Bio‐Derived Wood‐Based Gas Diffusion Electrode for High‐Performance Aluminum–Air Batteries: Insights into Pore Structure

Gas reactant transport plays a crucial role in various gas‐consumption reaction‐based electrochemical devices, but the pivotal performance limitation still centers on gas diffusion electrodes (GDEs). To this end, natural cross‐cut wood to prepare GDEs for aluminum–air batteries is introduced by utilizing the ordered structure with microchannels. With cobalt–nitrogen co‐doped carbon nanotubes as the oxygen reduction reaction catalysts grown inside the channels of carbonized wood slice and wettability gradient modification, cherry‐based GDE achieves higher power density (267 vs. 236 mW cm−2) than the commercial carbon fiber paper‐based electrode. By bridging the identified characteristics of pore structure via deep‐learning image recognition technology with the permeability of other three typical kinds of (ash, pine, and oak) wood‐based GDEs, it is revealed that the ratio of effective porosity to the average pore size is key to the performance. This work demonstrates the feasibility of bio‐derived wood‐based materials for fabricating high‐performance GDEs and provides insights into pore structure for the rational design of structured electrodes.


Introduction
Gas consumption reactions are of significance for energy conversion and environmental protection technologies such as fuel cells, [1] metal-air batteries, [2] CO 2 reduction, [3] and hydrogen peroxide production. [4][11][12] To improve the performance, besides improving the intrinsic activity of catalysts, the elimination of the undesired mass transfer resistance of gas species should be equally important.To this end, gas diffusion electrode (GDE), usually consisting of commercial carbon fiber paper (or cloth) and catalysts, has been proposed and commercially applied to enhance the gas transfer owing to their ideal electrical conductivity and mechanical strength.It should be noted that the cost of carbon paper (1400 $ m −2 ) for both cathode and anode would take up more than half of the total cost of a fuel cell device, even higher than the catalyst coating membrane (CCM) section. [7,13]However, the commercial carbon fiber paper is often fabricated by randomly knitting short carbon fibers, thus the pore size, as well as the gas diffusion channels, are usually disordered, which would induce unwanted sinuous or discontinuous routes.According to the Kozeny-Carman (KC) theory for a typical single-phase diffusion model, [14,15] the gas permeability  is inversely proportional to the tortuosity , where  is defined as the ratio of the sinuous channel length to the sample length.[18][19][20][21] Recently, the emerging 3D-printing techniques have demonstrated great potential for fabricating the required ordered structures as compared with traditional manufacturing, [22][23][24] with substantially improved precision, speed, and building volume.However, even though some precise structures have been manufactured from metal or carbon materials (e.g., graphene), [25] the commercial printable materials are still mainly limited to polymers for most scenarios.More importantly, large-scale manufacturing is challenging for industrial applications.Thus, the development of highly ordered porous and low-cost GDEs to replace commercial carbon fiber paper is of urgent importance.
[33] For example, with fluent electrolyte transport paths based on vertically-aligned channels of the original crude wood material, Hu et al. proved the 3D-wood electrode ensures the feasibility in vanadium flow battery. [34]Similarly, Cui et al. [35] introduced wood slice as the template to fabricate ultrathick bulk LiCoO 2 cathode by utilizing their vertical microchannels as the highway for ultralong Li-ion transport pathway.[38] We can envision that wood materials with such natural structure would be ideal "breathable" candidates as GDEs, [39] but the relevant progress is rather limited, and lack of systematic investigation of identified structure characteristics on performance.
Herein, we demonstrate the feasibility of wood-based GDEs for aluminum-air batteries, with cobalt-nitrogen co-doped carbon nanotubes as ORR catalysts inside the channels of carbonized cross-cut wood slice and gradient wettability modification.Cherry-based GDE exhibited higher power density (267 mW cm −2 ) than the commercial carbon fiber paper-based electrode, which can be attributed to the superior gas transport property based on the aligned vertical pore structure and the elongated localized gas-liquid-solid three-phase contact lines inside the micro-channels.As further combined with deep-learning image recognition technique and gas permeability measurement, the characteristics of the pore structure of cherry and other three typical (ash, pine, and oak) wood-based GDEs were identified.It is revealed that the ratio of effective porosity to the average pore size is key to the GDE performance, as evidenced by the superior performance of the abundant porosity and effective small pore size (45 ± 15 μm) of cherry.We believe that this work demonstrates the feasibility of bio-derived wood-based materials for fabricating high-performance GDEs, and provides rational structural manufacturing and modification strategy for further optimization of GDEs.

Results and Discussion
The typical fabrication process of wood-based GDEs is schematically illustrated in Figure 1a.A piece of cross-cut 300μm-thick wood was first carbonized at 800 °C for 60 min (Figure S1, Supporting Information); the original 3D porous framework of wood with vertically aligned microchannels is well kept, as revealed by the top-and side-view scanning electron microscopy (SEM) images (Figure 1b,c).Co-and N-co-doped carbon nanotubes (CoN-CNTs) [40][41][42][43] were chosen as the model catalyst for ORR, and Co(OH) 2 nanosheets were grown on the carbonized wood slice as the catalyst for CNTs' growth (Figure S2a, Supporting Information), which was followed by the pyrolysis of melamine as the carbon and nitrogen source under 650 °C. [44,45]As characterized by SEM, the two sides' surface and the inner surface of microchannels are densely covered with curved CNTs (Figure 1d; Figure S2b, Supporting Information), and transmission electron microscopy (TEM) image indicates that Co nanoparticles are located at the top area of CNTs.High-resolution transmission electron microscopy (HRTEM, Figure S2c, Supporting Information) further displays the typical lattice spacing of 0.216 and 0.391 nm, corresponding to Co(100) and (120) of CNTs.The powder X-ray diffraction (XRD) pattern shows characteristic peaks at 26°and 42°, which can be indexed into (002) and (100) facets of CNT, and 44°, 51°, 75°, and 92°corresponding to Co(111), Co(200), Co(220) and Co(311) facets, suggesting the successful growth of CoNC-NTs (Figure 1e).The nitrogen content was measured as 13.73% from X-ray photoelectron spectroscopy (XPS), which is similar to the energy dispersive spectroscopy (EDS) result (12.91%).(Table S1 and Figure S3a,b, Supporting Information).The N peak can be divided into graphitic N, pyrrolic N, and pyridinic N (Figure S2d, Supporting Information), and cobalt element present in the form of +2 or +3 valence (Figure S4a,b, Supporting Information).
The intrinsic ORR activity of as-prepared Co-N-CNTs was evaluated with a rotating disc electrode in 0.1 m KOH (Figure S5a,b, Supporting Information), exhibiting the onset potential of ≈0.83 V and the half-wave potential of ≈0.74 V, which is comparable with reported results. [44]The electron transfer number was calculated via the Koutechy-Levich (KL) equation according to the limiting current density at different rotational speeds, which was found as ≈3.8 at different potentials (0.3 V, 0.35 V, 0.4 V, and 0.45 V) (Figure S5c, Supporting Information), indicating ideal ORR performance.
For the application of metal-air batteries, the wettability modification is helpful for mass transport and elongating the threephase contact boundaries inside the channels.The as-prepared cherry wood-based GDE was further modified with gradient wettability by putting on a shallow PTFE solution (10 μL) film instead of completely soaking and roasted at 350 °C for 0.5 h to get partially wetted from only one side via the capillary force.The water droplet contact angle of the PTFE penetration side is measured as 112°, and that of the other side is 55°, indicating the wettability difference between the two surfaces (Figure 2a), as aerophilic (AI) side and aerophobic (AO) side, as further evidenced by the apparent "white" surface of AI side induced by the strong light reflection from continual gas film coating (Figure 2a inset).Linear elemental mapping of fluorine along the cross-sectional direction (Figure 2b) verifies that the fluorine content exhibits a decreasing trend from AI side to AO side.After soaking with aqueous soluble dye (Pyronin) and drying (Figure 2c), the fluorescence signal mapping along the cross-section exhibits the opposite trend of fluorine element distribution, further verifying the successful modification of PTFE gradient.With such gradient wettability, the cherry-based GDE exhibits the onset potential at 0.78 V, and a rapid current density growth reaching a current density of 15 mA cm −2 at 0.5 V versus the reversible hydrogen electrode (RHE), superior over both of the aerophobic and the aerophilic electrodes (Figure 2d).The peak power density of the assembled aluminum-air battery (Figure 2e; Figure S6a,b, Supporting Information) based on the as-prepared cherry-based GDE can reach up to 267 mW cm −2 at the open-circuit voltage of 1.9 V, higher than that of 236 mW cm −2 based on commercial carbon fiber paper electrode, easily driving LED light and electric fan (Figure 2f; Figure S6c, Supporting Information).With a charge/discharge current density of 10 mA cm −2 , cycling performance is essentially stable for 45 000 s (12.5 h), demonstrating ideal application potential (Figure S7, Supporting Information).48] To further get insights into the structure influence on GDE performance, four typical wood pieces including cherry, oak, ash, and pine, as well as commercial carbon fiber paper were prepared as GDEs via the same fabrication procedure.As revealed from the SEM images (Figure 3a-e), the selected four wood materials generally have characteristic pore size distribution.While the pore size of commercial carbon fiber paper is randomly distributed owing to the random knit fabrication, namely, the irregular curved channels for gas transportation.In order to obtain precise statistical information of pore size distribution, a deeplearning image recognition technique was applied for the treatment of SEM images of carbonized wood materials (Figure 3f; Figure S8a-h, Supporting Information).Briefly, the original images were first sharpened with the homomorphic filtering algorithm, to make the edges of the apertures more informative and to allow more details to be seen.Second, small apertures were eliminated and breaks in the contour lines were filled by smoothing out parts of the aperture contours through closing operation, bridging narrower breaks and slender gullies.The entire images were then binarized to give a distinct black-and-white effect, with whitened around and black holes sorted as pores, thus highlighting the target contours.Finally, the statistical information was obtained according to the plotting scale after the rectangular out-cut treatment (see more details in Supporting Information).As revealed by the statistics of pore area distribution (Figure 3g) and porosity distribution (Figure 3h; Figure S9d, Supporting Information), cherry has the dominant macropores with a relative medium size of ≈45 ± 15 μm (Figure 3a inset); ash and oak have the dominant macropores larger than 100 μm, in detail, 150 ± 10 μm for ash, and 100 ± 10 μm or oak (Figure 3g inset).While for pine, the pore size is much smaller, with an average value of 15 ± 5 μm.The statistics of porosity can be ideally assumed as equal to the area fraction of pores at the wood slice surface, which can be used to estimate the droplet contact angles on the hydrophobic side of wood-based GDEs (Figure 3i; Figure S9a-c, Supporting Information) according to the Cassie-Baxter theory.
where f pore is the area fraction of pores at the wood slice surface, obtained from image recognition;  PTFE is the contact angle of a polymer in the form of a flat and smooth film(≈120°for a PTFE film);  pore is the contact angle of the pore, which is assumed as 180°in the Cassie-Baxter state.The calculated contact angles are highly consistent with the experimental results, confirming the accuracy of the obtained pore structure from deep-learning image recognition.Electrochemical impedance spectroscopy (EIS) measurements were performed with the GDEs placed on the surface of the electrolyte.[51] Also, the cherry-based GDE has the most rapid current density growth, reaching a current density of 15 mA cm −2 at 0.5 V versus the reversible hydrogen electrode (RHE), higher than the carbon fiber paper 4b).To further investigate the pore structure influence on the performance, we further evaluated the gas permeability of the woodbased GDEs with a homemade device (Figure 4c), by fixing the wood-based GDEs in the pipe and placing different weights to provide the driving pressures.All the wood-based GDEs show superior permeability than that of commercial carbon fiber paper at various pressures, verifying the ideal gas diffusion property of aligned vertical pore structures.While from the view of pore structure, the wood-based GDEs can be considered as slices with multiple vertical round channels, and such single-phase flow can be estimated by the semiempirical model via Kozeny-Carman (KC) equation, [52] relating to porosity, specific surface area, tortuosity, and a geometric factor.As derived from the solution for laminar flow in a round pipe, the Hagen-Poiseuille equation, and Darcy's law (see more details in Supporting Information), the permeability  can be expressed as: where r is the pore radius, A is the cross-sectional area,  is tortuosity (defined as l/L, l is the length of the pore channel and L is the length of the permeable wood slice).For the vertical pore structure of wood-based GDEs (Figure 4d),  is 1, and the ratio of A is the porosity ϕ.Thus, the permeability  is r 2 8 , namely, proportional to the product of porosity and pore size.Based on the statistics of porosity and pore size from image recognition, the order of estimated gas permeability of ash-, oak-, and cherry-based GDEs is in accord with the experimental results (Figure 4e); While the experimental permeability of pine is far behind the estimated value.Given that most pores of pine are smaller than 25 μm, the practical gas resistance would be much larger than that of large pores, and such small-sized channels in wood would be easily blocked during the treatment processes and/or naturally blocked owing to their internal wall structures of an organism (Figures S10 and S11, Supporting Information).The result indicates that a significant proportion of the porosity of pine is ineffective (as marked in dash circle in Figure 4e), and should be recalibrated, demonstrated as a shadow part in Figure 3g,h.
On the other hand, as the thicknesses of GDEs are the same, the reactive sites (S) of different wood-based GDEs can be roughly estimated as proportional to the total circumferential length of all pores, which is the product of the circumferential length of single pore (2r) and the pore number N. Also, a given porosity can be accounted as the ratio of the average single pore area (r 2 ) to the corresponding surrounding area nearby each pore (R 2 ), namely, ϕ = r 2 /R 2 (Figure 4f inset).Hence, the pore number N is inversely proportional to the average surrounding area (1/R 2 ), and we can get such product as "r/R 2 ", namely, ϕ/r (Figure 4f inset).
In light of the above results, we can understand that among the four typical wood-based GDEs, the pore size of natural wood lower than 25 μm would be tiny for gas transportation probably due to the high transport resistance and/or easily be blocked owing to the inner wall structure; over-big pore size of wood (>150 μm) would conversely reduce the porosity, namely, the effective active sites; small enough pore size (50 μm) and abundant porosity of cherry, namely, the high ratio of "ϕ/r" is the key to top the performance.Also, with the asymmetric gradient wettability modification, the localized gas-liquid-solid three-phase contact lines inside the micro-channels would be effectively elongated, which would be much longer than that of the conventional random knit carbon fibers, resulting in higher utilization of active sites.
As further considering the applications in other gas consumption reactions (e.g., CO 2 reduction reaction, fuel cells), humidity and wetting state should also be concerned with the pore structure of GDEs.For CO 2 reduction reaction devices, the GDEs are usually under the similar condition like metal-air batteries, with one side toward electrolyte and one side toward gas, the small sized pores would be over-immersed owing to the capillary effect, which could be promoted by wettability gradient modification or introducing compulsive flow; [53] For fuel cells, the target gas typically diffuse the GDE in a relatively dry environment, and the pores could be tuned ideally small enough.Hence, the rational design and deeper understanding of pore structured (like woodbased) materials are promising candidates in gas consumption reactions and would open the door to new possibilities.

Conclusion
In summary, we introduced bio-derived wood slices as the current collector to fabricate high-performance GDEs for aluminum-air batteries, with Co-N-CNTs as the ORR catalysts grown inside the channels and wettability gradient modification.The carbonized wood-based GDEs exhibit superior gas permeability than the commercial carbon fiber paper.For the application of aluminum-air batteries, the power density of cherry-based GDEs reached 267 mW cm −2 , higher than 236 mW cm −2 of the carbon fiber paper-based electrode.Comparing other three typical kinds of wood, ash, pine, and oak, it is verified that the performance is highly related to small enough pores and abundant porosity.While the pore size lower than 25 μm of natural wood would easily be blocked and induce high transfer resistance in practical applications.The overbig pore size would conversely re-duce the effective active sites owing to the limited porosity.This work exhibits the huge application potential of wood-based materials in gas consumption reaction devices, but also gives insights into the rational design of ordered structures of GDEs practical application scenarios.

Experimental Section
Chemicals: Urea was purchased from Macklin.Co(NO 3 ) 2 -6H 2 O was purchased from Xilong Scientific.Melamine was purchased from Henan Junhua Development Co., Ltd.Aluminum flakes (2 mm) were purchased from Tianjin Guangfu Fine Chemical Research Institute.Hydrophilic carbon paper (HP-030) was purchased from Toray.Commercial Pt/C was purchased from Suzhou Yilongsheng Energy Technology Co., Ltd.Deionized water with a resistivity ≥18 MΩ was used to prepare all aqueous solutions.All the reagents were of analytical grade and were used directly without further purification.
Preparation of Wood-Based GDEs: The wood slices were crosssectionally cut into 2 × 2 cm 2 squares with a thickness of 300 μm and carbonized in a tube furnace at 800 °C for 60 min with the fixation of nickel foam as a multilayer sandwich structure.Then, dense cobalt hydroxide nanosheet arrays were grown on the carbonized wood using a solvothermal method, with 40 mL methanol solution containing 0.29 g cobalt nitrate hexahydrate and 0.6 g urea, at 120 °C for 12 h.After cooling, rinsing, and drying, CNT catalysts were loaded via chemical vapor deposition.With 2 g of melamine as the carbon and nitrogen resource and cobalt hydroxide nanosheet arrays as pyrolysis catalyst, Co-N-CNTs were grown in a tube furnace at 600 °C with an increasing rate of 5 °C min −1 .Afterward, for the wettability gradient modification, the obtained GDEs were repeatedly sprayed with a small amount of 5% PTFE solution from one side after drying three times, and finally got roasted at 350 °C for 0.5 h.

Characterization:
The morphology and structure of the materials were examined using a Zeiss SUPRA 55 scanning electron microscope (SEM) and a Hitachi HT7700 transmission electron microscope (TEM), respectively.Thermogravimetric Analysis (TGA) was performed using Discovery TG instruments under the continuous gas flow of 10 mL min −1 of N 2 under a heating rate of 10 K min −1 .Differential Scanning Calorimetry (DSC) was performed using METTLER TOLEDO instruments under continuous gas flow of 50 mL min −1 of N 2 under a heating rate of 10 K min −1 .X-ray diffraction patterns were recorded on a Shimadzu XRD-6000 Ray Diffractometer.X-ray photoelectron spectroscopy (XPS) measurements were performed on an ESCALAB 250 from THERMO VG, USA.Contact angles were tested on a CA100D optical contact angle tester.Electrochemical properties were measured on a Shanghai C&H CHI 660.
Electrocatalytic Experiment: RDE tests were conducted at room temperature using a PARSTAT2273 rotating disc tester (diameter of working area 5 mm) with an oxygen-saturated 0.1 m KOH electrolyte.The catalyst was first scraped off the unmodified electrode, then 5 mg of catalyst was added to 0.49 mL of DMF, followed by 10 μL of Nafion solution (5 wt.%) and sonicated for 30 min to obtain the ink.A combination of a Pine modulated speed rotator and a CHI 660E was used for testing.Cyclic voltammetric curves (CV) were first measured at a scan rate of 50 mV s −1 and then the catalyst polarization curves (LSV) were taken at a scan rate of 5 mV s −1 and a rotational speed of 1600 rpm.
For the aluminum-air cell device assembly process, aluminum sheets of >99% purity were selected for the anodes and the prepared samples were used as air electrodes with a 1 cm spacing between the cathodes and anodes.A 6 mol L −1 potassium hydroxide solution was circulated through the device under a peristaltic pump to test the cell's performance and its application.

Figure 1 .
Figure 1.a) Schematic diagram of the fabrication process of cherry wood-based GDE.b) Top-and c) side-view SEM images of cherry wood-based GDE.d) SEM image of CNTs grown inside pores.e) X-ray diffractions of carbonized wood (CW) and CoNCNTs@carbonized wood (green patterns represent Co metal, black patterns represent C, and red patterns represent CNT).

Figure 2 .
Figure 2. a) Scheme of the GDE with gradient wettability modification, and the corresponding contact angles and underwater images.b) Linear mapping of fluorine elements in cherry wood fractures, inset: the image of SEM diagram of wood vertical channel slope.c) Confocal fluorescence microscopy image of gradient wetted cherry-based GDE.d) LSV curves of carbonized wood with different infiltration modifications.e) Comparison of cell performance between carbon fiber paper air and cherry wood electrodes.f) LED lighting application of aluminum-air battery based on wood-based GDE.g) Schematic diagram of the principle of vertical pore gradient wetting carbonized wood air electrode compared with conventional carbon paper air electrode.

Figure 3 .
Figure 3. a-d) Top-view SEM images of carbonized cherry, oak, ash, and pine, inset: high magnification SEM images of characteristic pore structures and camera images.e) Top-view SEM images of carbon fiber paper; inset: camera image of commercial TGH-H-060 carbon fiber paper.f) Original SEM image and computer algorithm processing of cherry wood slice.g) Statistics of the pore size distribution of the four different wood-based GDEs, inset: magnified pore size distribution of cherry, oak, and ash.h) Statistics of porosity distribution of the four different wood-based GDEs.i) Experimental and calculated contact angles of the four different wood-based GDEs.

Figure 4 .
Figure 4. a) Permeability test of different carbonized woods and carbon paper.b) Experimental and calculated permeability of different wood.c) Schematic diagram of simplified permeability model.d) Schematic diagram of floating electrode device.e) Bode plot of carbonized woods and carbon paper.f) Comparison of cell performance between carbon fiber paper air and cherry wood electrode.