Catalyst integration within the air electrode in secondary Zn-air batteries

The air electrode of a Zn-air battery facilitates the O2 reduction and evolution reactions during battery discharge and charge, respectively. These reactions are kinetically sluggish and appropriate catalysts are essential at the air electrode to increase battery efficiency. Precious metals are traditionally used, but increasingly attention has shifted towards non-precious metal catalysts to decrease the cost and increase the practicality of Zn-air batteries. However, loading of the catalyst onto the air electrode is equally as important as catalyst selection. Several methods can be used to deposit catalysts, each with their own advantages and disadvantages. Example methods include spray-coating, electrodeposition, and impregnation. These can be categorized as indirect, direct, and hybrid catalyst loading techniques, respectively. Direct and hybrid loading methods generally provide better depth of loading than indirect methods, which is an important consideration for the porous, air-breathing electrode of a Zn-air battery. Furthermore, direct methods are free from ancillary materials such as a binder, required by indirect and hybrid methods, which translates into better cycling stability. This review examines the various techniques for fabricating catalyst-enhanced air electrodes with an emphasis on their contributions to battery performance and durability. More durable Zn-air battery air electrodes directly translate to longer operational lifetimes for practical Zn-air batteries, which is an important consideration for the future implementation of electrochemical energy storage in energy systems and technologies. Generally, direct catalyst loading techniques, which integrate catalyst material directly onto the air electrode structure, provide superior cycling performance to indirect catalyst loading techniques, which distribute an ex-situ synthesized material onto the top layer of the air electrode. Hybrid catalyst loading techniques, which grow catalyst material directly onto nanostructured supports and then integrate them throughout the air electrode architecture, offer a compromise between direct and indirect methods.


Introduction
Strategies for achieving a carbon neutral energy landscape often involve renewable energy sources, such as wind and solar, which are inherently intermittent.As a means to capture excess renewable energy and release the stored energy on-demand, energy storage is perceived as the key to widespread adoption of renewable energy [1][2][3].Of the various energy storage technologies available, electrochemical batteries are favoured for their versatility, scalability, and ease of use [4].Currently, Li-ion batteries dominate the rechargeable (or secondary) battery market, replacing the aging technology of Pb-acid batteries [1,5].However, several key drawbacks of Li-ion batteries include high cost [3], limited material reserves, and safety concerns [6].As a possible future replacement for Li-ion batteries, Zn-air batteries (ZABs) boast far higher energy density (1086 Wh kg −1 vs. 296 Wh kg −1 for Li-ion) [7][8][9], lower cost and highly abundant materials (mineral reserves of Zn are nearly twenty-times greater than for Li) [10,11], and safe operation (non-flammable components, especially the electrolyte) [6].However, challenges remain in the commercialization of ZABs due to Zn rechargeability issues, atmospheric complications (humidity, CO 2 , etc.), and, most critically, poor air electrode kinetics [6].
ZABs, and more generally any metal-air battery, combine a metallic electrode with an air-breathing electrode to establish an electrochemical cell [12].An alkaline electrolyte is often chosen to reduce the overpotential for the air electrode reactions [12].Of the various metallic electrodes stable in an alkaline solution, Zn provides the largest electrochemical potential and thus highest cell voltage [13].During battery discharge, the Zn electrode undergoes oxidation in an alkaline environment according to equation (1) [7].At the air electrode, O 2 is reduced according to equation (2) [14].It is important to note that the O 2 reduction reaction (ORR, equation ( 2)) requires chemical species in three different phases: gaseous O 2 , liquid water, and solid electrons [12].Thus, ORR can only proceed at three-phase boundary regions where all three phases coexist.In practical ZABs, this three-phase boundary area lies within the gas-permeable membrane that makes up the air-breathing electrode [12].
O 2(g) + 2H 2 O (l) + 4e − ⇌ 4OH − (aq) The air electrode of a metal-air battery has three requirements to facilitate ORR: (1) Electronic conductivity for supplying electrons, (2) gas permittivity to enable O 2 diffusion into the cell, and (3) hydrophilicity to provide access to water molecules [6].However, the maximum concentration and mobility of O 2 species is much higher in a gaseous environment than a liquid one (e.g.dissolved O 2 in water) [7,15].Therefore, a careful balance of hydrophobicity is crafted by applying a gradient of hydrophobic agents, the most common choice being polytetrafluoroethylene (PTFE) [6].The outside air-facing surface is highly hydrophobic to enable gaseous O 2 flow into the electrode, while the interior electrolyte-facing surface is more hydrophilic to improve the interface between the air electrode and electrolyte and thereby reduce the interfacial resistance [12,16].This electrode structure, often called the gas diffusion layer (GDL) [14], is frequently made of carbon materials to satisfy electronic conductivity [6,7,17].Another improvement made to the GDL is the development of a so-called microporous layer, where the interior electrolyte-facing side of the GDL contains a high-surface area structure of carbon particles bound together [6,14], forming pores on the order of 20-50 nm [18].It should be noted that this pore structure size is actually classified as mesoporous in traditional chemistry [19], but the nomenclature of a microporous layer persists nonetheless.This high surface area construction improves the solid-electrolyte interface and additionally provides a support for catalyst loading [6,14,20].This is important since ORR is a notoriously sluggish electrochemical reaction and benefits greatly from electrochemical catalysts [7,14,15].
A Zn electrode features the most active metal that can be electrodeposited in an aqueous electrolyte [13], permitting reversal of the discharge reaction (reverse direction of equation ( 1)) and recharging of the ZAB.At the air electrode, ORR is reversed in what is known as the O 2 evolution reaction (OER, reverse direction of equation ( 2)), since O 2 is formed [14,21].While the Zn electrode has its fair share of complexities and rechargeability issues [6,22], the O 2 reduction and evolution reactions are sluggish and require significant activation overpotentials [15].Therefore, the investigation of ORR and OER catalysts is a key focus of ZAB research [14,23].It is possible to physically separate the discharge and charge processes of the air electrode [24], but this adds to the size of the battery and reduces the effective energy and power density [7,16].Thus, if both charge (OER) and discharge (ORR) occur at a single electrode, a bifunctional catalyst is required [14].The list of contenders for bifunctional ORR/OER catalysts is extensive, but generally lower cost and highly abundant materials are selected for practical ZAB application [14,23].Commonly studied catalysts that meet this criteria are transition metal oxides, in particular oxides of metals that lie in the 4th period of the Periodic table (e.g.Mn, Fe, Co, Ni) [23].When experimentally evaluating the electrochemical performance of a bifunctional catalyst, it is common practice to compare results with a Pt and RuO 2 mixture as a benchmark catalyst [6,7,12,14,[25][26][27].
The GDL microporous layer serves as a scaffolding for catalyst loading and, thus, catalysts are usually applied at the interior electrolyte-facing side of the GDL [14].The three-phase boundary area, where ORR can occur, is generally dictated by the hydrophobicity of the GDL and where liquid electrolyte penetration meets air flow [7].The recharge reaction, OER, only requires two chemical phases (liquid hydroxide ions and solid electrons) and can occur anywhere the electrolyte is in contact with the electronically conductive GDL material.It is ideal to have catalyst material distributed throughout the GDL such that ORR and OER can always be catalyzed and activation overpotentials minimized.As such, catalyst material at the interior face (i.e.where the electrolyte meets the electronic conductor) is favourable for OER, while catalyst material deeper within the GDL is desirable for catalyzing ORR, since this is where the three-phase boundary region exists [28,29].With the right hydrophobic treatment, ORR occurs primarily close to the microporous layer such that any occurrence of ORR is enhanced by the catalyst (figure 1(a)).However, extensive charge and discharge cycling of a ZAB can deteriorate the PTFE treatment of the GDL and can lead to a phenomenon known as flooding [6].During flooding, the electrolyte is able to penetrate deeper into the pores of GDL and pushes the three-phase boundary area (where liquid electrolyte meets gaseous O 2 ) to regions further away from the interior electrolyte facing side.This migration of the three-phase boundary area forces ORR to occur at areas deeper within the air electrode, where catalyst material may not have been deposited (figure 1(b)) [28,30].As a result, the activation overpotential of ORR increases and the overall efficiency of the ZAB is reduced.This phenomenon commonly appears as a loss in battery performance and efficiency over the course of many cycles.To reduce this effect, several different catalyst loading techniques can be used to apply catalyst material throughout the entire depth of the GDL, so that ORR catalyst is present at any region where the three-phase boundary area migrates [30].ZABs that have less performance loss over the course of many charge and discharge cycles are often described as being highly durable or exhibiting stable performance [21].
The deposition of catalysts onto/into the GDL can generally be classified as either indirect or direct [17].Indirect methods synthesize the catalyst material ex-situ (such as hydrothermal methods) and apply the catalyst through a traditional deposition technique, such as drop-casting or spray-coating [21].Direct methods create the catalyst material using the GDL as a substrate (i.e.in-situ synthesis) [31].This results in strong adhesion and generally better distribution of catalyst.However, a hybrid method of catalyst development has also been identified, where catalyst particles are synthesized on conductive nanomaterials and subsequently introduced into the GDL.While the method is technically indirect, since the GDL is not the catalyst substrate, the integration of the catalyst loaded nanomaterial into the GDL is effective and thorough for hybrid methods, leading to similar benefits as direct deposition methods such as good adhesion and excellent catalyst distribution throughout the thickness of the GDL.

Indirect methods
The classical method of depositing a catalyst material onto the air electrode substrate is via an ink suspension (figure 2) [21,26].The catalyst powder is dispersed in a solvent along with a conductive filler and binder agent [17,32].The conductive filler, often a carbon material [33][34][35][36], compensates for catalysts that are inherently nonconductive, such as metal oxides [14,23], to reduce the contact resistance between the catalyst surface and substrate material [37].The binder agent enables the formation of a homogenous suspension and also works to reduce the contact resistance between the catalyst particles and the conductive filler [38].Both PTFE and Nafion are used as binders [38,39]; the latter material is favored for polymer electrolyte membrane fuel cell electrodes [40,41].Typical solvents include water [42], glycerol [41], ethanol [33,43], isopropanol [20,36], or any mixture of these [44][45][46].Ultrasonication is a common step in synthesis to uniformly disperse the catalyst, binder, and filler and create a homogenous suspension [41,44,[47][48][49].The ink is then applied to the air electrode substrate either using drop-casting [35,46,49,50], where the suspension is deposited onto the surface drop-wise, or spray-coating [36,51], where the suspension is sprayed with an airbrush and painted onto the substrate [44,52].After coating, the solvent evaporates and a residue of catalyst, filler, and binder remains on the surface to act as the catalyst layer [20,40].Since the catalyst material is synthesized ex-situ and applied to the air electrode substrate afterwards, the ink-based method is considered as an indirect method of catalyst loading [17].Related to this is the paste method, which creates a more viscous slurry using the same components, which is then rolled or pressed onto the air electrode substrate [29,33,34,38,53,54].
A beneficial aspect of indirect methods for ZAB catalyst loading is the freedom to synthesize the catalyst material using a variety of methods.Since catalyst formation occurs independent of the final air electrode substrate [17], high temperatures [55] or aggressive solvents [42,45] can be utilized.Any number of synthesis methods are available, but common solution-based ones include sol-gel techniques [56], hydrothermal formation [53,54,57], and chemical reduction [34].As well, most researchers prepare benchmark noble metal catalyst air electrodes via an indirect ink and spray-coating method [58][59][60][61].
The additional materials required during indirect loading of catalysts (i.e.binder and filler) can negatively impact overall battery performance [6,16,58].Binders can block catalytic active sites [16,21,41] and degrade during ORR [6,16,32,38], while carbon-based filler materials increase the overall mass of the air electrode [16,32] and corrode during OER [6,14].For example, Wang et al prepared a bifunctional La 0.8 Sr 0.2 Co 0.4 Mn 0.6 O 3 perovskite catalyst using a sol-gel method [56].They subsequently used an indirect paste method to create a ZAB air electrode [62].Over the course of 100 cycles, the OER potential increased by 150 mV when cycled at 10 mA cm −2 in 6 M KOH.The authors attributed this performance degradation to the oxidation of the carbon support used during preparation of the catalyst slurry [56,62].More critically, however, indirect methods generally only deposit catalyst material on the top-most surface of the air electrode [30,63].This results in severe performance loss during extended cycling, when electrolyte flooding migrates the three-phase boundary past the catalyst layer [30].This is illustrated in a report by Prabu et al who deposited a graphene-supported CoMn 2 O 4 (CMO) catalyst on GDL via an airbrush (figure 3).The efficiency of the assembled ZAB at the beginning of the cycling test is shown in figure 3(a) and is calculated as the discharge potential divided by the charge potential.For this example, the efficiency is 65% when cycled at 20 mA cm −2 in 6 M KOH (figure 3(a)).As the cycling test continues, with alternating periods of charge and discharge, the charge and discharge potentials deviate from the standard potential by greater and greater amounts.This is understood as the overpotential of the battery cell and is also denoted as the C-D potential gap in figure 3(b).The battery efficiency calculated at the 200th cycle is only 47% (figure 3(b)) [51].This represents a 28% loss in efficiency over the period of 200 cycles, or 14% per 100 cycles.The indirect catalyst  loading technique experiences a large degradation in battery performance during cycling, as the three-phase boundary migrates past the spray-coated surface layer and into the uncatalyzed GDL substrate.

Direct methods
In contrast to indirect methods, direct loading of ZAB catalysts involves synthesizing an electrocatalyst directly onto or even within the final air electrode substrate.This can be accomplished by an array of techniques including electrodeposition, electroless deposition, chemical vapour deposition (CVD), or physical vapour deposition (PVD) [17].The air electrode substrate is often a carbon-based GDL as described previously; however, other candidate substrates include metal foams or meshes using alkaline-stable metals such as Ni, Ti, or stainless steel [6,21,24,64].A hydrophobic treatment or barrier layer is applied after catalyst synthesis to provide the necessary wet-proofing of the air electrode [21,65].An obvious advantage of direct catalyst loading is the avoidance of additional materials such as binders or conductive fillers.These materials, susceptible to degradation during battery operation, are excluded so that the lifetime of the air electrode is extended [58].Another advantage of direct catalyst synthesis is the improved interface between the current collector and catalyst material [32].Electron flow from catalyst particles to the current collector is not impeded by any binders or secondary materials [64,66], reducing the resistance of the air electrode [6,58].As well, direct catalyst loading methods are generally faster and can simplify electrode design [58,66], which is advantageous for commercializing and scaling up ZABs [6,17,32,64].More importantly, however, direct deposit methods can deposit catalyst material throughout the air electrode structure, as opposed to only surface-level deposition, increasing the catalytically-active surface area and improving the stability of the air electrode against flooding [6,32,64,66].The main drawback of direct catalyst synthesis on the air electrode substrate is the sensitivity of the air electrode material.Aggressive deposition techniques, such as those involving high temperatures or oxidizing chemicals, can damage carbon-based GDLs and hydrophobic agents (e.g.PTFE) [17,67].The use of non-carbon GDLs and ex-situ hydrophobic treatments is a potential workaround to these limitations.
Depositing ZAB catalysts directly on the air electrode substrate can be achieved by either electrodeposition or electroless deposition.Used extensively in industry for applying coatings [68,69], these techniques operate in the liquid phase by converting dissolved chemical species into a solid material.In electrodeposition, the voltage difference and current flow between electrodes are manipulated, resulting in deposition of material on the working electrode [17].In electroless deposition, reducing agents in solution cause spontaneous deposition of material on the working electrode [69].In either case, if GDL is used as the working electrode, ZAB catalysts can be directly coated onto the GDL.Electrically conductive substrates are required for electrodeposition, but not for electroless deposition [17,69].GDLs are usually conductive to act as a current collector for the air reactions and are thus suitable substrates for either process.Xiong and Ivey published work on a Co-Fe catalyst for OER, electrodeposited directly on carbon GDL (figures 4(a) and (b)) [37].Co-based catalysts benefit greatly from Fe doping, but synthesis procedures are often complex and involve additives such as carbon black or PTFE.In their work, Xiong and Ivey achieved ZAB cycling results comparable to a Pt-C commercial catalyst using a simple one-step electrodeposition process [37].Improving upon their work, an ORR-active catalyst (MnO x ) was also directly electrodeposited onto GDL (figures 4(c) and (d)), in addition to the OER Co-Fe catalyst [70].By distributing the MnO x catalyst on the GDL first, followed by the Co-Fe catalyst (figures 4(e) and (f)), regions of ORR activity are catalysed by MnO x while regions of OER activity benefit from the Co-Fe catalyst, similar to the depiction in figure 1.This double-layered catalyst air electrode provided superior ORR and OER performance than either catalyst layer alone based on cyclic voltammetry testing.Furthermore, ZAB rate testing in 6 M KOH revealed that the double-layered catalyst provides superior bifunctional efficiencies than a commercial Pt-C catalyst, with 62% and 60% bifunctional efficiencies at 10 mA cm −2 for the electrodeposited catalyst and Pt-C catalyst, respectively [70].When cycled at 5 mA cm −2 , a ZAB constructed with the MnO x /Co-Fe catalyst had similar bifunctional efficiencies to one with a commercial Pt-C catalyst.The average discharge potential for the MnO x /Co-Fe ZAB was 1.18 V, compared with 1.21 V for the Pt-C ZAB, while the average charge potential was 1.98 V for the MnO x /Co-Fe ZAB compared with 2.03 V for the Pt-C ZAB.After 13 h of cycling, the ZAB with MnO x /Co-Fe performed better than the ZAB with Pt-C in discharge-charge polarization testing [70].Thus, the electrodeposition technique for the MnO x /Co-Fe catalyst resulted in a more stable ZAB than the indirect spray-coated Pt-C electrode.
A double-layer electrodeposition process was also explored by Kim et al where they used a gel-mediated electrodeposition process to deposit ORR-active MnO 2 and OER-active Co 3 O 4 directly onto Ni foam.A ZAB constructed with this catalyst showed only 0.05 V of degradation during either charge or discharge at 1 mA cm −2 over 400 h in 6 M KOH, maintaining approximately 60% bifunctional efficiency [71].The direct electrodeposition process provided a strong connection between the catalyst and Ni foam substrate, resulting in stable cycling performance.Some researchers have applied a bifunctional catalyst directly onto a carbon substrate via a single-step electrodeposition process [59,72,73], while others have used electrodeposition to add a second catalyst to a previously synthesized catalyst material [74,75].Electroless deposition was used by Karajagi et al to create a Ni interlayer on carbon air electrodes, protecting the electrodes from carbon corrosion and also providing OER activity [76].In a different study, Lee et al used an electroless process to grow Co 3 O 4 nanowire arrays directly on a stainless steel mesh.When applied as an air electrode, the directly grown catalyst was stable for 600 h, exhibiting 97% charge and 94% discharge retention over nearly one month of cycling in 6 M KOH at 18 mA cm −2 .Comparatively, a spray-coated Pt-C electrode failed after only 25 h, when the charging potential exceeded 3.0 V [58].The excellent longevity of the electrodeposited catalyst exemplifies the benefit of a direct catalyst loading technique, while the poor cycling performance of the spray-coated Pt-C electrode highlights the inferiority of indirect catalyst loading techniques.
Since the GDL must be gas permeable for O 2 exchange between the cell and exterior environment, a gas-based deposition technique would be suitable for loading catalyst material throughout the GDL structure.CVD is a method of depositing thin film coatings on a substrate by employing gaseous reactants.Both surface-based and gas-based chemical reactions occur in CVD which lead to film growth as precursors are supplied to the surface [77].A few examples of catalysts synthesized via CVD are found in the literature, but rather than use CVD to coat the GDL substrate, CVD is used as a method of synthesis for ZAB catalysts.Sometimes the catalysts are developed directly with the air electrode material [78,79] and other times they are deposited indirectly after ex-situ synthesis [80][81][82][83].A specialized form of CVD known as atomic layer deposition (ALD) isolates only surface-based chemical reactions to grow coatings, eliminating gas-phase reactions or thermal decomposition of precursors.As a result, the conformality or step-coverage of coatings via ALD is superior to traditional CVD [77].Thus, ALD is ideally suited for high-aspect ratio substrates, such as the high surface area microporous layer of GDL.Direct coating of the GDL substrate with an ALD film was studied by Clark et al in their publication of a MnO x catalyst active towards ORR.They found that a forming gas (5% H 2 in N 2 ) plasma pre-treatment was required to maintain saturating growth from the water-based MnO x ALD process.Without the plasma pre-treatment, MnO x ALD coatings on GDL were found to agglomerate carbon particles and reduce the porosity and surface area of the air electrode.The optimized ALD process instead uniformly coated the air electrode, maintaining the porosity and also enabling deeper penetration of catalyst material into the air electrode depth [84].When assembled into a ZAB, the MnO x ALD coated air electrode displayed superior polarization behaviour compared with a Pt-Ru-C benchmark catalyst at current densities above 200 mA cm −2 .High current densities consume O 2 quickly and battery performance relies on O 2 availability.The benchmark Pt-Ru-C air electrode was prepared via spray-coating and does not provide as much O 2 access to catalyst material as the ALD coated electrode, with deep catalyst loading into the electrode porosity.Thus, the ALD MnO x air electrode provided a superior peak power density of 170 mW cm −2 compared with 158 mW cm −2 for Pt-Ru-C [30].To further improve the performance of the air electrode, CoO x was added to the ALD process.Overall, the ALD coated electrode, with direct catalyst loading, was resilient to flooding, maintaining a stable discharge potential of ∼1.25 V at 10 mA cm −2 for 20 h in 6 M KOH and only decreasing to ∼1.17 V after 50 h of cycling.In comparison, a spray-coated Pt-Ru-C benchmark catalyst, with poor distribution of catalyst material from indirect loading, suffered greatly from flooding during cycling.The initial discharge potential of ∼1.25 V quickly dropped to 1.1 V after only 20 h at 10 mA cm −2 [30].
Direct deposition of ZAB catalysts using ALD was also studied by Labbe et al by directly depositing an OER-active Fe 2 O 3 catalyst on GDL.However, their ALD process utilized an O plasma which etched the carbon-based GDL substrate.To overcome this, a protective sublayer of ALD MnO x was deposited prior to FeO x ALD [85].The MnO x sublayer also enhanced the growth characteristics of the FeO x process and it was found that the FeOx ALD process experienced multiple stages of island growth, a phenomenon not typical in ALD [67].By mixing the ORR-active MnO x ALD layer developed by Clark et al [84] with the OER-active FeO x ALD layer, a bifunctional catalyst was developed via direct ALD.Optimization of the mixed MnO x -FeO x ALD process yielded a 30:10 mixture of MnO x cycles to FeO x cycles.Electron diffraction analysis in the transmission electron microscope (TEM) indicated that the mixed oxide is a Mn spinel structure with Fe substitution ((Mn,Fe) 3 O 4 ).Energy dispersive x-ray (EDX) analysis in the scanning TEM (STEM) (figures 5(a)-(d)) revealed a thin, uniform distribution of the ALD coating on the carbon particles of the GDL.Cycling results of a ZAB with the (Mn,Fe) 3 O 4 catalyst indicate a long lifetime for the directly grown catalyst, with stable cycling results even after 600 h (1540 cycles) of 10 mA cm −2 cycling in 6 M KOH (figure 5(e)) [86].A Pt-Ru-C benchmark catalyst, on the other hand, deposited by indirect spray coating experienced performance loss and reached similar charge and discharge potentials as an uncoated electrode after 300 h of cycling.
Other ALD researchers preloaded a ZAB air electrode with carbon nanotubes (CNT) to provide a high surface area scaffolding.Both Co 9 S 8 and NiS x were deposited in this way [87,88], yielding stable bifunctional performance in 6 M KOH and, in the case of Co 9 S 8 , very competitive peak power and efficiency values of 197.6 mW cm −2 and 62.5% at 10 mA cm −2 , respectively [87].While ALD is classified as a direct catalyst loading method, some researchers have employed ALD more as an indirect catalyst synthesis method, where the catalyst is grown via ALD ex-situ and deposited on the air electrode by an indirect method, such as drop-casting [89][90][91][92].One reason to use ALD as an indirect method of synthesis is because of aggressive O-based reactants employed in ALD.Carbon based organometallic ALD precursors, chosen because of their volatility and lack of contaminating Cl or N species [93,94], often require the use of ozone or O plasma reactants to combust the carbon ligands of the precursor molecule and avoid carbon inclusion in the growing ALD film [93][94][95][96].While effective for this purpose, these highly reactive O-based reactants can also oxidize carbon substrates, such as GDL, during deposition [30,67].Likewise, the PTFE in GDL can be damaged by ozone or O plasma reactants [30,97,98].Both of these effects reduce the efficacy of the air electrode during ZAB cycling [30,67].
In contrast to CVD, which involves chemical reactions to directly grow a film, PVD involves the physical redistribution of atoms from a target source onto a substrate.This vacuum deposition technique operates by vaporizing the target atoms through either heating (thermal evaporation), high energy lasers (pulse laser deposition, PLD), or a beam of high energy electrons.Alternatively, target atoms can be removed via sputtering, where high energy ions impart kinetic energy to the target atoms [99].Often metallic or alloy targets are used in PVD and the application of this deposition technique towards ZAB catalysts reflects this.Several papers by Chen et al explored the direct deposition of an Ag-Cu nanoalloy onto Ni foam air electrodes using PLD [100,101].They report ZAB performance values comparable to a Pt-C benchmark catalyst but with better stability, maintaining stable battery performance over 1200 cycles of charge and discharge in 6 M KOH, while noticeable deterioration of the Pt-C sample occurred after only 4 cycles [102].Pham et al also employed PLD, depositing a Ni 3 Pt alloy directly onto Ni foam for the air electrode of a ZAB which ran for over 450 cycles [103].
The terms 'free standing,' [104,105] 'self supporting,' [105] or 'binder free' [32,74,103] are frequently used in the literature to describe a direct deposition catalyst, where the substrate material is often synthesized in-situ during catalyst development [106,107].Metal-organic frameworks (MOFs), which are porous crystalline structures comprised of metal nodes and organic linkers, are sometimes employed in these studies [108,109].They provide high surface area and catalytic activity from both functional groups on the organic linkers and the uncoordinated metal centers.In particular, zeolitic imidazolate frameworks are frequently employed for their N functional groups on the organic components.Furthermore, metal or metal oxide nanoparticles can be attached within or on the surface of the framework structure to provide additional catalytic activity [108].Synthesis routes for MOFs are generally hydrothermal/solvothermal [108], with electroless deposition reagents added for anchoring metal oxide nanoparticles [104,105,109,110].

Hybrid methods
As the name suggests, hybrid catalyst loading techniques are a combination of both indirect and direct catalyst loading techniques.Hybrid methods directly grow catalyst material onto nanostructured supports (e.g.CNTs, carbon fibers (CF)) and then distribute the catalyst coated nanostructure into the air electrode.The ex-situ distribution into the air electrode provides the ability to manufacture the catalyst loaded nanostructures in a variety of environments and conditions, regardless of air electrode substrate material.When coupled with an effective distribution technique, competitive performance to direct deposition methods can be achieved because the catalyst material is integrated into the bulk of the air electrode and not surface level only.The nanostructured support is typically conductive (e.g.carbon-based) and thus conductive additives can be omitted from the electrode integration process [61].In addition, the direct growth of catalyst material onto the nanostructured support improves the interfacial contact between the catalyst and electronic conductor similar to direct catalyst loading methods.The main downside of hybrid catalyst loading is the continued necessity of polymeric binders, which can degrade during battery cycling, as occurs with indirect catalyst loading techniques [6,16,32,38].
One hybrid method, reported by Aasen et al begins by growing metal oxide nanoparticles on N-CNTs via a simple ultrasonication method.The suspension of nanoparticle coated N-CNTs is then passed through a porous carbon GDL via vacuum filtration, essentially using GDL as filter paper (figure 6(a)).By doing so, the nanoparticle coated N-CNTs are distributed deep into the bulk of the GDL, up to 100 µm into the microporous layer [61].Additionally, the GDL samples are presoaked in the catalyst suspension, increasing the loading of material within the GDL (figure 6(b)) [61].The researchers found that defect sites, created by doping CNTs with N, are a requirement for anchoring metal oxide nanoparticles onto the nanostructured support (figure 6(c)) [61].This hybrid deposition method provides a large catalyst surface area, improving overall catalytic performance, and maintains the three-phase boundary during extended battery cycling, leading to highly stable cycling performance [111].For example, an ORR-active Mn 3 O 4 catalyst on N-CNTs was cycled in a trielectrode configuration in 6 M KOH, where OER and ORR occur on independent electrodes, and showed only a 30 mV decrease in discharge potentials over 200 cycles at 20 mA cm −2 .This hybrid catalyst loading technique was further explored by Aasen et al in their investigation of a bifunctional (Co,Fe) 3 O 4 catalyst decorated onto N-CNTs (figures 6(d)-(g)).Bifunctional cycling showed negligible discharge loss and less than 0.1 V of charge degradation after 500 cycles, maintaining ∼58% efficiency at 10 mA cm −2 in 6 M KOH (figure 6(h)) [111].In addition to its high durability, the simplicity and low cost of this hybrid preparation method facilitates quick assessment of a wide variety of catalyst chemistries.Two additional reports by Aasen et al and one by McDougall et al explored metal oxide nanoparticle chemistries of NiMnO x , NiFeO x , NiCoFeO x , NiMnFeO x , NiMnCoO x, MnCoFeO x , ZnCoO x , ZnMnO x , ZnMnCoO x , ZnCoFeO x , and ZnNiMnCoO x [25,60,112].In all reports, the bifunctional cycling stability of the hybrid synthesized catalyst was superior to a spray-coated Pt-Ru-C comparison.
The above soaking and filtering impregnation technique was also extended to nanoparticle-decorated hollow carbon structures in several reports by He et al.Instead of using commercially obtained N-CNTs as nanostructured supports, He et al synthesized hollow mesoporous carbon spheres (HMCs) and hollow carbon cubes (HCCs, figures 7(a) and (b)) [113][114][115].Reactive chemicals (such as HF) were used in combination with high temperatures (as high as 915 • C).These reaction conditions are generally not compatible with direct catalyst deposition methods on GDL.Yet, the hybrid impregnation technique ensured good distribution of the catalyst loaded nanostructures throughout the GDL and resulted in stable cycling performance.For a Mn 3 O 4 catalyst on HMCs, trielectrode ZAB cycling at 20 mA cm −2 in 6 M KOH revealed a 3.3% loss in discharge potential (1.21 V to 1.17 V) after 235 cycles (figure 7(c)).A spray-coated Pt-Ru-C comparison, however, experienced 4.2% loss (1.19 V to 1.14 V) during the same cycling conditions (figure 7(d)) [113].For a Co 3 O 4 -decorated HMC catalyst, bifunctional ZAB cycling at 10 mA cm −2 revealed an initial efficiency of 63% with a final efficiency of 57% after 200 cycles (figure 7(e)).A spray-coated Pt-Ru-C comparison displayed initial and final efficiencies of 62% and 50%, respectively, after only 100 cycles at 10 mA cm −2 (figure 7(f)) [114].Lastly, a CoNi nanoparticle decorated HCC catalyst, impregnated into the GDL via the soaking and filtering hybrid loading technique, displayed an initial bifunctional efficiency of 59% during 10 mA cm −2 ZAB cycling in 6 M KOH.This was reduced to 55% after 90 h of cycling, yielding an overall 3.4% drop in efficiency (figure 7(g)).A spray-coated Pt-Ru-C electrode, on the other hand, had a 14.5% drop in efficiency (from 56% to 42%) after only 60 h of cycling (figure 7(h)) [115].These studies all reveal the ability of hybrid loading to provide the best of both worlds between indirect and direct catalyst loading techniques.Synthesis conditions not directly compatible with GDL material can be employed ex-situ to develop state-of-the-art catalysts, while the hybrid loading method is effective in delivering the catalyst material deep into the GDL to provide stable long term cycling performance.
Another hybrid method was recently explored by Abedi et al wherein metal oxide nanoparticles were grown on activated CFs via a sonication method (figures 8(a)-(d)) [116], which was similar to the processes reported by Aasen et al [61] or Li et al [117].No sonication in the synthesis procedure produced a discontinuous coating of catalyst material on the CF substrates, while too much sonication resulted in partial delamination of the coating.The optimized sonication time and Mn:Co salt ratio created a uniform, thick coating of MnCo 2 O 4 material on the CF supports.It was proposed that the Mn species exist exclusively in the 2+ oxidization state, while Co species were only in the 3+ state.This combination resulted in high bifunctional activity, outperforming the Pt-Ru-C catalyst [116].For this particular hybrid catalyst loading technique, Abedi et al transformed the catalyst loaded CFs into a paste with a small amount of carbon black filler and PTFE binder.It should be emphasized the nanoparticle coated CFs are already conductive, so that the carbon black filler is largely for mechanical purposes as opposed to a conductive aid.The paste was then rolled onto a hydrophobic carbon-paper backing layer to produce a ZAB air electrode with catalyst material distributed throughout the thickness of the GDL [116].The homemade GDL showed good bifunctional Reprinted with permission from [115].Copyright (2022) American Chemical Society.activity and excellent cycling performance, far superior to a spray-coated Pt-Ru-C benchmark [116].After a minor efficiency loss during the first few cycles, the hybrid paste MnCo 2 O 4 /CF exhibited almost identical charge and discharge potentials throughout 200 cycles at 10 mA cm −2 in 6 M KOH (figure 8(e)), while a Pt-Ru-C benchmark, prepared via spray-coating, experienced severe performance loss and an efficiency reduction from 57% to only 41% after 200 cycles at 10 mA cm −2 (figure 8(f)) [116].This GDL was even compatible with a gel electrolyte (polyacrylic acid with KOH) to yield a solid state ZAB, which displayed stable cycling behaviour for 130 cycles at 10 mA cm −2 , after which some degradation in performance is observed (figure 8(g)).A similar solid state ZAB with a spray-coated Pt-Ru-C catalyst on the air electrode, however, was not stable during cycling at 10 mA cm −2 , where degradation began after only 20 cycles and eventual failure occurred at 110 cycles (figure 8(h)) [118].The use of a gel electrolyte was also found to increase the power density of the ZAB, with the gel MnCo 2 O 4 cell exhibiting a peak power density of 240 mW cm −2 compared with only 127 mW cm −2 for the aqueous version, and only 165 mW cm −2 for the solid state Pt-Ru-C cell.As an additional benefit, the gel electrolyte was tolerant to low temperatures, enabling stable cycling performance at temperatures as low as −25 • C (figure 8(i)).In this case, after 200 cycles of 2 mA cm −2 bifunctional cycling, the starting and final efficiencies were 53% and 49%, respectively.At −45 • C (figure 8(j)), the ZAB was less stable and less efficient, but was able to withstand 200 cycles at 2 mA cm −2 without reaching the charge and discharge cutoff voltages of 3.0 V and 0.5 V, respectively [119].Similar hybrid paste methods have been employed by other researchers, who developed their catalyst loaded nanostructures via microwave [43], hydrothermal [120], solvothermal [27,121], and sol-gel [122] methods.The hybrid paste methods differ from traditional indirect paste methods because catalyst particles are directly anchored onto nanostructured (often conductive) supports, which are then transformed into the microporous layer of the GDL.Indirect paste methods, which disperse the catalyst (often non-conductive) in a mixture of a conductive filler and an insulating binder [33,34,53], do not possess the synergy between the catalyst and nanostructured support and may only coat the outermost surface layers of the microporous layer [53].

Battery performance comparison
A comparison of reported cycling performances for various air electrode catalysts explored in the literature is provided in table 1. Catalyst loading onto the air electrode is categorized as either indirect, direct, or hybrid, based on the methodology reported in the article.The testing conditions for ZAB cycling are not standardized and, as such, various different current densities and cycle lengths are employed [6].For example, Lee et al cycled their ZAB for 100 cycles at 6 h per cycle, yielding an overall cycling time of 600 h [58].Clark et al also cycled their ZAB for 100 cycles, but at 0.5 h per cycle, totalling only 200 h [30].Another variable among the studies, in addition to the deposition method and cycle parameters, is the chemistry of the catalyst employed.Some catalysts are more electrocatalytically active than others and provide higher bifunctional efficiency.Therefore, table 1 features a calculation of the efficiency loss per 100 cycles.This normalizes the data for each catalyst such that the stability of the air electrode is emphasized.For indirect deposition methods, the efficiency loss per 100 cycles is on the order of 5%-10%, representing poor stability where the efficiency of the battery decreases to ∼50% of the initial value after only 500 cycles.Direct deposition methods, on the other hand, have efficiency loss per 100 cycles values on the order of 0.5% to 5%.This represents much higher ZAB stability, where the efficiency value drops to ∼90% of the original value after 1000 or more cycles.For hybrid deposition methods, the efficiency loss per 100 cycles is on the order of 1% to 5%, which is on par with direct deposition methods.Thus, in terms of cycling stability, hybrid catalyst loading techniques are competitive with direct loading methods, both of which are far superior to indirect loading methods.This can also be seen in figure 9, which plots the efficiency retention against the total number of cycles tested.In this case, a higher value is desirable and represents an air electrode which can maintain its original bifunctional efficiency.The indirect air electrodes (shown as empty circles) populate the lower-left corner, which corresponds to low efficiency retention at low cycle numbers.The direct air electrodes (shown as black squares) are primarily found in the top portion of the graph, representing highly stable ZABs.The hybrid catalyst loaded air electrodes (shown as triangles) are also in the upper section of the plot, but are also at the far right edges, demonstrating long battery lifetimes.
The benefits afforded by directly loading catalyst material onto the air electrode, as opposed to indirect loading methods, are best illustrated by comparative studies.In one such study, Lee et al scraped off their directly grown Co 3 O 4 nanowires on Ni foam, prepared an ink with the catalyst particles, isopropyl alcohol, and Nafion, and spray-coated the mixture onto carbon GDL.For both pulse cycling (with 5 min charge-discharge periods) and extended cycling (with 3 h charge-discharge periods), the spray-coated catalyst exhibited severe degradation in cycling performance while, at the same time, the directly-grown catalyst was extremely stable, retaining over 90% of the initial performance after 600 h of extended cycling at 18 mA cm −2 in 6 M KOH [58].In another comparative study, Sumboja et al grew MnO x directly on GDL via electroless deposition.They also prepared MnO x powder using the same electroless technique and applied it to GDL ex-situ by drop-casting a Nafion and ethanol ink [123].Again, both pulse cycling and extended cycling (15 mA cm −2 discharge and 7.5 mA cm −2 charge, both in 6 M KOH) showed that the directly loaded MnO x was far superior in terms of stability than the indirect MnO x counterpart.In particular, the directly loaded MnO x catalyst had no more than a 5% change in charging or discharging potential after 350 pulse cycles, while the drop-casted MnO x experienced a 15% reduction in discharge potential after only 170 pulse cycles, with the charging potential reaching the 2.5 V cut-off potential within 110 pulse cycles.In the extended cycling, the directly deposited MnO x electrode showed less than 7% change in the charge and discharge potentials over 15 extended cycles, while the indirect MnO x electrode reached the charging cut-off value before the end of the first cycle [123].It should be emphasized that the above two comparative studies carefully controlled the mass loading between the direct and indirect methodologies to ensure fairness.In another comparative study, Meng et al utilized electrodeposition, calcination, and an immersion method to synthesize a bifunctional Co 4 N decorated CF network catalyst directly on carbon cloth.The catalyst coated carbon cloth was shredded, formed into a paste with Nafion in alcohol, and applied to a separate carbon cloth substrate to represent an indirect deposition catalyst.The cycling behaviour of the paste method was remarkably worse than the case where the directly synthesized carbon cloth electrode was used.After 83 h of cycling at 10 mA cm −2 in 6 M KOH, the indirect paste electrode provided only 33% bifunctional efficiency, while the direct catalyst electrode retained a stable ∼60% efficiency for 136 h at 10 mA cm −2 [104].In essence, these studies reveal the benefit of directly deposited catalysts compared with chemically equivalent indirectly loaded catalysts.A key feature of direct deposition methods that translates into stable cycling performance is the integration of catalyst material deeper within the air electrode structure compared with only surface bound catalyst particles deposited by indirect methods.When flooding occurs during prolonged cycling, the architecture of catalyst loading enables the continued catalysis of the three-phase boundary area and an overall stable cycling performance [30].This concept guided the work of Yu et al in their investigation of asymmetric air electrodes for ZABs [63].They employed a NiFe layered double hydroxide (LDH) electrocatalyst directly grown on a carbon paper substrate via a liquid phase chemical process.The  asymmetric electrode was developed by growing NiFe LDH inward from the hydrophilic face of the carbon paper, achieving a 3-dimensional (3D) interface of catalyst material (the third dimension being the thickness of the electrode; figure 10(b)).For comparison, a conventional air electrode (with a 2-dimensional (2D) interface) was prepared via drop-casting of NiFe LDH particles (figure 10(a)).Polarization tests in a ZAB with a 6 M KOH electrolyte revealed that the asymmetric electrode had more active sites and better mass transfer than the conventional electrode, with max power densities of 93.9 and 42.1 mW cm −2 for the asymmetric and conventional electrodes, respectively [63].ZAB cycling results reinforced the importance of catalyst architecture in the air electrode in achieving long-term stability.For the conventional drop-casted air electrode, the charge and discharge potentials were only stable for 650, 200, 80, and 25 cycles when cycled at 5, 10, 25, and 50 mA cm −2 , respectively.The asymmetric electrode, on the other hand, exhibited stable charge and discharge potentials for more than 2000, 600, 300, and 100 cycles, respectively, at the same current densities.The indirect catalyst loading method of the conventional electrode resulted in performance loss due to exfoliation and destruction of the electrocatalyst, while the direct catalyst loading process of the asymmetric electrode enabled high current and long cycle life in ZABs [63].
Hybrid catalyst loading techniques also develop these 3D catalyst architectures and provide similar benefits to the direct catalyst loading methods.For example, Aasen et al compared the ZAB rate performance of indirect spray-coating versus their hybrid soaking and filtering process for Mn 3 O 4 coated N-CNTs.At all current densities investigated (2-10 mA cm −2 ), the soaked and filtered air electrode showed a 0.1 V lower ORR overpotential than the spray-coated comparison [61].It was revealed that their soaked and filtered electrode had abundant catalyst material up to 35 µm away from the surface of the GDL substrate, illustrating a 3D catalyst architecture using a hybrid deposition technique.Thus, hybrid catalyst loading is competitive with direct loading and both produce a tailored electrode architecture that is resistant to flooding and improves catalyst access in the air electrode [61,63,111].
Using direct or hybrid methods for catalyst loading at the air electrode of a ZAB is far superior to indirect loading methods.Not only is catalyst material distributed more uniformly and deeper within the electrode substrate, enhancing long-term cycling behaviour, but the improved interface between the catalyst and current collector affords reduced Ohmic losses during battery operation [32].However, the choice between direct or hybrid loading should be evaluated based on the available synthesis method (or methods) for a particular catalyst and the compatibility of the substrate material with that synthesis method.For example, Co-Fe oxide particles were anchored onto N-CNTs and integrated into a carbon-based GDL using the hybrid soaking and impregnation technique.This technique employed mild solvents and low temperatures which did not damage the carbon material or PTFE treatment of the air electrode.As a result, a ZAB using the Co-Fe oxide air electrode in 6 M KOH was able to successfully cycle for over 250 h at 10 mA cm −2 [111].It is also feasible that an ALD process could deposit a Co-Fe oxide catalyst directly onto a GDL substrate, possibly preloaded with N-CNTs to increase the area for deposition [87].However, the ALD process would require the use of an O-plasma reactant for both FeO x and CoO x subcycles [30,85].The use of an O-plasma reactant, without any buffer layer integrated into the ALD process, has been demonstrated to be very damaging to a carbon substrate [67].Thus, the direct loading of a Co-Fe oxide catalyst onto carbon GDL using an O-plasma ALD process would result in a poor performing air electrode.In this case, the hybrid soaking and impregnation technique is a better choice than a direct ALD process.Electrodeposition can also be used to directly deposit a Co-Fe catalyst (with a surface layer of Co-Fe oxide), which was shown to be compatible with a carbon GDL.A ZAB assembled with 6 M KOH and the electrodeposited Co-Fe GDL showed stable bifunctional cycling for over 20 h at 5 mA cm −2 [37].In this case, the decision between a hybrid soaking and impregnation technique and a direct electrodeposition process comes down to the amount of deposited material, the penetration depth of catalyst loading, and whether a binder-free process is preferred.Rather than change the synthesis process, the substrate material can also be varied.For example, an ALD process for a Co-Fe oxide catalyst could easily be applied to a metal foam substrate [124,125].In this case, the O-plasma reactant is not detrimental to the substrate and the direct loading technique, with all its benefits, could be applied to the air electrode.Thus, the future of catalyst integration into the air electrode of ZABs will depend both on the synthesis method for state-of-the-art catalysts and the best practices for air electrode materials.

Conclusions
The loading of catalyst material on the air electrode of a ZAB is as critical as the choice of catalyst chemistry.Indirect catalyst loading methods, such as an ink and spray-coating techniques, are associated with poor stability due to binder decomposition, carbon corrosion, and a 2D catalyst interface with the electrolyte.Direct catalyst loading methods, such as electrodeposition or ALD, provide improved stability over indirect methods due to binder-free integration, enhanced conductivity pathways, and a 3D interface with the electrolyte that is less susceptible to flooding.Additionally, hybrid catalyst integration methods, such as a reported soaking and impregnation technique, can deliver performance on par with direct methods due to the synergy with nanostructured substrates and deep penetration of catalyst material within the air electrode structure.ZABs with either direct or hybrid catalyst loading exhibit superior cycling stability to indirect loading counterparts and are the key to long lifetime, stable ZABs.

Figure 1 .
Figure 1.Depiction of an idealized air electrode during charge and discharge.(a) A careful balance of hydrophobicity, ORR catalyst distribution, and OER catalyst distribution ensures that the two-phase boundary area between liquid electrolyte and the solid current collector during OER is catalyzed by OER catalyst particles, and the three-phase boundary area between liquid electrolyte, the solid current collector, and gaseous O2 during ORR is catalyzed by ORR catalyst particles.(b) During flooding, the electrolyte penetrates deeper into the air electrode, shifting the three-phase ORR process into the un-catalyzed region, reducing battery performance.

Figure 2 .
Figure 2. Process flow of indirect catalyst loading techniques.

Figure 3 .
Figure 3. ZAB cycling for an indirectly loaded catalytic air electrode.(a) The first 200 min of charge and discharge potentials and (b) the total charge and discharge potential curves after 200 cycles (33 h).Adapted with permission from [51].Copyright (2014) American Chemical Society.

Figure 5 .
Figure 5. Direct ALD of a (Mn,Fe)3O4 catalyst on GDL carbon particles.(a) STEM image of the GDL particles with the ALD coating represented by dark bands around the particles, (b-d) EDX mapping of Fe, Mn, and O, respectively, showing the coverage of the carbon particles with the thin ALD coating.(e) Cycling performance of ZABs with (Mn,Fe)3O4, Pt-Ru-C, and uncoated GDL air electrodes.

Figure 6 .
Figure 6.(a) Process flow of soaking and filtering hybrid loading technique devised by Aasen et al [61].(i) Nanostructured support, (ii) direct growth of catalyst particles anchored onto nanostructured support, (iii) suspension of nanoparticle loaded nanostructures, (iv) soaking GDL in the suspension, and (v) filtration of the suspension through GDL.(b) Scanning electron microscopy (SEM) image of the GDL surface after soaking and filtering, (c) TEM image of (Co,Fe)3O4/N-CNTs, with yellow arrows indicating nanoparticles and blue arrows indicating N defects in the CNT wall.(d) STEM image and (e)-(g) O, Co, and Fe EDX mapping, respectively, of (Co,Fe)3O4/N-CNTs.(h) ZAB cycling of the hybrid loaded (Co,Fe)3O4/N-CNT catalyst [111].John Wiley & Sons.© 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim.

Figure 9 .
Figure 9. Plot of bifunctional efficiency retention after cycling against the total number of charge-discharge cycles.Indirect catalyst loading data are shown as empty circles, direct loading as black squares, and hybrid methods as triangles.The numbers beside data points refer to the reference number for that work.

Table 1 .
Compilation of reported cycling performance for different ZAB air electrode catalysts, categorized based on the deposition method employed.Values in square brackets are indirectly obtained based on reported cycling details but are not explicitly stated.If efficiency values are not directly reported in an article, an estimate is provided based on the published figures, similar to the method shown in figure3.All studies utilize a liquid 6 M KOH electrolyte.