the theoretical equation that the pressure drop increases with increasing the air ow velocity: where m and U 0 are gas viscosity and velocity, L and d f are the thickness of the lter and diameter of ber, respectively, and f(a) represents the pore size distribution. 37 Interestingly, the silk broin nanober lters possessed a relatively lower pressure drop than the PAN one although they shared a similar ber size and packing density. This implied that there were other undetected but critical factors that control the air ow resistance.
To quantitatively compare the ltration performance of the SF-based nanober membrane and PAN nanober membrane, the quality factor was calculated aer each experiment. The quality factor (QF) is a standardized parameter that correlates the removal efficiency and pressure drop to evaluate the ltration performance of an air lter, QF ¼ Àln(1 À E)/DP, where E is removal efficiency and DP is the pressure drop. As the air ow velocity increases, the QF of SF-based nanober membrane increases more steeply than that of PAN nanober membrane (Fig. S3d, ESI †). The above results indicate that the SF-based nanober membrane is able to improve the ltration efficiency of PM pollutants while decreasing the air-ow resistance.
To further understand the ltration performance improvement of SF-based brous ltration media, special attention was paid to the morphological changes of the uffy soot aggregates during the ltration process. Concretely, FE-SEM was employed to observe the morphologies of PM captured by nanober membrane at different time during a continuous ltration process. Fig. 5a shows the SF-based nanober membrane before capturing PM pollutants and the time sequence of PM capture is shown in panels (a-f) in Fig. 5. At the initial stage, as the smoke owed past the nanober net, PM particles were randomly captured by the ber and wrapped tightly on the nanober net. Subsequently, as the capture kept going on, the particles captured on the nanobers are able to deform and spread over the nanober surface, without forming any clogging and destroying the tortuous channel. Finally, the cactus-like structure became smooth and the diameter increased uniformly from 400 nm to 600 nm aer capturing the PM particles. These observations demonstrate that the PM pollutants could deform and spread on the silk broin nanobers. Schematic illustration of the capture, deformation, coalescence, and growth of the uffy soot aggregates during the ltration process was shown in Fig. 5g. The shape variations were mainly determined by the competition between the surface tension of pollutants and the adhesion between pollutants and nanobers. For the pollutants captured by the ltration membranes, the general form of the energy is E ¼ ∯g pv dA pv þ ∯g sp dA sp þ ∯g sv dA sv þ ∰rghdV where g pv , g sp , and g sv are the interfacial tension of pollutantvapor, solid-pollutant, and solid-vapor, respectively. A pv , A sp , and A sv refer to the contact area between the pollutant-vapor, the solid-pollutant, and the solid-vapor interfaces. The fourth term represents the gravitational potential energy and once the pollutants are captured by the nanober, the variation of gravitational potential energy can be neglected. According to the principle of energy minimization, the surface energy released in the deformation process is where DA pv , DA sp , and DA sv refer to the variations of area of the pollutant-vapor, the solid-pollutant, and the solid-vapor interfaces, respectively. In the process of PM particles deforming and spreading over, the area of the solid-vapor interface reduced is equal to that of solid-pollutant interface increased. Then the released surface energy is During the spreading process, the volume of the PM particle remains unchanged, so the above equation can be transformed as where R and r refer to the radius of nanober and PM particle before deformation and L and d refer to the length and thickness of the PM particle aer spreading over (details of the equation solution procedure are provided in ESI †). A surface plot of the result was generated by the matplotlib module in Python to visually illustrate the impact of r and R (Fig. 5h). As the r increases, the released surface energy will increase, which favors the viscous dissipation of pollutants and gives kinetic energy to overcome the inertial surface tension forces. Thus, the uffy soots are able to spread over and surround the ber axisymmetrically and the nanometer-scale channels formed by the SF-based nonober were well reserved. For PAN nanober air lters, the time sequence of PM capture is shown in panels (a-f) in Fig. 6. As the smoke owed past the PAN nanober, PM particles were rst randomly bound around the smooth surface of PAN nonober or stuck on one side of the nanober only. As the ltration process went on, more PM pollutants were trapped by the nonober membrane. Then the small particles aggregated to form larger ones and nally reach to a stable spherical shape, forming beaded bers and blocking some of the sinuous channel eventually. Fig. 6g depicts a simplied illustration of the capture, coalescence, and growth process of the uffy soots on PAN nanober air lter. When calculating the energy of this system, the gravitational potential energy is neglected as well. During the coalescence process, the surface area of the two small PM particle contracts and the volume stays unchanged. So the variation of interfacial energy in this process can be written as: where g pv , g sp , and g sv are the interfacial tension of pollutantvapor, solid-pollutant, and solid-vapor, respectively. DA pv , DA sp , and DA sv refer to the variations of area of the pollutantvapor, the solid-pollutant, and the solid-vapor interfaces.
According to the classical Young-Dupre equation 38 that cos q ¼ (g sv À g sp )/g pv then the above equation can be transformed as: with DA pv < 0, cos q < 0, DA sp < 0, and the interfacial tension being a positive number. Aer a simple calculation, we can state that DE S < 0 always holds. In other words, the uffy soots can deform and aggregate spontaneously with the release of surface tension as driving force. With the continuous feeding of smoke, more pairs of adjacent uffy soots coalesced into larger ones to minimize their surface energy. As a result, the air-ow resistance increases because the sinuous channels were blocked by these big lter cake. According to the experimental phenomena and analysis of energy law as we demonstrated above, the excellent ltration performance of SF-based brous ltration media is mainly related to the surface tension of silk broin and the adhesion between pollutants and SF nanobers. Silk broin is enriched of various functional groups, including hydroxyl (-OH), amide bond (-CO-NH-), phenolic hydroxyl (-C 6 H 4 OH) etc., which are excellent target sites for the coupling of various particles or chemicals. Pollutants could easily deform and spread on the hydrophilic SF nanobers but they tend to aggregate, contract and nally form ball-shaped conformations on the hydrophobic PAN nanobers. Compared with the well reserved pathways inside the SF nanober membrane, the sinuous channels inside the PAN nanobers membrane were heavily blocked. In short, the conformation difference of the pore structures between SF membrane and PAN membrane during the ltration process plays a key role in changing the air ow streams and thus affecting the pressure drop. There are two main waste management strategies, landlling and incineration. Each of these methods has dangerous side effects. Landlling is prone to producing toxins, leachate and toxic gases while burning waste emits toxic gases and particulates into the air. When a traditional air lter runs out of its use  time, it will be disposed and burned, making the air pollution much worse. Here, we shed a light on the biodegradability of the brous ltration media. In vitro degradation was carried out by incubating the SF and PAN nanober membrane in 50 ml of phosphate-buffered saline (PBS) containing 2.3 U ml À1 protease XIV at 37 C. The results (Fig. 7) indicated that aer incubating in protease XIV solution for 24 h, weight loss of the SF nanober membrane was about 78% while there was a loss of about 20% in weight aer 2 weeks in PBS solution, which can be attributed to the degradation of the residual silk I and/or non-crystalline regions in silk broin nanober. We should mention that PAN nanober membrane incubated in buffered protease XIV solution or PBS only were stable through the whole incubation period. Compared with the synthetic polymer lter, SF air lter can directly degrade into peptides and amino acids, which is non-toxic and can be easily metabolised by microorganisms and macroorganisms in natural environment. [39][40][41] This indicates that the biodegradation products of silk broin materials do less or no harm to the environment. As shown above, silk broins biomaterials are desirable for air ltration because it offers the possibility to be eliminated in the natural environment.

Conclusions
In summary, eco-friendly membranes based on silk broin for high-performance ltration media have been successfully fabricated by a green electrospinning process, which uses water as solvent. This nanober air lter showed a superior ltration performance than the state-of-the-art due to its richness in functional groups and good wettability to pollutants. Signicantly, a mechanism describing the air ow resistance was proposed based on the principle of energy minimization. Ultimately, the resultant silk broin brous ltration media are able to improve the ltration efficiency (99.99%) for particle matter with a broad range of size from 0.3 mm to 10 mm while decrease the pressure drop (75 Pa) and present the advantage to be biodegradable. We envision that the green fabrication process and the use of sustainable material would pave the way for the next-generation air ltration system, which exhibited robust ltration performance, excellent biodegradability, and cost-effective.

Conflicts of interest
There are no conicts to declare.