Chem. Rev. - ACS Publications - American Chemical Society

Jan 31, 2018 - A comprehensive review of recent advances in the field of oxygen reduction electrocatalysis utilizing nonprecious metal (NPM) catalysts...
0 downloads 0 Views 9MB Size
Review Cite This: Chem. Rev. 2018, 118, 2313−2339

pubs.acs.org/CR

Nonprecious Metal Catalysts for Oxygen Reduction in Heterogeneous Aqueous Systems Andrew A. Gewirth,*,†,‡ Jason A. Varnell,† and Angela M. DiAscro† †

Department of Chemistry, University of Illinois at Urbana−Champaign, Urbana, Illinois 61801, United States International Institute for Carbon Neutral Energy Research (WPI-I2CNER), Kyushu University, Fukuoka 819-0385, Japan



ABSTRACT: A comprehensive review of recent advances in the field of oxygen reduction electrocatalysis utilizing nonprecious metal (NPM) catalysts is presented. Progress in the synthesis and characterization of pyrolyzed catalysts, based primarily on the transition metals Fe and Co with sources of N and C, is summarized. Several synthetic strategies to improve the catalytic activity for the oxygen reduction reaction (ORR) are highlighted. Recent work to explain the active-site structures and the ORR mechanism on pyrolyzed NPM catalysts is discussed. Additionally, the recent application of Cu-based catalysts for the ORR is reviewed. Suggestions and direction for future research to develop and understand NPM catalysts with enhanced ORR activity are provided.

CONTENTS 1. Introduction 2. Pyrolyzed Nonprecious Metal Catalysts 2.1. Catalyst Preparation and Performance 2.1.1. Multiple Precursors 2.1.2. N-Rich Polymeric Precursors 2.1.3. Metal−Organic Frameworks 2.1.4. Hard Templates 2.1.5. Heteroatom Doping 2.1.6. Carbon/Nitrogen Catalysts 2.2. Active-Site and Mechanistic Studies 2.2.1. Characterization of Active Species 2.2.2. Catalyst Degradation 2.2.3. Oxygen Reduction Reaction Mechanism 3. Oxygen Reduction Reaction Catalysts Based on Copper 4. Overview and Recommendations (A Research Proposal) Author Information Corresponding Author ORCID Notes Biographies Acknowledgments References

this lack has many origins but most of these relate to a single molecule: oxygen.1 Figure 1 shows a typical fuel cell, in which a fuel such as H2 is oxidized at the anode, releasing protons, which cross the membrane or separator, and electrons, which can be used for power. At the cathode, the protons and electrons are used to reduce oxygen via the oxygen reduction reaction (ORR). Different types of fuel cells include solid oxide fuel cells (SOFC), which typically operate at high temperatures where the kinetics of both oxidation and reduction are not problematic. However, durability remains an issue for an SOFC.2 A low-temperature fuel cell can be of two types. The

2313 2314 2315 2315 2318 2318 2319 2320 2321 2321 2321 2325 2326 2326 2328 2328 2328 2328 2328 2328 2329 2329

1. INTRODUCTION Fuel cells are an attractive way of extracting electrical energy from a chemical fuel. If fully implemented, fuel cells could provide motive power for automobiles, trains, and ships. They could also be part of a process to store electrical energy generated from renewable resources in a chemical fuel such as hydrogen or a hydrocarbon when the renewables power an electrolyzer. However, this promise, now approaching two centuries in duration, has not been fully realized. The reason for © 2018 American Chemical Society

Figure 1. Schematic of hydrogen/air fuel cell. Hydrogen is oxidized at the anode, producing protons and electrons used to provide electricity and to reduce oxygen to water at the cathode. Special Issue: Oxygen Reduction and Activation in Catalysis Received: June 19, 2017 Published: January 31, 2018 2313

DOI: 10.1021/acs.chemrev.7b00335 Chem. Rev. 2018, 118, 2313−2339

Chemical Reviews

Review

first, usually called a polymer electrolyte membrane (PEM) fuel cell, operates at or slightly above (ca. 80 °C) room temperature and features a hydrated fluoropolymer such as Nafion as the electrolyte.3 The second, an alkaline fuel cell (AFC), uses KOH as the electrolyte.4 When H2 is used as a fuel, the catalyst for the hydrogen oxidation reaction (HOR) is Pt. Because the HOR is very efficient, relatively little Pt is required on the anode. Alternatively, when another molecule, such as CH3OH, is used as the fuel [a so-called direct methanol fuel cell (DMFC)], differentoften less effectivecatalysts such as PtRu are utilized.4 The attraction of a fuel cell is at least 2-fold. The fuel cell provides a high-efficiency method to extract electrochemical energy from a fuel. Of course, energy can be obtained from a fuel by combustion, but the efficiency of this process is limited by the Carnot cycle and typically does not rise much greater than 20%.5 Fuel cells, on the other hand, offer the promise of efficiency close to 100%.6 The efficiency of a fuel cell (ζ) is partially determined by the overpotentials observed for the HOR and the ORR at the electrodes, given in eq 1: ζ=1−

this planet.11 Both the oxygen evolution reaction (OER) and the ORR are utilized by living organisms to sustain vital functions. Life uses oxygen via the agency of the ORR, which extracts energy from a fuel and uses O2 as the electron acceptor. Oxygen reduction is carried out in living organisms by enzymes such as cytochrome c oxidase (CCO) and laccase. In these enzymes, the overpotential for the ORR is significantly reduced compared to manmade catalysts, approaching the thermodynamic value of 1.23 V. The low overpotentials demonstrated by enzymes in nature are proof that huge improvements in catalyst activity for the ORR in fuel cells are a legitimate target. However, in these systems the electrochemical rates of the ORR are relatively slow due to H+-transfer limitations, and when they are employed as catalysts for fuel cells, the current densities are low due to the large size of the enzymes.11 As a result, several other types of catalysts have been investigated for the ORR in fuel cells. The first type involves porphyrins/phthalocyanines reprising some or all of the active sites based on CCO. A second class of catalysts seeks to mimic the active site in multicopper oxidase (MCO) enzymes such as laccase. The third class of ORR catalysts involves pyrolyzed materials, usually based on Fe or Co. The pyrolyzed materials exhibit the greatest promise, with activity and stability currently approaching that of Pt. Because of the promise of these NPM ORR catalysts, the area has been well reviewed in recent years.12−22 One recent review in Chemical Reviews focused on the direct application and performance of several types of catalysts, including NPM catalysts, in PEM fuel cells.23 To complement this recent review, we summarize advances reported since 2015 in the preparation of NPM ORR catalysts that have led to increases in activity, and we discuss the possible nature of the active sites that have been reported in recent literature.

ηanode + ηcathode ΔE°

(1)

Here ηanode and ηcathode are the overpotentials at the anode and cathode, respectively, and ΔE° is the difference between formal potentials at the anode and cathode. For a H2/O2 PEM fuel cell, ΔE° = 1.23 V. As stated, ηanode is typically small when H2 is the fuel, on the order of 20 mV or less. On the other hand, ηcathode is large, typically between 300 and 600 mV for a real PEM fuel cell. The high value of ηcathode lowers the overall efficiency of the PEM fuel cell to values around 45%. This low efficiency obviates some of the advantages attendant the fuel cell idea. The large value of ηcathode is a direct consequence of the slow kinetics of the ORR at the cathode. Developing catalysts to lower this overpotential has been a focus of research for over 60 years. The earliest and still most common catalyst is based on Pt or its alloys. Recent work has seen substantial advances to reduce Pt loading in fuel cells and decrease the value of ηcathode by careful, atomic-level control of the Pt environment, as described in recent reviews.7−10 While considerable progress has occurred to optimize the performance of Pt catalysts, many issues remain with both the amount of Pt required and the overall durability of the cathode, which is subject to degradation and poisoning in the course of ORR operation. Because of cost, activity, and durability issues surrounding the use of Pt in a PEM fuel cell, considerable effort has been directed at designing and synthesizing ORR catalysts not based on Pt. In contrast to Pt-based catalysts, an ideal ORR catalyst would be cheap, durable, and very active. During the time in which ORR catalysts have been developed, catalysts based on early transition metals such as Fe, Co, and Cu have been investigated. In a fuel cell, or almost any practical device, these catalysts must be supported or attached to the electrode surface in order to achieve high current densities required for its application. The supported catalyst facilitates electron transfer from the cathode to the catalyst/O2 complex and does not interfere with the H+ or OH− transport required for fuel cell operation. Catalysts based on first-row transition metals, or so-called nonprecious metal (NPM) ORR catalysts, find their inspiration in many sources, not the least of which is biology. Oxygen electrochemistry is ubiquitous in nature and is central to life on

2. PYROLYZED NONPRECIOUS METAL CATALYSTS Jasinski’s 1964 discovery24 that metal phthalocyanines were active catalysts for the ORR was the start of over 50 years of interrogation of catalysts of this type. Researchers later discovered that heat-treating the adsorbed metal phthalocyanine increased the activity and stability of the catalysts.25,26 Since then, there have been numerous studies directed at both elucidating the catalyst active site(s) and developing more active and more stable materials. Numerous studies report on the heterogeneity in the catalysts, a feature that has impeded progress. Spectroscopic studies show that the catalysts feature multiple chemical species, some of which may or may not be active under different conditions. Consequently, there remain substantive questions as to the source of the observed activity. In the literature, differences between the activity of various catalysts in acidic and alkaline conditions are extremely common. In particular, higher ORR onset potentials are typically observed for catalysts operating in alkaline conditions. This is especially true for catalysts based on pure carbon or nitrogen-doped carbon materials, which exhibit much higher activity in base than they do in acid.27 The differences in activity between acid and base are less significant for the Fe- or Co- based NPM ORR catalysts. These trends in the literature point to the possibility of having metal-free sites involving C or N atoms that can be active in base, whereas it is generally found that incorporating a metal is necessary to achieve high ORR onset potentials in acid conditions.27 Over the past few years, the number of new catalysts exhibiting high activity for the ORR in either acidic or alkaline 2314

DOI: 10.1021/acs.chemrev.7b00335 Chem. Rev. 2018, 118, 2313−2339

Chemical Reviews

Review

Figure 2. Synthesis summary for typical NPM ORR catalysts.

conditions has burgeoned, in part due to the use of new synthetic methods. These methods share the general requirement for a source of metal (typically Fe or Co), C, and N in the synthesis. A general summary of typical NPM catalyst synthesis is shown in Figure 2. Here, we focus on discussing several recent papers following a previous Chemical Reviews paper23 that provide important insight into the active site and mechanism of the ORR on NPM catalysts across a wide variety of synthesis methods and catalyst morphologies in both acidic and alkaline conditions. First, we report on many recently developed or improved methods used to generate active ORR catalysts providing guidance for future catalyst syntheses. Typically, the highest activity comes from Pt and/or one of its alloys. Most NPM catalysts exhibit activity lower than Pt, particularly when measured in acid. This lower activity is reported as a shift in the half-wave potential or in the “onset” potential (typically where the reduction is a few percent lower than baseline). Tables 1 and 2 provide a summary of catalyst preparation and activity for several selected NPM catalysts exhibiting high ORR activity from recent literature, as measured in either alkaline or acidic media. For convenience and to highlight the differences between catalyst properties and performance in acid and base, the results summarized in the tables have been separated, with Table 1 containing selections for alkaline conditions and Table 2 containing selections for acidic conditions. The parameters (onset and half-wave potential) reported in the tables are typically obtained by use of rotating (ring) disk electrodes, R(R)DE. Another metric reported for NPM catalysts is the peroxide yield, which is the amount of peroxide produced during the ORR. For good 4 e− ORR catalysts, this number is typically a few percent. For worse catalysts, this number can approach unity. Figure 3 shows linear sweep voltammograms (LSVs) obtained by use of RRDE for one of the most active ORR catalysts reported to date, with onset and half-wave potentials rivaling that of commercial Pt along with a low H2O2 yield.28 Second, we discuss a selection of recent reports that provide information on the role of precursors and synthesis steps to generate an active catalyst and/or discuss the mechanism and active site/species during the ORR. Such studies are vital to direct the formation of enhanced catalysts with higher activity and a higher density of active sites. Additionally, we include recent work evaluating the contributions of chemical/structural features in the catalyst to the overall stability and degradation mechanisms of the catalyst, which is of crucial importance for long-term operation of NPM catalysts in fuel cells. In particular, the role of catalyst surface species and resistance to oxidation is

described, as well as the likely role of peroxide in catalyst degradation via Fenton-type reactions with Fe. 2.1. Catalyst Preparation and Performance

With the establishment of the basic method of NPM catalyst synthesisinvolving precursor mixing and pyrolysissubsequent work has been directed at modifying the precursor mixture in an attempt to alter the activity of the final, pyrolyzed product. In many cases, altering the precursors and preparation procedure produces more active catalysts. The reason for this increase in activity is linked to many factors, including (1) nature of the active site or active species, (2) density of the active site or active species in the catalyst or on its surface, and (3) other properties such as surface area and porosity of the resultant material, as well as catalyst conductivity. While the importance of surface area and porosity as well as conductivity is apparent, the role of the metal species and the identity of the active sites are less clear in the literature. Here we describe several recent preparation methods highlighting the various strategies for developing active and stable NPM ORR catalysts. 2.1.1. Multiple Precursors. The use of multiple precursors as the sources of metal, N, and carbon support has been well described in previous literature, but many studies continue to investigate the effects of different precursors on the activity of the resulting catalysts.29 As an example, Jiang et al.30 combined carbon nanotubes (CNT), glucose, melamine, and an iron salt and observed that as the amount of Fe precursor was increased, the ORR activity increased. This increase corresponded to the presence of more crystalline Fe species. The crystalline Fe species were suggested to enhance the activity of the N and FeNx sites surrounding Fe nanoparticles, as shown in Figure 4. Another study examined the effects of using sulfate salts of different transition metals as precursors, along with aminopyridine as the N source and Black Pearls 2000 as the C support. The authors found that the Fe and Co salts introduced a higher content of pyridinic N and MeNx species, which led to better activity relative to catalysts prepared from other transition metals or those prepared without the addition of a metal.31 Kramm et al.32 used Fe and Co porphyrin and oxalate precursors along with elemental S and used a second heattreatment step in 10% H2, followed by a second acid-washing step to remove metallic species and generate a catalyst with a high loading of FeNx sites. The catalyst then exhibited high turnover frequency for the ORR. Another paper utilizing similar precursors showed that an intermediate acid leaching step before the second pyrolysis helped decrease the formation of reduced metallic species, which subsequently increased the stability of the catalyst.33 Using an already established precursor method based on an iron phenanthroline complex adsorbed on Black Pearls, other researchers showed that, for catalysts 2315

DOI: 10.1021/acs.chemrev.7b00335 Chem. Rev. 2018, 118, 2313−2339

Chemical Reviews

Review

Table 1. Catalyst Summary in Alkaline Conditionsa metal Fe Fe Co Fe Fe Fe Fe Co Fe Fe Fe Fe Fe Fe Co Co Co/Ni Co Fe Fe Fe Fe Fe Fe Fe Co Fe Fe

e e

onset (V)

E1/2 (V)

suggested active siteb

Precursor Method 1.00c 0.899 FeNx@Fe/Fe3C 0.98 0.83 FeNx 0.974 0.858 Co/CoOx 1.0c 0.89c Fe3C@C 1.027d 0.934d Fe/Fe3C 1.20c 0.934 FeNx or pyr-N 0.956d 0.868d pyr/graph-N 0.97 0.82 Co−N (Co2N) 0.99 0.85 Fe−N (Fe3N) Polymer Method 1.02d 0.912d FeNx 0.99 0.83c graph-N/Fe3C 0.96c 0.87c pyr/pyrr-N 0.960 0.877 graph/pyr-N 0.96c 0.84c Fe−C, Fe3C, and Fe3O4 Metal−Organic Framework Method 0.96c 0.90 Co−N/C 0.982 0.881 CoNx 0.996d 0.877d Co/Ni NPs 0.976 0.849 pyr/graph-N Hydrogen-Bonded Organic Framework Method 0.962 0.83c FeNx, graph-N Iron Phthalocyanine Method 0.98 0.87 pyr-N 1.057d 0.987d N−C (pyr/pyrr) Template Method 0.970 0.869 FeNx 0.95c 0.85c N or FeNx 0.98 0.82c graph-N, FeNx 0.95c 0.88 FeNx Heteroatom Method 0.98 0.88 Co NPs and CoNx/Sx 1.02 0.848 Fe or N/C or S/C 0.95 0.79c Fe−P and P−C 0.94 0.82 N/P−C C/N Method 0.912 0.737 edge/defect-C 0.877d 0.797c,d edge/defect-C 0.867c,d 0.787c,d pyr/graph-N 0.942 0.79 N

Table 2. Catalyst Summary in Acidic Conditionsa ref

metal Fe Fe Fe Fe Fe Fe Fe Fe Fe Fe/Mo

30 36 37 38 39 41 44 46 45 103 106 107 108 111

Fe Fe Fe Fe

156 157 159 160

Fe Fe Co Fe

170

Fe

199 200

Fe Fe Fe Fe

204 128 206 209

Fe Co Fe

228 229 232 234

onset (V)

E1/2 (V)

suggested active siteb

Precursor Method 0.88 0.74c FeN4 0.82c 0.72c FeN4 0.858d 0.738d FeN4 0.86c 0.77c Fe−N 0.928 0.785 Fe3C@C 0.865d 0.748d FeNx 0.88c 0.793 FeNx or pyr-N 0.879 0.68c FeNx 0.859d 0.629d pyr-N, FeNx 0.845 0.674 FeNx, MoNx Polymer Method 0.83c 0.72c pyr-N/FeNx d 0.864 0.704d FeNx 0.92 0.81c pyr-N/FeNx 0.849d 0.769d Fe NPs Metal−Organic Framework Method 0.97c 0.90c FeNx 0.95 0.82 FeNx 0.875 0.675 pyr/graph-N, Co NPs 0.95 0.81 FeNx Covalent Organic Framework Method 0.84 0.73c FeN4 Template Method 0.84c 0.73c FeNx 0.908d 0.688c,d FeNx 0.90c 0.75 FeNx 0.88c 0.79 FeNx Heteroatom Method 0.94 0.86 Fe, N, S 0.88 0.78 Co NPs and CoNx/Sx 0.85 0.66 Fe or N/C or S/C 0.90 0.69c N/P−C

ref 32 33 34 35 38 39 41 42 47 48 102 103 105 110 28 158 161 162 169 197 201 208 209 227 228 229 234

a

Synthesis method, metal precursor, onset potential, half-wave potential, and proposed active site for selected catalysts in recent literature exhibiting high ORR activity bPyr-N, pyridinic N-atom; Graph-N, graphitic N-atom; NPs, nanoparticles. cValue estimated from cyclic or linear sweep voltammograms (CV or LSV) shown in given reference. dValue converted to reversible hydrogen electrode (RHE) from potential scale given in reference, where ideal reference electrodes are assumed.

268 269 272 277

a

Synthesis method, metal precursor, onset potential, half-wave potential, and proposed active site are listed for selected catalysts in recent literature exhibiting high ORR activity. bPyr-N, pyridinic Natom; graph-N, graphitic N-atom; pyrr-N, pyrrolic N-atom; NPs, nanoparticles. cValue estimated from cyclic or linear sweep voltammograms (CV or LSV) shown in given reference. dValue converted to reversible hydrogen electrode (RHE) from potential scale given in reference, where ideal reference electrodes are assumed. eMetal utilized during synthesis steps to give the resulting catalyst.

led to higher active-site density, presumably caused by the higher gas-phase pressure and retention of N in the catalyst. Researchers used 1,4,8,11-tetraazacyclotetradecane, urea, and graphene as precursors and observed that a combination of high surface area and surface N species yielded a better descriptor for higher activity than simply total N content.36 Other researchers found that, by adding ethylenediaminetetraacetic acid (EDTA) to a mixture of cobalt nitrate, Dglucosamine hydrochloride, and melamine, the surface area and encapsulation of Co metallic and oxide species by graphitic carbon were increased following pyrolysis, again leading to enhanced activity.37 Zhu et al.38 used dicyanamide, ferric chloride, and Black Pearls as precursors and showed that the formation of Fe3C aggregates encapsulated within N-doped CNTs, shown in Figure 5, functions as an ORR catalyst with high activity in both acid and base conditions. Pan et al.39 studied catalysts prepared from starch, urea, and ferric chloride pyrolyzed at different temperatures and found that graphitic N and Fe nanoparticles formed at higher temperatures correlated

pyrolyzed multiple times, the first pyrolysis determines the Fe speciation in the final catalyst, while a second pyrolysis in NH3 affects several other factors such as the relative surface abundance of the Fe species, the surface N content, and electrochemically accessible surface area, all of which contribute to increased catalyst activity.34 Gumeci et al.35 prepared a catalyst from typical precursors including iron acetate, melamine, and Ketjen Black. Using highpressure pyrolysis in a sealed quartz tube, these researchers found that putting a larger amount of the mixture in the reactor 2316

DOI: 10.1021/acs.chemrev.7b00335 Chem. Rev. 2018, 118, 2313−2339

Chemical Reviews

Review

with the observed activity in base, while FeNx formed at lower temperatures correlated with activity in acid. Another group also investigated the ORR activity of a catalyst pyrolyzed at different temperatures. They observed that the ORR activity in base decreased as the pyrolysis temperature increased above 600 °C. Using extended X-ray absorption fine structure spectroscopy (EXAFS), they suggested that the decomposition of FeNx species to from nanoparticles at higher temperatures was responsible for the decrease in activity.40 Other researchers found that mixing carbon black and ferric chloride along with melamine and terephthalaldehyde to form a network via a condensation reaction led to a catalyst with pyridinic N species exhibiting good ORR activity in base but lesser activity in acid, ostensibly due to the protonation of the pyridinic N in the active sites.41 Sun et al.42 used precursor-based methods with ferrous chloride hydrate to synthesize Fe3C and Fe2N catalysts with and without N doped into the carbon support and showed that the presence of N in the support stabilizes the formation of FeNx species, giving rise to high ORR activity. Alternatively, Fe2N and Fe3C alone are not very active. However, other researchers found that by including a W precursor along with Fe and Co, WC, Fe3C, and Co3C phases were created, leading to some enhancement in activity.43 He et al.44 synthesized a catalyst using ferric chloride along with ionic liquids based on N-methylimidazole, shown in Figure 6. By changing the substituents on the imidazole, invoking precursor rigidity, the authors altered Fe distribution and

Figure 3. RRDE measurements recorded in O2-saturated electrolyte, pH 1, for a series of prepared NPM catalysts and commercial Pt: (a) percentage of H2O2 detected and (b) polarization curves for ORR. Adapted with permission from ref 28. Copyright 2015 Nature Publishing Group.

Figure 4. Correlation between atomic percent of N−Fe coordination, as determined by X-ray photoelectron spectroscopy (XPS), and E1/2 values for catalysts featuring different Fe coordination environments (Fe−N for Fe@C-FeNC-1, both Fe−N and Fe−Fe for Fe@C-FeNC2, and Fe−Fe for Fe@C-FeNC-3). Reprinted with permission from ref 30. Copyright 2016 American Chemical Society.

Figure 6. Synthesis and chemical structures of ionic liquids based on N-methylimidazole used as precursors to form a NPM catalyst. Reprinted from ref 44. Published 2016 Royal Society of Chemistry under license CC BY-NC 3.0.

overall graphitization of the pyrolyzed catalyst. Different substituents also changed the amount of graphitic and pyridinic N species, the mesopore content, and the catalyst conductivity, all of which helped achieve better activity. Another group also used an ionic liquid as a precursor with Fe and Co, and they suggest that Fe3N and Co2N are active species.45,46 Yang et al.47 synthesized a new N-rich ligand, 6,7-di(pyridin-2-yl)pteridine2,4-diamine, which forms ferrous complex polymers when coordinated with Fe and ostensibly both prevents Fe from aggregating and stabilizes the N species during pyrolysis. The catalyst exhibiting the highest pyridinic N content also exhibited the highest activity of the catalysts studied. Similarly, Lin et al.48 used the substitution of Fe(II) into a Mo(V) coordination polymer containing 11′-bis(dipyrido[3,2-a:2′,3′c]phenazine) to prevent the aggregation of Fe and produce an active catalyst with MeNx sites, as shown in Figure 7. These

Figure 5. (a) Transmission electron microscopic (TEM) image of a pyrolyzed NPM catalyst, showing Fe3C aggregates encapsulated by Ndoped graphitic carbon, and (b) high-resolution TEM image showing lattice spacing of Fe3C in an encapsulated nanoparticle. Reprinted with permission from ref 38. Copyright 2015 Royal Society of Chemistry.

2317

DOI: 10.1021/acs.chemrev.7b00335 Chem. Rev. 2018, 118, 2313−2339

Chemical Reviews

Review

carbon black. These researchers found that precursors leading to FeNx structures with Fe(III) and high quaternary N content produced more active catalysts.104 Another group also used a PANI-derived catalyst and treated it with NH3 during pyrolysis. This group found that NH3 treatment increased the activity, ostensibly by increasing the surface area and creating micropores to host FeNx species. In this catalyst, the overall N content was actually decreased by the treatment.105 Also using PANI, researchers generated a catalyst with encapsulated Fe3C particles with graphitic N on the surface that exhibited activity similar to Pt in base.106 Osmieri et al.107 used poly(vinylpyrrolidone) (PVP) along with mesoporous carbon and iron acetate to show that the total N and Fe content are not limiting factors for ORR activity but rather that the specific presence of pyridinic and pyrrolic N species, along with Fe exposed by an increase in microporosity from a second heat treatment, correlate with high catalyst activity. Another group used polypyrrole and ferric chloride as precursors that were pyrolyzed and then acid-leached; they found that several factors including amount of graphitic carbon, conductivity, and surface area were related to the observed catalyst activity. In this case the presence of graphitic and pyridinic N incorporated by the metal precursor resulted in the highest activity Fe species.108 The effects of acid washing to increase the activity were also found with a PVP catalyst following removal of amorphous carbon from the catalyst. This treatment increased the graphitic order of the surface and exposed more active species.109 Two other studies using catalysts produced via the electrospinning of a precursor mixture to form fibers show that treatments in H2 or air also increase the activity of the catalysts, ostensibly by creating pores that contain active species. The authors suggest that the treatment produces active centers, such as carbon defect sites and exposed metallic species, possibly without the requirement of Me−N coordination.110,111 Other examples of catalysts derived from polymeric materials have also been reported.112−154 2.1.3. Metal−Organic Frameworks. Another common strategy in recent work is the use of metal−organic frameworks (MOFs) and other similar frameworks as sources of metal, N, and carbon to help increase the dispersion of the metal sites within the resulting porous catalysts. In an early example, Zitolo et al.28 used a zeolitic imidazolate framework (ZIF) into which a small amount of an iron phenanthroline complex was dispersed via ball-milling, followed by two heat treatments in Ar and NH3, respectively, to obtain a catalyst with a preponderance of FeNx sites and high activity in acid. More recent work from the same group showed that there is some correlation between the pore volume of the ZIF before pyrolysis and the ORR activity of the produced NPM catalyst.155 Another group used ZIF isomorphs with direct incorporation of 5% Co in the ZIF to achieve high activity from the dispersion of Co and retention of CoNx sites after pyrolysis.156 A similar work also found that direct incorporation of Co into the ZIF led to increased formation of single metal atom CoNx sites, enhanced by the formation of vacant N sites left by volatilization of the Zn during pyrolysis as shown in Figure 9.157 Similarly, Wang et al.158 showed that by substituting Fe(II) into ZIF-8 particles and using a surfactant to control the particle size, the aggregation of Fe during pyrolysis was minimized, generating a more active catalyst. Another group demonstrated that the intercalation of other transition metals (Ni, Fe, Zn, and Cu) into a Co-based ZIF led to the formation of protected transition metal (TM)/Co alloyed nanoparticles that were

authors found enhanced graphitization at lower temperatures caused by the presence of Mo. Finally, several other reports

Figure 7. Process for formation of a coordinated Mo precursor and ion exchange with Fe, followed by pyrolysis and acid leaching to form a NPM catalyst with Fe and Mo. Reprinted with permission rom ref 48. Copyright 2016 American Chemical Society.

utilize different mixtures of metal, N, and C precursors to give active catalysts.49−99 2.1.2. N-Rich Polymeric Precursors. The use of N-rich polymeric precursors, such as polyaniline (PANI), gives catalysts with high activity.100 Since this initial report, many other N-rich polymeric precursors have been examined. Kuroki et al.101 attempted to use 15N magic-angle spinning to interrogate the N species present in a catalyst prepared with 15 N-labeled PANI. They observed signals from several N species, including pyridinic and graphitic N, as well as Fe3+−N coordination. One study used different polymeric precursors, shown in Figure 8, with Fe precursors and correlated the polymer structure with the ORR activity of pyrolyzed catalysts.

Figure 8. (a−c) Polymerization schemes for different phenylenediamine monomers, leading to different N coordination and available N species in the precursor mixture. (d) Schematic of polymer coating of carbon black. Reprinted with permission from ref 102. Copyright 2016 Elsevier.

In this report, better metal dispersion and higher pyridinic N content led to the formation of more FeNx sites and higher activity.102 Another group used a copolymer network based on 2,4,6-tripyrrole-1,3,5-triazine and pyrrole along with iron acetate and found that better activity was associated with the increased presence of FeNx sites created by the N-rich polymer.103 The influence of the Fe precursor salt was investigated by mixing different salts with polyaniline and 2318

DOI: 10.1021/acs.chemrev.7b00335 Chem. Rev. 2018, 118, 2313−2339

Chemical Reviews

Review

sites.164 Researchers used a Co/Zn ZIF to combine the effects of graphitization from Co and porosity from Zn and found that the optimal activity was achieved with 25% Co in the ZIF.165 Another work mixed Cu nanowires into ZIF-8 and pyrolyzed the mixture to give a catalyst with a variety of N sites and high ORR activity in base.166 Eisenberg et al.167 used a Mg MOF with nitrilotriacetate linkers and showed that the level of Ndoping in the final catalyst corresponded to changes in the observed activity. Another group used a Zr MOF with Cu− porphyrin linkers and found that annealing with a small air impurity led to increased activity caused by the etching of carbon material and increased exposure of the Cu(II) sites.168 Ren et al.169 used a covalent organic framework with Fe− porphyrin linkers to form a catalyst with FeN4 species exhibiting high activity in acid but which also yielded a relatively high peroxide yield of about 10%. Liu et al.170 used a hydrogen-bonded organic framework made from melamine and trimesic acid mixed with Fe and excess Zn salts to produce a catalyst with a high content of graphitic N and a high surface area featuring high activity. The authors speculated that this activity was possibly promoted by the presence of Fe remaining in the catalyst. The use of MOFs to generate catalysts with high activity is also evident in other recent work.171−194 2.1.4. Hard Templates. Another strategy to prepare active NPM ORR catalysts features the use of hard templates along with other precursors. The hard templates are suggested to promote the formation of pores and increase catalyst surface area. SiO2, which can be removed via KOH or HF washing, has been commonly utilized as the hard template in recent years due to previous development of the method.195 Stariha et al.196 used ferric nitrate and mebendazole precursors with a fumed silica template. The materials were combined via ball-milling and pyrolyzed before being leached in HF to remove the silica and produce a catalyst. Scanning electron microscopy (SEM)/ focused ion beam (FIB) cross sections and calculation of the two- and three-dimensional (2D and 3D) Euler numbers suggested the presence of a high concentration of MeNx sites within highly connected pores in the most active materials. Sebastian et al.197 showed that additional acid washing and NH3 treatment steps on a silica-templated catalyst derived from ferric nitrate and aminobenzimidazole slightly improved the activity due to the removal of amorphous carbon and metallic species, possibly resulting in more exposed FeNx sites. Another group showed that an even distribution of Fe in micropores created from silica templates increased the activity and lowered the H2O2 yield from catalysts made from pyrolyzed iron phthalocyanine (FePc).198 Osmieri et al.199 used different transition metals as precursors with a porous silica template. They found that the role of the metal was to promote formation of pyridinic N and decompose the organic precursor to form micropores. In this case Fe exhibited the highest activity. Yang et al.200 used a template to produce ordered mesoporous carbon and showed that without the presence of N or Fe the catalyst is not active for the ORR. However, after N-doping (without a metal) the activity increased, and by inclusion of an Fe source the catalyst achieved high activity. This activity remained high after acid washing, leading the authors to suggest that catalysts containing C−N sites and graphitic carbon are most active. Liu et al.201 used a silica template and different ionic liquids with ferric chloride to synthesize catalysts and found that increased Fe dispersion to form small particles and FeNx species (with pyridinic N) resulted in the highest activity. Anibal et al.202

Figure 9. Schematic showing (a) formation of Co nanoparticles from a Co-containing MOF (ZIF-67) and (b) formation of single metal atom Co sites from a bimetallic MOF with both Co and Zn. Reprinted with permission from ref 157. Copyright 2016 Wiley−VCH.

uniformly distributed in the resulting catalyst. The Ni/Co particles exhibited the highest degree of graphitization of the carbon matrix and the highest activity.159 Li et al.160,161 used a Co/Zn ZIF with different ratios of Co (0.25−2) deposited onto Co/Al layered double hydroxide nanoplatelets and then pyrolyzed these to form a honeycomblike structure with incorporated Co nanoparticles and pyridinic and graphitic N species. Other researchers used an Fe-doped ZIF and showed that supporting the ZIF crystals on CNTs helped to improve the conductivity of the pyrolyzed catalyst. However, this treatment had minimal effect on activity.162 The use of a porous carbon matrix along with a ZIF as a precursor was also found to produce a catalyst with high activity.163 Another study wrapped ZIF crystals with graphene oxide, as shown in Figure 10, and found that this led to higher retention of N after pyrolysis. The resultant material included more pyridinic and graphitic species as well as an increase in 2−5 nm diameter pores that enabled better accessibility to the N

Figure 10. Schematic showing formation of ZIF-8 particles encapsulated by sheets of graphene oxide. Reprinted with permission from ref 164. Copyright 2016 American Chemical Society. 2319

DOI: 10.1021/acs.chemrev.7b00335 Chem. Rev. 2018, 118, 2313−2339

Chemical Reviews

Review

of template in HF and a second pyrolysis were found to increase catalyst durability due to the high degree of graphitization. Sa et al.209 applied a silica coating to an iron porphyrin adsorbed onto CNTs, as shown in Figure 12, to confine the precursors during pyrolysis and prevent Fe

found that different templates could be used to vary the pore sizes in the resulting catalyst, shown in Figure 11. The size of the pores in the catalyst reflected the size of the template, but a

Figure 12. Synthetic scheme for preparation of NPM catalyst. (Upper pathway) A silica coating was used to confine the precursors on the surface of carbon nanotubes (CNT) and prevent the formation of Fe aggregates. (Lower pathway) Same synthesis without the use of a silica coating, which results in the formation of Fe and Fe3C species. Reprinted with permission from ref 209. Copyright 2016 American Chemical Society.

aggregation, which resulted in more FeNx sites and higher activity. Hard templates derived from silica and other materials have been utilized in other recent syntheses for NPM catalysts as well.210−226 2.1.5. Heteroatom Doping. Heteroatom doping is another method used to alter the activity of NPM catalysts. Research in this area includes attempts to incorporate dopants such as S and P into the catalyst surfacein addition to or in place of the metal and N typically used. As an example, Kwak et al.227 showed that S-doping from thioacetamide with iron porphyrin precursors inside a porous silica template increased the pyridinic N content (as determined by XPS), leading to a small increase in the ORR activity. The codoping was thought to increase the favorability of pyridinic N site formation. Another group added different S precursors to a Co MOF, as shown in Figure 13. They found that incorporation of S led to higher N content and a greater percentage of pyridinic N. In

Figure 11. Pore-size distributions for carbon (metal-free) and catalysts (Fe) resulting from the use of several different silica particles as hard templates, as determined from (a) Barrett−Joyner−Halenda (BJH) and (b) density functional theory (DFT) methodologies. Reprinted with permission from ref 202. Copyright 2016 Elsevier.

second pyrolysis after the template was removed resulted in pore collapse. These authors also found that catalysts with intermediate size pores (∼10 nm) outperformed catalysts with smaller and larger pores. Other researchers found that pore size and shape greatly affected the activity of a variety of templated catalysts, with ultramicro pores (