Renewable Energy from Livestock Waste Valorization: Amyloid-Based Feather Keratin Fuel Cells

Increasing carbon emissions have accelerated climate change, resulting in devastating effects that have become tangible on an everyday basis 1 . This is mirrored by a projected increase in global energy demand by ca. 50% in a single generation timespan 2,3 , urging a shift from fossil-fuel derived materials toward greener materials and more sustainable manufacturing processes 4 . Bio-based industrial byproducts, such as food industry side-streams 5 , are attractive alternatives with a strong potential for valorization owing to their large volume, low cost, renewability, biodegradability, and intrinsic material properties 6 . Here we demonstrate the reutilization of industrial chicken feathers waste into proton conductive membranes for fuel cells 7 , protonic transistors 8 , and water-splitting devices 9 . Keratin was isolated from chicken feathers via a fast and economical process, converted into amyloid �brils upon heat treatment, and further processed into membranes with imparted proton conductivity through a simple oxidative method 10 . The functionality of the membranes is demonstrated by assembling them into a hydrogen fuel cell capable of powering several types of devices using hydrogen and air as fuel. Additionally, these membranes could be used to generate hydrogen by water splitting, and in protonic eld-effect transistors as a thin-lm modulating protonic conductivity via the electrostatic gating effect. We believe that by converting industrial waste into renewable energy materials at low cost and high scalability, our green manufacturing process can contribute to a fully circular economy with neutral carbon footprint.


Full Text
Global energy demand has been on the continuous rise and is anticipated to further increase by an astonishing 47% in the near future 2 .Current primary energy sources, which are still derived from nonrenewable fossil fuels, are unsustainable and detrimental to the environment due to the large emissions of greenhouse gases (GHG) into the atmosphere.Hence, to curb global warming, there is an urgent need for new technologies based on renewable and sustainable energy generation sources.
Fuel cells are considered a promising sustainable technology for energy conversion to generate electricity without CO 2 emissions using electrochemical reactions 11 .Among the various types of fuel cells, hydrogen fuel cells are particularly attractive from a decarbonization point-of-view, with efforts focusing on the production of green hydrogen as the next renewable fuel 12 .The heart of a fuel cell lies in the membrane electrode assembly (MEA), which consists of a proton exchange membrane (PEM) that selectively allows only the passage of protons between the electrodes.Per uorosulfonic acid (PFSA) membranes such as Na onÔ and Aquivion Ò remain the benchmark state-of-the-art PEMs used in fuel cells due to their high proton conductivity, chemical inertness, and mechanical stability.However, the production of Na on is very expensive due to the use of uorinated chemicals 13 and pose serious environmental concerns relating to poor biodegradability and environmental toxicity of per uorinated materials 14 .Moreover, the functionalization of polymers through sulfonation usually involve highly corrosive reagents, reducing their environmental friendliness, [15][16][17] boosting the search for greener and more economical materials for large-scale applications.
Inspired by the ubiquity of proton conduction in the living world 18,19 as well as their intrinsic biodegradability, studies have explored biopolymers (proteins and polysaccharides) as potential proton conductive materials 20,21 .The proton conductivity of naturally occurring proteins is generally too low to be useful in practical applications but has been improved with functional group modi cations 8,22 .
Alternatively, high proton conductivity can be achieved in biopolymers through de novo design of peptides utilizing nanostructure-aided conductivity 23 or metal oxide-peptide composites 24 , but these approaches still face challenges regarding cost and scalability.
Chicken feathers are produced at a staggering rate of 40 million tons annually as a byproduct from the poultry industry 5 , but have been underutilized due to their unsuitable nutritional pro le.Furthermore, they face disposal challenges due to the generation of toxic sulfur dioxide from incineration.Efforts to valorize chicken feathers include their use as hydrogels 25,26 , biosorbents for heavy metals 27,28 , functional lms 29 and bers 30 .These studies point out that feather keratin may be re-con gured into highly functional materials, notably exploiting the structural properties and high cysteine (Cys) content of keratin proteins.However, no attempts have been made to date to revalue keratin into renewable energy devices.In this work, we chose chicken feathers as a starting material for fabricating proton conductive biomaterials.Keratin proteins were extracted from industrial chicken feathers using an environmentally friendly and scalable process and converted into amyloid brils that were subsequently processed into free-standing amyloid bril membranes.Proton conductive properties were imparted through a postoxidative treatment which converted Cys thiols into sulfonic acid groups, using a benign environmental process with harmless and inexpensive chemical compounds.The performance of the membrane was demonstrated in a fuel cell device capable to transform by electrochemical reaction H 2 and O 2 (the latter directly from air) into electrical power and mechanical work (with H 2 O as sole byproduct).To demonstrate the general potential of this approach, we highlight two additional applications of these materials, as protonic eld-effect transistors, as well as in the generation of H 2 via water splitting.

Protein extraction and isolation
An overview of the production of feather keratin proton conductive membranes is depicted in Fig. 1.
Keratin was extracted and isolated from chicken feathers and heat treated to produce keratin amyloid brils (Fig. 1a), after which they were mixed with a crosslinker and dopant and cast onto a substrate to produce a free-standing membrane.The membrane was then heat cured followed by an oxidative treatment to produce a modi ed membrane with imparted proton conductive properties (Fig. 1b), which was then assembled into a fuel cell (Fig. 1c) or other devices utilizing proton conductivity.
Among protein-containing industrial waste byproducts, chicken feathers have the highest protein content (ca.90%) and are most abundant in Cys amino acid (ca.8%, Fig. 2a) 5 .However, the extraction and isolation procedure of feather keratin on a large scale has faced challenges, notably the use of harsh toxic solvents (e.g.sodium sul de) and reducing agents such as b-mercaptoethanol and dithiothreitol, or long isolation dialysis treatments 31 .Here, we extracted feather keratin using a basic solvent consisting of urea and thioglycolate.Urea acted as a chaotropic solvent disrupting the hydrogen bonds within the compact structure of feathers, while thioglycolate served as a reducing agent to reduce intra-and intermolecular disul de bonds, resulting in the separation of protein inter-chains and eventually dissolution.The supernatant was then precipitated and washed to obtain a crude keratin isolate.We reextracted this isolate in ammoniacal solution and precipitated it to obtain a pure keratin isolate, which displayed only a single protein band around 10 kDa as analyzed by electrophoresis (Fig. 2b), demonstrating minimal protein degradation during the entire process.An amino acid analysis of the obtained keratin isolate revealed 7 mol.% of Cys (Supplementary Fig. S1), in agreement with the expected 8 mol.% from earlier reports 32 .Our process produced a feather keratin isolate almost fully soluble in dilute acid, in contrast to keratin isolates obtained from other extraction procedures, which require concentrated acids 33 .Furthermore, the process does not require dialysis, enabling scalability and fast production rates.

Protein brillization
Within the family of keratin proteins, hair and wool keratins belong to a-keratins composed of a-helix intermediate laments, while keratin from feathers, claws, and beaks belong to b-keratins dominated by b-sheet secondary structures 34,35 .While early x-ray diffraction (XRD) studies on feather barbs suggested the presence of amyloid-like brils 36 , the self-assembly of regenerated feather keratin into amyloid brils has not yet been reported to date.Sequence analysis of feather keratin from the UniProt database (P02450) using a variety of protein structure prediction algorithms (Supplementary Fig. S2) indicated that feather keratin could be assembled into amyloid brils.The structure of feather keratin has been reported to consist of a central region with predominantly hydrophobic residues adopting a b-sheet conformation 32 , with Cys residues primarily located at the N-and C-termini 37 .The highest propensity for b-sheet aggregation was predicted to occur between residues Val 32 -Leu 43 , which lie within this central region and have also been identi ed as a hotspot for amyloidogenic aggregation.Thus, we postulated that feather keratin could be self-assembled into amyloid brils, which was veri ed by the facile self-assembly of keratin monomers into amyloid brils after heat treatment at 90°C under acidic conditions.Using thiazole orange as a molecular probe for amyloid brils 38 , the brillization process proceeded relatively quickly and reached a plateau after only 2h, exhibiting a faster kinetic than that of animal-derived b-lactoglobulin 39 or plant-derived proteins 40 (Fig. 2c).Solutions of feather keratin amyloid brils also exhibited birefringence under cross-polarized light, corroborating the presence of amyloid brils (inset of Fig. 2c), which were found to be several micrometers in length after 5h as imaged by Atomic Force Microscopy (AFM) and Transmission Electron Microscopy (TEM) (Fig. 2d and 2i).The nano brils exhibited an average height of about 7 nm (Fig. 2e), a semi exible behavior with comparable contour (Fig. 2f) and persistence (Fig. 2g) lengths in the range, and periodic pitches of 90 nm along the bril axis (Fig. 2h), as expected for most amyloid brils.

Membrane fabrication
We then mixed feather keratin amyloid brils with glyoxal and mercaptosuccinic acid and cast the solution to obtain a clear free-standing exible membrane (Supplementary Fig. S3a), followed by thermal curing (Supplementary Fig. S3b).As we hypothesized that proton conduction could be imparted into the membrane upon conversion of thiol groups into sulfonic acids, we immersed the membranes in peracetic acid to oxidize thiols and disul de bonds to sulfonic acids (Supplementary Fig. S3c) 10,41 .The nal membrane is denoted as Keratin-M.
Oxidation into sulfonic acid was corroborated by both Fourier Transform Infrared Spectroscopy (FTIR) and Raman Spectroscopy, with the appearance of the S=O peak at 1040 cm -1 42 (Fig. 3a) in the FTIR spectrum, and the disappearance of the Raman shift peak at 2550 cm -1 assigned to the thiol 43 (Fig. 3b).
This was further supported by XPS measurements, which showed the major peaks of -SO 3 -2p 3/2 and 2p 1/2 doublets at 168.2 and 169.6 eV, respectively, after deconvolution [44][45][46][47] .Peaks of sulfone and sulfate groups were also observed, possibly due to side reactions in the fabrication process (Fig. 3c).As the modi cation of thiols into sulfonic acids confers a negative charge, the difference in surface potential before and after modi cation was detected by Kelvin Probe Force Microscopy (KPFM).Before modi cation, the keratin casting solution deposited onto the Highly Oriented Pyrolytic Graphite (HOPG) substrate exhibited a relative positive potential, which likely arose from positively charged amino acids Arg and Lys.Upon oxidative treatment, the membrane exhibited a relative negative potential that may have resulted from the conversion of neutral thiols into negatively charged sulfonic acids (Fig. 3d and Supplementary Fig. S5).
Throughout the fabrication process, the membrane maintained its shape and mechanical integrity while still possessing birefringence arising from amyloid brils under cross-polarized light (Supplementary Fig. S3d), suggesting that the amyloid brils were preserved without signi cant alteration.The XRD pattern of the membranes displayed two dominant peaks at 9.3° and 19.6°, corresponding to d-spacings of 4.5 Å and 9.5 Å, respectively, that re ect the signature cross-b structure of amyloid brils 48,49 (Fig. 3e).A smooth surface was observed under SEM for keratin-M membranes (Supplementary Fig. S6), whereas at higher magni cation a brillar network morphology was identi ed on the surface with pore sizes of 7-20 nm (Fig. 3f).In contrast, neat keratin membranes showed a denser morphology with lower porosity (Supplementary Fig. S7).

Membrane properties
The modi cation of neutral thiols into negatively charged sulfonic acids was also indicated by the increase in ion exchange capacity (IEC) from 0.18 meq g -1 to 0.84 meq g -1 after oxidative treatment, while a further increase to 1.56 meq g -1 was achieved with the addition of MSA due to the contribution of additional thiol groups (Table 1).The boost in IEC also resulted in a higher water uptake due to enhanced hydration and charge repulsion from the anionic sulfonic acid groups, which could possibly explain the larger pore size from keratin-M due to higher swelling compared to neat keratin membranes.Keratin-M membranes maintained good barrier properties despite the presence of nanopores, demonstrating almost no permeability to dye molecules such as Rhodamine B (Supplementary Fig. S8).They also exhibited 3fold less permeability to small molecules such as methanol compared to Na on (Supplementary Fig. S9), demonstrating their potential in direct methanol fuel cells in which Na on performs less e ciently due to its high methanol permeability 50 .The membranes were mechanically robust and could be readily manipulated without damaging them.Their thermo-mechanical characterization is reported in the supplementary information (Young's Modulus and tensile strength in Supplementary Table S1 and Fig. S10; thermal behavior and glass transition in Fig. S11).Supercritical CO 2 -dried keratin membranes measured with TGA displayed a 10% decrease in weight until 100°C attributed to the evaporation of water, followed by thermal degradation and deamidation with drastic weight loss (~ 60%) between 300°C to 400°C 51 (Supplementary Fig. S12).The main parameter in a fuel cell is the proton conductivity (s), which is essential for any polymer electrolyte membrane.After the successful conversion of thiols into sulfonic acids, feather keratin membranes exhibited an enhanced proton conductivity that increased with the MSA content from 0.02 mS cm -1 to 6.3 mS cm -1 in water (Fig. 4a).The appearance of the semi-circle in the Nyquist plot of neat membranes indicate a more capacitive behavior (Fig. 4b), as also evidenced in the large phase angle from the Bode plot (Supplementary Fig. S13a).This characteristic shifted towards a membrane with a resistor-type behavior for keratin-M membranes, as seen from an almost absent semi-circle in the Nyquist plot (Fig. 4c) and a phase angle approaching 0° at high frequencies in the Bode plot (Supplementary Fig. S13b).Comparing across biomaterials reported to date, the proton conductivity of keratin-M is the highest among all biopolymers, with the exception of speci cally engineered peptides (Supplementary Table S2 and S3).But in contrast to the latter, feather keratin membranes are processed from low-value waste materials with a strong potential for scaling-up.Remarkably, because this comes from a waste stream intended for incineration (and CO 2 emissions), this process takes place with an overall negative carbon footprint, adding value to both sustainability and environmental friendliness.Furthermore, the proton conductivity of the membranes could still be further improved by doping with acid electrolytes, notably with sulfuric acid which provided the most signi cant increment to 22.8 mS cm -1 , and which may nd applications in vanadium redox ow batteries.

Fuel cell performance
Having demonstrated a suitable proton conductivity for the ensued amyloid keratin membranes, we then tested their in-situ performance as a polymer electrolyte membrane in a hydrogen fuel cell (Supplementary Fig. S14).The open circuit voltage (OCV) of a fuel cell has been reported to provide a useful indication of the integrity and degradative behavior of the membrane such as thinning and pinhole formation 52 .The cell assembled with keratin-M membrane displayed an OCV between 0.95 -1 Vcomparable to Na on -and polarization curves were obtained at increasing temperature.With air at the cathode, the cell generated an increased power density with temperature, reaching up to 20 mW cm -2 at 65°C (Fig. 4d).Similarly, the peak power of 25 mW cm -2 was generated with pure oxygen at 55°C, after which it decreased to 21 mW cm -2 at 65°C (Fig. 4e).This could be due to an increase in reactive oxygen species generated with increased oxygen concentration, leading to the formation of free radicals such as OH×, H×, and HOO× that have been reported to initiate membrane degradation 53,54 .Further increase in temperature to 80°C resulted in a drop in OCV to 0.85 -0.9 V along with a decreased power density lower than that at 65°C.To ensure a high OCV and optimal performance, a membrane of low hydrogen permeability is desired.Utilizing staircase voltammetry with N 2 at the cathode (Supplementary Fig. S16), the hydrogen permeability of the keratin membrane was assessed and calculated using the intercept, yielding 3.8 ± 0.3 ´ 10 -10 mol cm -2 s -1 , (compare to Na on -3.8 ´ 10 -9 mol cm -2 s -1 ) 55 while the electrical resistance obtained from the slope was approximately 3200 ± 140 Ohm cm 2 (Supplementary Fig. S17) 56,57 .
To demonstrate the applicability of the keratin-M membrane in hydrogen fuel cells, we assembled the membrane into a commercial test fuel cell setup.With hydrogen and air as the respective fuels at the anode and cathode, the cell was able to generate power to turn on both red and white LED lamps (Supplementary Fig. S18).In addition, the cell was responsive to the presence of fuel, turning the LED lamp on and off with the introduction and absence of hydrogen (Supplementary Movie S1).Most importantly, the fuel cell could perform mechanical work, e.g.drive a fan setup driven by a motor (Supplementary Movie S2) as well as a fuel cell toy car ((Supplementary Movie S4 and Fig. 4f).Furthermore, the keratin-M PEM membranes could be further used as an electrolyzer for the production of hydrogen and oxygen from water when applying an electric bias, as observed by the formation of bubbles at the respective outlets (Supplementary Movie S4 & Supplementary Fig. S19).
Transistor performance Finally, we demonstrate the application of keratin-M as a solid-state lm for the fabrication of protonic eld-effect transistors (H + -FETs).Devices were fabricated by casting keratin-M lm between two gold/palladium hydride (PdH x ) electrodes pre-deposited using e-beam evaporation process on a 50 nm hafnium oxide (HfO 2 ) gate dielectric layer (Fig. 4g).The conductivity of the keratin-M lm was observed to be dependent on relative humidity (RH), displaying a surge in current when the RH increased from 50% to 90% (Fig. 4h and Supplementary Fig. S20), as a result of the increase in proton conductivity with increased hydration, which is commonly observed in most proton conductive materials 20,22,58 .The transfer characteristics of the FET device (Fig. 4i) measured for negative and positive gate voltages (V GS ) at a source to drain voltage (V SD ) of 0.5 V exhibited a ON-to-OFF ratio of ~ 1.4 with a low leakage current (Supplementary Fig. 21).The output characteristics of the device displayed a similar trend whereby the channel conductivity increased upon application of negative V GS (Supplementary Fig. 23 and 24) .Biasing the device with negative V GS induced positive charges into the channel proportional to V GS as the keratin-M and Si gate form a capacitor with HfO 2 as the dielectric.This resulted in an increased positive charge carrier concentration and thereby channel conductivity.The electrostatic gating effect highlights the potential use of keratin-M in H + -FETs.

Conclusions
We have demonstrated the fabrication of proton conductive membranes isolated from chicken feathers keratin through a green and low-cost process with high scalability.Amyloid-based keratin membranes were successfully used in a hydrogen fuel cell to power several devices with only hydrogen and air as the fuel, and also demonstrated pronounced electrostatic gating effect in a transistor and hydrogen generation in an electrolyzer.While the ensued proton conductive membranes do not yet fully match the performance (particularly at high temperature) of commercial benchmarks such as Na onÔ and Aquivion Ò , the membranes pose a radically different environmental and sustainability footprint and at a potentially much lower cost: our current estimates on the total production from raw material to membrane give ca.USD 850/m 2 .This value is at least half that of Na onÔ (USD 1500 -3000/m 2 ) and is anticipated to be much lower when scaled up and optimized, hence highlighting another long soughtafter feature of these original materials.Our study unambiguously demonstrates that livestock industrial waste can be reprocessed into valuable materials for renewable energy devices using a fully sustainable process, and brings us one step closer towards a circular economy concept based on zero -or even negative-carbon emissions.

Materials and Methods
Full details on Materials and Methods are provided in the Supporting Sections A-D.Hereby a summary of the main methods used is provided.

Protein extraction and isolation to membrane fabrication
Protein extraction and isolation from chicken feathers obtained from a local farm are detailed in Supplementary Section AI.Feather keratin amyloid brils were prepared by heating keratin isolates (Supplementary Section AII) and fabricated into membranes as described in Supplementary Section AIII.
Protein and membrane characterization are provided in Supplementary Section B and C, respectively.Electrochemical measurements of membranes are described in Supplementary Section D.

Fuel cell measurements
For the in-situ fuel cell measurements, membranes were placed inside a 1 cm 2 cell between two platinumcoated commercial GDE's (JM ELE0244, Johnson Matthey, United Kingdom) with a nominal loading of 0.4 mg Pt/cm 2 .Fuel cell tests were performed using a custom fuel cell test bench with a constant ow rate of 0.4 ln/min for all gasses.Polarization curves were recorded at 100% RH and a pressure of 1.5 bar.Hydrogen crossover currents were measured using staircase voltammetry at 100% RH and a pressure of 1 bar.Additional details are provided in Supplementary Section IV.To demonstrate powering of the devices, membranes were placed between two platinum-coated carbon paper electrodes of 0.5 mg/cm 2 (Fuel Cell Store, USA) to form the membrane electrode assembly in the Flex-Stak Electrochemical cell (Fuel Cell Store, USA).Pure H 2 and Air were supplied at the anode and cathode, respectively.Metal tabs were connected to a mini DC/DC (LiPower, Sparkfun) converter with an output of 3.3 V.

Transistor fabrication
Field-effect transistor devices (channel length of 100 mm and width of 1 mm) were fabricated by depositing gold/palladium metal electrodes using electron-beam evaporation on top of hafnium oxide (50 nm) coated silicon substrate.Palladium electrodes were exposed to pure hydrogen gas for 5 mins in a closed chamber for their transformation to PdH x .0.5 mL of amyloid bril mixed with glyoxal and MSA were deposited onto the device and air-dried.The substrate was then cured at 150°C for 50 min and immersed in peracetic acid solution at 37°C for 5 h.The device was then immersed in 0.5 M sulfuric acid for 1 h and then rinsed thoroughly with water.

H + -FET Characterization
Output and transfer characteristic of the transistor devices were recorded using the Keysight B2901A precision source measure unit.Relative humidity level was kept at 95% in a custom-built dry box chamber using a humidi er (BioAire Lifestyle).

Supplementary Files
This is a list of supplementary les associated with this preprint.Click to download.SupplementaryMovie1.mp4

Figure 1 General
Figure 1

Figure 2 a
Figure 2

Figure 3 a
Figure 3

Figure 4 a
Figure 4

Table 1 .
Feather keratin membrane properties before and after treatment