Fundamental Mechanistic Understanding of Electrocatalysis of Oxygen Reduction on Pt and Non-Pt Surfaces: Acid versus Alkaline Media

Complex electrochemical reactions such as Oxygen Reduction Reaction (ORR) involving multi-electron transfer is an electrocatalytic inner-sphere electron transfer process that exhibit strong dependence on the nature of the electrode surface. This criterion (along with required stability in acidic electrolytes) has largely limited ORR catalysts to the platinum-based surfaces. New evidence in alkaline media, discussed here, throws light on the involvement of surface-independent outer-sphere electron transfer component in the overall electrocatalytic process. This surface non-specificity gives rise to the possibility of using a wide-range of non-noble metal surfaces as electrode materials for ORR in alkaline media. However, this outer-sphere process predominantly leads only to peroxide intermediate as the final product. The importance of promoting the electrocatalytic inner-sphere electron transfer by facilitation of direct adsorption of molecular oxygen on the active site is emphasized by using pyrolyzed metal porphyrins as electrocatalysts. A comparison of ORR reaction mechanisms between acidic and alkaline conditions is elucidated here. The primary advantage of performing ORR in alkaline media is found to be the enhanced activation of the peroxide intermediate on the active site that enables the complete four-electron transfer. ORR reaction schemes involving both outer- and inner-sphere electron transfer mechanisms are proposed.


Introduction
Oxygen reduction reaction (ORR) on noble and nonnoble metal surfaces remains as one of the well-investigated electrochemical processes. While under acidic conditions Ptbased systems remain the mainstay as catalyst materials for ORR, in alkaline electrolyte a wide range of nonnoble metals and their oxides are stable enough for practical applications [1,2]. ORR pathway rather than ORR mechanism has typically been addressed in the literature due to the easy accessibility of the former from rotating ring-disk electrode (RRDE) studies, and the complexity in understanding the latter [3]. ORR pathway is found to be similar in both acid and alkaline media on Pt-based materials [3,4]. Based on the initial propositions by Damjanovic et al. [4][5][6], rate determining step (rds) on Pt electrodes is widely agreed to be the first electron transfer step to the adsorbed molecular O 2 with or without rapid proton transfer [4][5][6]. A major alternative viewpoint to rds in ORR was proposed by Yeager et al. [7], wherein it was proposed that ORR on Pt surfaces is likely to involve dissociative chemisorption of molecular O 2 with the initial adsorption of O 2 (with or without an electron transfer) as the rds. ORR on Pt-based catalysts is understood to proceed via "parallel" routes with the 4e − "direct" or "series" pathway as the predominant route and a minor route involving 2e − pathway to peroxide. Various kinetic models for ORR have been developed to understand the reaction pathways. The first model was developed by Damjanovic et al. [8] following which Wroblowa et al. [9], and Anastasijević et al. [10,11], proposed extensive models.
For electrocatalytic reactions proceeding via inner-sphere electron transfer mechanism, it is typically assumed that 2 Advances in Physical Chemistry either molecular adsorption of reactant species (dissociatively or nondissociatively) or an electron transfer is the first step [12]. For neutral, nonpolar species like molecular O 2 , direct molecular O 2 adsorption is likely to be inhibited relative to, for example, the adsorption of charged superoxide radical anion (O 2 •− ) unless the free energy of adsorption of O 2 molecule is very exothermic on a specific catalyst surface. This is especially true under fuel cell conditions, where the cathodic reaction typically occurs at potentials well positive of the potential of zero charge (pzc). Multistep, multielectron transfer processes like ORR that involves many adsorbed intermediates are undoubtedly classified as an inner-sphere electron transfer reaction. However, among the many elementary reaction steps involved in ORR, there could be a surface-independent outer-sphere electron transfer component in the overall electrocatalytic 4e − inner-sphere electron transfer reaction. In that perspective, O 2 reduction by one-electron transfer to superoxide (O 2 •− ) is observed at E • = −0.3 ± 0.03 V versus SHE corresponding to ΔG • = 30± 2 kJ mol −1 with both O 2 and O 2 •− remaining in the aqueous phase [13,14]. Given the pH independence of this redox couple (O 2 /O 2 •− ), the potential of this reaction does not change as the pH is varied from zero to fourteen [15]. Due to the occurrence of four proton transfer steps in O 2 reduction to H 2 O/OH − , its standard reduction potential changes by 0.828 V from 1.229 V to 0.401 V versus SHE as the pH value changes from zero to fourteen. This causes the overpotential for the first electron transfer step (O 2 /O 2 •− ) to decrease from 1.53 V at pH = 0 to 0.7 V at pH = 14 indicating a sharp decrease in overpotential at alkaline pH conditions. Blizanac et al. [15] argued based on a modified Pourbaix diagram approach that the above-mentioned decrease in overpotential is the primary thermodynamic reason for the applicability of a wide range of nonnoble materials in alkaline media. Due to the high overpotential required for O 2 /O 2 •− redox couple in acidic media, only certain specific catalyst surfaces such as platinum that offer high free energy of adsorption for O 2 can catalyze ORR in acidic media. While the decrease in overpotential for the first electron transfer in alkaline media is certainly significant, this argument is primarily of thermodynamic origin. The concept of involving the possibilities of outer-sphere electron transfer during ORR in alkaline media bears importance, and it was pointed out earlier by Bockris [16] and Appleby [17] that the exchange current density values in alkaline media exhibit near independence on a large number of electrode materials including silver, gold, manganese oxides, perovskites, and various carbon surfaces. So certain steps in the overall ORR process in alkaline media could proceed via a non-electrocatalytic mechanism [17].
The drive to replace expensive and scarce Pt-based catalysts for ORR has led to a class of electrocatalysts composed of first-row transition-metal ions stabilized by surface nitrogen functionalities on graphitic surfaces [2,[18][19][20][21]. The nature of the active site (obtained after heat treatment) in terms of its location on the carbon support (edge versus basal plane) [22], coordination number (Fe-N 4 versus non-Fe-N 4 environment) [23], and chemical identity of the nitrogen functional groups (pyridinic, pyrrolic, and quaternary) [24] have remained a key aspect of intense discussion. Several theories exist to explain the nature of the active site such as those proposed by van Veen et al. [25][26][27], McBreen et al. [28], Schulenburg et al. [29], Yeager et al. [2,30], Scherson et al. [31][32][33], and Dodelet et al. [22,[34][35][36][37][38][39][40][41][42][43][44][45]. Although some authors observed that ORR is conducted by sites comprised of surface nitrogen groups devoid of any metal ion centers [46,47], it is now widely accepted that the transition-metal ion centers coordinated to four nitrogen groups (Me-N 4 ) on graphitic surfaces constitute the active site [22,23,30,48], whereas chelation primarily serves to prevent the metal center from passivation/corrosion under electrochemical conditions [49]. Besides the exact structure of the active site, the low active site density or the metal loading that is obtained in these catalysts eludes clear understanding. A maximum in catalytic activity is obtained at a very low metal loading (∼5000 ppm by weight for inorganic precursors and ∼2 wt% for macrocycle precursors) [35]. Dodelet et al. [50] suggested that only the microporosity generated during pyrolysis upon gasification of disordered carbon content acts as a host for active sites. Irrespective of whatever the exact nature of the active site is, this does not necessarily explain the fundamental origin of ORR activity in this class of heattreated catalysts. It was recently pointed out by Bezerra et al. [51] and Garsuch et al. [52] that most studies on this class of materials have focused on the optimal synthesis conditions and structure necessary for maximum activity, whereas a more fundamental understanding will be of great help in developing innovative routes for new catalyst synthesis.
In this paper, we present a systematic study of ORR reaction mechanisms on Pt and non-Pt surfaces. Fundamental reasons for the so-called kinetic facility of ORR in alkaline media are investigated from the perspectives of the changes in the double-layer structure and the reaction mechanisms in transitioning from acidic to alkaline conditions. New evidence unraveling the influences of inner-sphere and outer-sphere electron transfer mechanisms during ORR is elucidated. Besides the efforts to understand the nature of the active site in heat-treated metal macrocycle-based catalysts, a combination of electrochemical and advanced X-ray absorption spectroscopy techniques has been utilized to understand (1) the fundamental origin of electrocatalytic ORR activity upon heat treatment and (2) the reasons for the low density of active metal sites. To this end, a comparative study of ORR on Pt and non-Pt surfaces in both acid and alkaline media has been performed to (1) analyze the influence of both inner-sphere and outersphere electron transfer mechanisms, (2) elucidate the ORR reaction mechanisms, and (3) understand the importance of stabilization of the reaction intermediates on the active site.

Experimental
2.1. Electrochemical Characterization. All electrochemical measurements were made at room temperature using a rotating ring-disk electrode (RRDE) setup from Pine Instruments connected to an Autolab (Ecochemie Inc., model-PGSTAT 30) bipotentiostat. Alkaline (0.1 M NaOH) and acidic (0.1 M HClO 4 ) electrolytes were prepared using sodium Advances in Physical Chemistry 3 hydroxide pellets (semiconductor grade, 99.99%, Sigma-Aldrich) and double-distilled 70% perchloric acid (GFS Chemicals), respectively. 30% Pt/C catalyst from BASF-ETEK (Somerset, NJ) was used as received. Ru/C catalyst was synthesized in-house via aqueous route at a metal loading of 20% by weight [53]. Iron(III) meso-tetraphenylporphyrin chloride (FeTPPCl) was procured from Alfa Aesar. FeTPPCl was mixed with Black Pearl carbon in the mass ratio 1 : 4 and ball milled for 2 hours at 400 rpm followed by pyrolysis at 800 • C for 2 hours under argon atmosphere. Catalyst inks were prepared by dispersing 25 mg of the catalyst in 10 mL of 1 : 1 millipore H 2 O : isopropyl alcohol mixture along with 100 μL of 5 wt% Nafion (R) solution as a binder. 10 μL aliquot of the catalyst ink was dispensed on Glassy Carbon (GC) disk of 5.61 mm dia. Gold ring electrode was held at 1.1 V versus RHE in alkaline electrolyte and at 1.3 V versus RHE in acidic electrolyte to detect stable peroxide intermediate. Collection efficiency of the disk-ring electrode was 37.5%. All potentials are referred to reversible hydrogen electrode (RHE) scale prepared from the same solution as the bulk electrolyte unless otherwise stated.

X-Ray Absorption Spectroscopic (XAS) Measurements.
The in situ XAS studies at Fe K-edge (7112 eV) were performed at X19A beamline of National Synchrotron Light Source (NSLS, Brookhaven National Laboratory, NY). Detailed information on the spectroelectrochemical cell design are given elsewhere [54]. Spectra at Fe K-edge were collected in fluorescence mode using a PIPS detector. Argonor oxygen-saturated 0.1 M NaOH was used as the electrolyte. Complete details on data analysis of X-ray absorption near edge structure (XANES) and extended X-ray absorption fine structure (EXAFS) are available elsewhere [54]. Typical k-range window during EXAFS fit was 2.500-12.500Å −1 (Kaiser-Bessel). Data analysis for Delta-mu (Δμ) studies at Fe K-edge involved specific normalization procedures detailed elsewhere [54,55]. Briefly, this involves careful calibration of edge energy (Fe K-edge 7112 eV), alignment to standard reference scan to account for any drift in the beam energy. A postedge normalization procedure was then applied to the aligned scans via a cubic spline function which normalizes the oscillations over a specific energy range (typically 25 to 200 eV with respect to E 0 ) on a per-atom basis. Difference spectra were obtained using the equation is the XANES spectra of the catalyst at various potentials and μ(0.1 V) is the reference XANES signal at 0.1 V at which potential no evidence for electrochemical adsorbates (H upd , O ads , OH ads ) were found on iron-based catalysts. Theoretical delta-mu curves (Δμ t ) were constructed using the FEFF 8.0 code [56]. This was accomplished using the relationship Δμ t = μ(O ads -Fe-N x -C)−μ(Fe-N x -C), where the oxide species (O ads or OH ads ) is in a specific binding site on Fe.

Results and Discussions
3.1. ORR on Carbon-Supported Platinum Nanoparticles: Acid versus Alkaline Media. Figure 1 provides a snapshot of the cyclic voltammetry (CV) and ORR on 30% Pt/C in 0.1 M NaOH and 0.1 M HClO 4 electrolytes. As shown in Figure  1(a), CV of 30% Pt/C catalyst features the typical hydrogen underpotential deposition/stripping region below 0.5 V versus RHE in both of the electrolytes followed by oxide formation on Pt at potentials above 0.7 V versus RHE. Oxide formation on Pt in acidic media is due to oxidation of the solvent water molecules (water activation) [57] and in alkaline media is due to specific adsorption of hydroxide anions from the supporting electrolyte [58,59]. The onset potential of Pt-OH formation is similar in both electrolytes, although in alkaline media oxide formation current exhibits a characteristic peak shape whereas in acidic media oxide formation current is relatively more flat. Half-wave potential (E 1/2 ) of Pt-OH in 0.1 M NaOH is 0.775 V whereas in 0.1 M HClO 4 E 1/2 of Pt-OH formation is shifted slightly more positive to 0.810 V. Figure 1(b) shows the ring current measured during ORR on Pt/C at 900 rpm in 0.1 M HClO 4 and 0.1 M NaOH. The corresponding ORR polarization curves are shown in the inset of Figure 1(b). The ORR onset potential in both of the electrolytes is ∼1 V versus RHE which is followed by mixed kinetic-diffusion region between the potentials 0.7 V and 1 V. Well-defined diffusionlimited current density region is observed below 0.6 V versus RHE. The marginally higher Pt-OH coverage in alkaline media (between the potential region of 0.9 V to 0.8 V versus RHE) causes a penalty of ∼25 mV higher overpotential at a kinetic current density of i k = 1 mA cm −2 geo for ORR in 0.1 M NaOH relative to 0.1 M HClO 4 . The most interesting observation in this experiment using Pt/C is in the ring currents measured during ORR. As seen in Figure 1(b), in 0.1 M HClO 4 the ring current due to peroxide oxidation is lower in the entire potential region and does not show any significant increase until the disk potential enters the H UPD region. On the contrary, the ring current due to peroxide oxidation in 0.1 M NaOH electrolyte shows a sharp increase at 0.9 V in the cathodic scan. The sharp increase in ring current at 0.9 V during ORR in alkaline media is closely related to Pt-OH formation from specific adsorption of hydroxide anions as seen from the CV in 0.1 M NaOH. As shown in Figure 1(a), in 0.1 M NaOH electrolyte, Pt-OH formation commences at ∼0.7 V versus RHE and reaches a peak current at 0.81 V and a plateau at ∼0.9 V. On the cathodic direction, the peak potential for Pt-OH reduction is ∼0.75 V versus RHE. As seen in Figure 1(b), on the cathodic ORR scan, the increase in ring current commences at 0.9 V and reaches a peak potential of 0.75 V versus RHE. This clearly establishes the interaction between molecular oxygen and the hydroxyl species on the surface. This peroxide formation in the mixed kinetic-diffusion region in alkaline media is more evident at a higher electrolyte concentration of 1.0 M NaOH as shown in Figure 1(c). At a higher concentration of alkaline electrolyte, that is, at 1 M NaOH, peroxide is detected at the ring electrode in both the anodic and the cathodic scan [60].
In order to explain the link between the specifically adsorbed hydroxide anions and the peroxide detected in the mixed kinetic-diffusion region in alkaline media, it is important to understand the double-layer structure of the electrode-electrolyte interface during ORR at high-pH  are typically well solvated and are classically expected to populate the outer Helmholtz plane (OHP). In ORR, the case of a typical electrocatalytic inner-sphere electron transfer mechanism involves strong chemisorption of molecular O 2 (with or without bond breaking) on oxide-free Pt active sites followed by 4e − /4H + transfer. This inner-sphere mechanism depicted in Figure 2 (inset (a)) is common to both acidic and alkaline media, wherein all the reaction intermediates remain adsorbed on the active site until the final product is desorbed into the bulk electrolyte. Also shown in Figure 2 (inset (b)) is the outer-sphere electron transfer mechanism that appears to be unique to the alkaline conditions. The solvated molecular O 2 , represented here as a cluster O 2 ·(H 2 O) n , could interact with the surface hydroxyl species (OH ads ) via a hydrogen bond between H atom in OH ads and O atom in the solvent water molecule. Such hydrogen bond energies (<35 kJ mol −1 ) are typically much lower than the energy associated with covalent bond strength such as in the case of direct chemisorption of O 2 on Pt (>300 kJ mol −1 ) [61]. This hydrogen bond formation stabilizes the solvated molecular oxygen O 2 ·(H 2 O) cluster in the OHP and promotes an outer-sphere electron transfer to form the superoxide species. Following this new evidence on the involvement of outer-sphere electron transfer in alkaline media, two ORR mechanisms at high-pH environment are proposed here. The first is the well-known electrocatalytic inner-sphere electron transfer mechanism, where molecular O 2 undergoes direct chemisorption on oxide-free Pt site leading to a direct/series 4e − pathway without the desorption of reaction  intermediates (such as peroxide) from the surface according to the following well-known reaction scheme (also depicted in Figure 3) [3]: The second is the outer-sphere electron transfer mechanism, where solvated molecular O 2 cluster O 2 · (H 2 O) n weakly interacts with adsorbed hydroxyl species to promote a 2e − reaction pathway to HO 2 − anion as a reaction product which desorbs from the surface and is eventually detected at the ring electrode. This reaction is formulated as follows: The first step in the previous reaction shown in (3) involves electron transfer (or tunneling) from the electrode surface across a thin oxide film and at least one layer of solvation shell to solvated O 2 . Equation (3) involves several elementary steps as written in the following: •− ) ads . The binding energy of (HO 2 •− ) ads on the oxide substructure of Pt is likely to be lower than that on oxide-free Pt site. This leads to the facile desorption of HO 2 − anion into the electrolyte which is eventually detected at the ring electrode. Also hydrogen peroxide anion formed via the homogeneous chemical disproportionation of superoxide anion would also be detected at the ring electrode. The interaction between the O 2 ·(H 2 O) n cluster and the surface hydroxyl species causes certain nonspecificity to the identity of the underlying electrode metal. This nonspecificity opens the gate to use a wide-range of nonnoble metals and their oxides as electrode materials for ORR in alkaline media. On noble metals such as Pt, this nonspecificity is observed only in the oxide formation region, and depending on the extent of −OH coverage both inner-and outer-sphere electron transfer mechanisms coexist in alkaline media. On nonnoble metals such as the first-row transition elements that are completely passivated by a layer oxide film upon immersion into the aqueous electrolyte, this outer-sphere electron transfer process is likely to be the dominant mechanism. Furthermore as identified in Figures 1(b) and 1(c), we consider the peakshaped ring current in the potential range of 0.6 V to 0.9 V in alkaline media to be a characteristic signature of the outer-sphere electron transfer reaction mechanism. In acidic media, the adsorbed OH ads species from water activation primarily serve only to block/inhibit the adsorption of molecular O 2 and other reaction intermediates via the wellknown site-blocking effect [62,63]. However, as shown here in alkaline media, the OH ads species not only block the direct adsorption of O 2 but also serve to promote the 2e − outer-sphere electron transfer reaction to peroxide. There are several precedents for electrochemical reactions which are mediated and/or promoted by specifically adsorbed anions and surface functional groups [64][65][66][67]. The anions or the surface groups act as an outer-sphere bridge between the reactant and the electrode surface. Furthermore as seen in Figures 1(b) and 1(c), peroxide intermediate is detected at the ring electrode at 0.5 V and also below 0.3 V versus RHE, and the reason for this was explained in detail in an earlier publication [53,60]. Briefly, in alkaline media, the peak in ring current at 0.50 V is due to the carbon support and the increase in ring current below 0.3 V is due to reorientation of the water molecules at the electrode surface [53,60].
In light of this new evidence of involvement of outersphere electron transfer mechanism during ORR in alkaline media, it is noted that, while this mechanism allows the possibility of nonnoble metals to be used as electrode materials, it predominantly seems to yield the 2e − peroxide intermediate as the final product. So, for efficient 4e − conversion of O 2 , it is important to facilitate direct adsorption of desolvated molecular O 2 on the OH-free active site and avoid the precipitous reaction of solvated molecular O 2 with the OH-covered active site. In agreement with Bard [12], this implies the necessity to promote the electrocatalytic inner-sphere electron transfer mechanism for efficient 4e − reduction of oxygen. Figure 4 shows the ORR on Ru/C catalyst in O 2saturated 0.1 M NaOH and 0.1 M HClO 4 electrolytes recorded at 900 rpm. The ORR onset potential for Ru/C in 0.1 M HClO 4 is 0.78 V versus RHE, whereas in 0.1 M NaOH it is 0.9 V versus RHE. In the alkaline media, mixed kineticdiffusion region between 0.9 V and 0.5 V is ensued by a welldefined limiting current region below 0.5 V versus RHE. In the acidic electrolyte, ORR is kinetically controlled as indicated by the absence of any clear diffusion-limited current region even at very high overpotentials. As shown in Figure  4(b) at a kinetic current density of 0.1 mA cm −2 geo , Ru/C exhibits ∼125 mV lower overpotential in 0.1 M NaOH than in 0.1 M HClO 4 . Primary reasons to this lower overpotential in alkaline media are delineated here. As seen in Figure 4(c) between the potential regions of 0.7 V to 0.8 V, the shoulder in ring current in 0.1 M NaOH electrolyte is characteristic of the outer-sphere electron transfer mechanism on oxidecovered Ru/C catalyst in alkaline media. This outer-sphere reaction mechanism was depicted in (3)-(5) giving rise to the peroxide intermediate (HO 2 − ) as the final product at oxide-covered Ru sites. This (HO 2 − ) aq formed is detected at the ring electrode and appears as a shoulder in ring current between 0.7 V and 0.8 V as shown in Figure 4(c). In acidic media, peroxide detected is very minimal and about an order of magnitude lower than that in 0.1 M NaOH. It should be noted that, in 0.1 M NaOH, this outer-sphere electron transfer occurs only at oxide-covered Ru sites, whereas at oxide-free Ru sites direct molecular O 2 adsorption should take place leading to efficient 4e − reduction of O 2 to OH − via an inner-sphere electrocatalytic pathway. So, in alkaline media, a combination of both inner-sphere and outer-sphere electron transfer mechanisms is operative. The consequence of this outer-sphere electron transfer in alkaline media is that this mechanism leads to higher concentration of HO 2 − to be generated near the electrode surface, that is, the double layer; however, no evidence for such an outer-sphere reaction is observed in acidic media. Higher activity of HO 2 − effectively shifts the potential of the electrode from that of the O 2 /HO 2 − couple to that of the HO 2 − /OH − redox couple by carrying out HO 2 − reduction to OH − on oxide-free Ru sites. This is shown in Figure 4

ORR on Heat-Treated Carbon-Supported Ru Nanoparticles.
Alkaline Media: As is well known, the standard reduction potentials of the above reactions are well positive of the 4e − reduction of molecular O 2 in both acidic and alkaline electrolytes. So, from a thermodynamic perspective, any peroxide intermediate formed should be immediately reduced further. This is also kinetically true on ruthenium as shown in Figure  4(d). Half-wave potential (E 1/2 ) of HPRR on Ru/C in 0.1 M NaOH is 40 mV positive compared to that of the E 1/2 of ORR [68]. So the kinetics of the reaction in alkaline media favors the immediate reduction of any peroxide intermediate generated during ORR. This is also true in acidic media although the shape of the HPRR profile in acidic media on Ru/C requires more explanation. The onset potential of HPRR on Ru/C in 0.  (7) On the contrary, this excess parallel generation of H 2 O 2 via the outer-sphere reaction mechanism is not observed in acidic media and, hence, the higher overpotential for ORR in acidic media. In general, this can also be further extended to other catalyst systems that exhibit lower ORR overpotential in alkaline media compared to that in acidic electrolytes. The prospects of modifying the carbon-supported ruthenium nanoparticle catalyst by chalcogen groups (S/Se) in order to suppress oxide formation and promote the inner-sphere electrocatalytic process by facilitating direct O 2 adsorption was discussed recently as a separate treatise [70]. Figure 5(a) shows a comparison in dilute acidic and alkaline electrolytes of the ORR activity of FeTPP/C catalyst pyrolyzed at 800 • C. As observed in Figure  5(a), the onset potential for ORR in 0.1 M NaOH is 0.95 V versus RHE, whereas in 0.1 M HClO 4 electrolyte it is 0.80 V versus RHE. This 150 mV lower overpotential in alkaline media is clearly reflected over the entire mixed kinetic-diffusion region. In 0.1 M NaOH electrolyte, the mixed kinetic-diffusion region is ensued by a well-defined diffusion-limited region. In 0.1 M HClO 4 , no clear diffusionlimited region could be discerned, which is indicative of kinetic control in acidic media even at high overpotentials. At a potential of 0.80 V versus RHE, the ORR kinetic current density of FeTPP/C (pyrolyzed at 800 • C) is clearly four orders of magnitude higher in 0.1 M NaOH electrolyte than that in 0.1 M HClO 4 . For a given catalyst, the four orders of magnitude difference in kinetic activity between acidic and alkaline media are intriguing. Figure 5(b) shows the HPRR activity of FeTPP/C catalyst (pyrolyzed at 800 • C) in both acidic and alkaline media in comparison to the corresponding ORR polarization curves. The onset potential for peroxide reduction in 0.1 M HClO 4 is 0.84 V versus RHE, whereas in 0.1 M NaOH it is 1.01 V versus RHE. Besides this onset potential difference, in 0.1 M NaOH electrolyte, the mixed kinetic-diffusion region for peroxide reduction is more anodic compared to that of ORR in the same electrolyte which is then followed by a reasonably discernable diffusionlimited current density region. This clearly indicates that peroxide reduction in alkaline media is kinetically favored such that any peroxide intermediate formed during ORR in 0.1 M NaOH will be immediately reduced to the 4e − product. On the contrary, the reduction of hydrogen peroxide in acidic media is kinetically unfavorable due to weak binding of H 2 O 2 on the active site leading to its desorption into the bulk electrolyte and/or catalytic decomposition to molecular O 2 . This clearly demonstrates that stabilizing the peroxide intermediate on the active site is important in effectively carrying out 4e − ORR process, and the higher stability of peroxide intermediate on the active site leads to the so-called kinetic facility in alkaline media but not in acidic media. Given the pK a values for the first and second ionization of H 2 O 2 at 25 • C (pK 1 = 11.69 and pK 2 = ∼20), the predominant peroxide species for pH < 12 is H 2 O 2 whereas at pH > 12 it is HO 2 − [58]. This leads to the formation of a stable Lewis acid-base adduct in alkaline media (Fe 2+ -HO 2 − ) that enables complete 4e − transfer. On the contrary, the lack of stabilization of (Fe 2+ -H 2 O 2 ) adducts in acidic media leads to desorption of H 2 O 2 into the bulk electrolyte. Figure 5(c) shows the ring current due to peroxide oxidation measured during ORR in both acidic and alkaline electrolytes. Clearly, the peroxide yield during ORR on FeTPP/C (pyrolyzed at 800 • C) is higher in acidic media by an order of magnitude compared to that in alkaline media. This observation is opposite to what was observed on the Pt/C and Ru/C catalysts discussed previously. The onset potential for ORR in 0.1 M NaOH is 0.95 V versus RHE, whereas the corresponding peroxide oxidation current does not begin until 0.8 V. In 0.1 M HClO 4 , the onset potential for both ORR and peroxide oxidation is 0.8 V. This is further proof for the instability of peroxide intermediate on the active site in acidic media because the weak binding of the H 2 O 2 intermediate on the Fe 2+ active site facilitates its desorption and/or decomposition into the bulk electrolyte. This desorption of H 2 O 2 into the bulk electrolyte due to its weak binding on the active site is the primary source of peroxide detected at the ring in acidic media. At very high overpotentials (<0.3 V), the ORR process in acidic electrolyte becomes very efficient and consequently the peroxide detected decreases [71,72]. The ring current profile in 0.1 M NaOH is more complex. Briefly, in alkaline media, the peak in ring current at 0.50 V is due to the carbon support and the increase in ring current below 0.3 V is due to reorientation of the water molecules [60,73]. More importantly, a weak shoulder in the ring current in 0.1 M NaOH is observed between 0.6 V and 0.7 V. Also shown in Figure 5(c) is the ring current profile of Pt/C for comparison. As discussed previously, the ring current peak due to Pt/C around 0.80 to 0.90 V is the signature for the peroxide formation via outer-sphere electron transfer mechanism at -OH-covered Pt sites. On Fe-N x -based catalyst, only a weak shoulder in ring current at more negative potentials of 0.60 to 0.70 V is observed indicating that the outer-sphere electron transfer mechanism is significantly muted. This indicates the facilitation of direct adsorption of O 2 on the Fe 2+ active site. This is because the specifically adsorbed OH − species on the Fe 2+ site acts as a labile ligand that is easily displaced by the molecular O 2 . This is ensued by the successful 4e − reduction of O 2 with minimal occurrence of outer-sphere electron transfer mechanism.

Pyrolyzed Metal Macrocycle for Promoting Inner-Sphere Electron Transfer.
The nature of the active site obtained upon pyrolysis and the fundamental origin of the activity is investigated here in detail. Figure 6 [30,31,74]. As seen in the SWV profile, the peaks at 1.260 V in 0.1 M HClO 4 and 1.508 V in 0.1 M NaOH correspond to the one-electron redox transition involving the delocalized π-electron system in the macrocycle ligand [74]. As observed in the literature [30], CV of the heat-treated catalysts does not yield useful information, since in most cases after heat treatment the double-layer charging current overwhelms the Faradaic currents. In order to overcome this limitation, SWV was performed as shown in Figure 7 in order to understand the evolution of the active site in these catalysts with increasing heat treatment temperatures. Figures 7(a)  and 7(b) are divided into two panels each due to different potential ranges that are required for baseline correction of SWV profiles. As seen in Figure 7(a) in alkaline media, after 300 • C pyrolysis, the Fe 2+/3+ redox couple at 0.31 V and the ligand oxidation peak at 1.5 V are similar to the case of non-heat-treated catalyst. After heat treatment at 500 • C and 600 • C, the magnitude of the peak currents decrease significantly due to the sublimation of some fraction of the metal macrocycle complex. More importantly, the Fe 2+/3+ peak potentials have shifted to more anodic potentials. As seen in Panel 1 of Figure 7(a), after pyrolysis at 500 • C and 600 • C, the peak potential (E p ) of the Fe 2+/3+ couple has shifted to 0.405 V and 0.427 V, respectively. Correspondingly, the ligand oxidation peak at ∼1.5 V, shown in Panel 2 of Figure 7(a), decreases in magnitude significantly due to the sublimation and/or destruction of the macrocycle. The ligand oxidation peak is not observed after heat treatment at temperatures greater than 600 • C indicating that the delocalized π-electron system of the macrocycle does not survive such high temperatures on the carbon support. After 800 • C pyrolysis, Fe 2+/3+ couple has shifted more anodically to E p = 0.48 V. As shown in Figure 7(a) (Panel 2), the most interesting observation is that after pyrolysis at 600 • C a new shoulder begins to emerge at ∼1.2 V. This shoulder resolves into a clear peak after 800 • C pyrolysis with a peak potential of 1.25 V versus RHE in alkaline media. This new peak is attributed to an anodic shift in the Fe 2+/3+ redox couple based on the in situ XANES experimental results shown later (vide infra). This indicates the presence of two Fe 2+/3+ redox couples, one at a low potential (∼0.48 V) and another at a high potential (1.25 V). The anodic shift in Fe 2+/3+ redox peak potentials shown in Figure 7(a) (inset) clearly indicates that after 600 • C pyrolysis the metal center exists in two redox environments. Qualitatively, similar behavior was observed in acidic media as shown in Figure 7(b). (Note that the ligand oxidation peak in acidic media that occurs above 1 V versus RHE is not shown in Panel 2 of Figure 7(b).) It is interesting to note that after heat treatment at 600 • C the Fe 2+/3+ redox peak potential in 0.1 M HClO 4 was shifted to an anodic potential of only 0.80 V versus RHE compared to the 1.25 V in alkaline media. Thus, there are two reasons for the lower ORR overpotential for FeTPP/C catalyst in alkaline media: (1) the higher redox potential of the Fe 2+/3+ metal center in alkaline media and (2) the improved stability of the peroxide intermediate on the active site. This translates into efficient 4e − ORR reaction with lower overpotential in alkaline media compared to acidic media.

Delta-Mu (Δμ) Studies.
Both the extended X-ray absorption fine structure (EXAFS) and X-ray absorption near edge structure (XANES) being bulk averaged techniques overlook the critical electrochemical reactions occurring on the catalyst surface. Delta-mu (Δμ) is a surface sensitive, spectral subtraction technique, where the bulk structure of the catalyst is effectively removed leading to information on the nature and site specificity of the surface adsorbates [54,55,75,76]. Figure 8(a) shows the XANES region of FeTPP/C catalyst pyrolyzed at 300 • C taken at two different potentials of 0.1 V and 0.9 V versus RHE. At 0.1 V, the metal center exists in the reduced Fe 2+ state with no adsorbates (neither hydrides nor oxides) at the axial position, and immediate coordination environment is reminiscent of the iron(II) phthalocyanine complex, where the preedge peak is muted. At 0.9 V, the metal center is oxidized to Fe 3+ state with an oxygen atom at the axial position, and the metal coordination environment is similar to that of the original porphyrin complex, where the preedge Fe(1s → 3d) forbidden transition at 7112.5 eV is observed. The delta-mu spectra is obtained by subtracting the XANES regions according to the equation Δμ = μ(0.90 V)−μ(0.10 V). In the delta-mu spectra of Figure  8(a), the positive peak feature (boxed portion) indicates the difference in absorption probability at the preedge energy (7112.5 eV). This positive peak feature could be safely assigned as a signature for the existence of the metal center in a centrosymmetric environment undergoing a transition from Fe 2+ -N 4 coordination geometry at 0.10 V to O-Fe 3+ -N 4 coordination geometry at 0.90 V. Ensuing this positive peak is a steep negative dip featuring a split peak below 20 eV. This negative dip characterizes charge transfer from the metal center to the adsorbed oxygen species [77]. As shown in Figure 8(b), the XANES spectra of the 800 • C pyrolyzed catalyst is predominantly characteristic of metallic iron at both 0.10 V and 1.10 V that precludes proper analysis of the active site. However, careful analysis of the corresponding delta-mu spectra clearly indicates the positive peak feature at the preedge energy indicative of the fact that the active site is Fe 2+ -N 4 where the metal center is in a centrosymmetric environment which is mildly disrupted by the presence of an axial oxygen atom. This clearly indicates that the Fe 2+ metal center surrounded by four nitrogen atoms is the active site that binds oxygen in the axial position and the redox transition from Fe 3+ to Fe 2+ triggers oxygen adsorption according to the redox mechanism [78]. Figure 9 shows the theoretical delta-mu (Δμ t ) spectra obtained using FEFF8.0 code [56]. The structural models used are shown in the insets of Figure 9. Only the atoms encircled were used in the theoretical FEFF8.0 modeling. These theoretical spectra were calculated from Fe-N 4 -C x models derived from prior crystallographic data adjusted to the EXAFS fitting results according to the relation Δμ t = μ(O ads -Fe-N 4 -C)−μ(Fe-N 4 -C), where the oxide species (O ads or OH ads ) is in a specific binding site. In all cases, the positive peak feature at the preedge energy was observed only when the adsorbed oxygen atom was placed in the axial position of the metal center.
No successful theoretical delta-mu fits could be obtained for Fe-N coordination numbers less than four or for oxygen adsorption modes other than at the axial position. As shown in the inset of Figure 9(a), after 300 • C pyrolysis the immediate coordination environment of the original precursor porphyrin macrocycle is clearly retained. This corresponds to an FeN 4 C 12 cluster, where the metal center is coordinated to four nitrogen atoms and each nitrogen atom in turn bonded to two carbon atoms. Finally, inclusion of the four methine carbon bridges gives a tally of C 12 . Figure 9(b) shows the theoretical delta-mu spectrum that likely mimics the line shape of the experimental delta-mu shown in Figure 8(b) for the 800 • C heat-treated catalyst. As shown in the inset of Figure 9(b), the molecular cluster used to simulate the theoretical delta-mu spectrum consisted of FeN 4 C 10 . While compositionally this cluster is not very different from the 300 • C pyrolyzed sample, the immediate coordination environment of the metal active site after 800 • C heat treatment is found to be reminiscent of the crystallographic atomic defects such as the divacancy on the graphitic surfaces. This is schematically depicted in Figure  9(c) [79,80]. Atomic defects such as monovacancy and divacancy on microporous carbon and carbon nanotubes are known to thermodynamically exist or can be induced via various chemical or physical processes [81][82][83][84][85]. The presence of a monovacancy in carbon atom creates three dangling bonds whereas divacancies create four dangling bonds. These dangling bonds give rise to unsaturated valences that then become favorable for nitrogen doping [80]. Consequently, these nitrogen-doped sites constitute defective pockets for metal coordination. Such atomic vacancies are either already present on the graphite surfaces or can be created during the  heat treatment step [81,83]. During heat treatment under inert atmosphere, carbothermic reaction causes desorption of oxygen functional groups along with creation of vacancy defects [83]. It has also been observed earlier that heattreated Fe-N x catalysts showed higher activity when supported on carbon that was previously treated in concentrated inorganic acids [52]. Since such acid treatment steps lead to oxygen functional groups on carbon support, it is likely that this yields higher number of defective sites during subsequent heat treatment. These defective pockets are likely the favorable zones for anchoring FeN 4 active sites. Presumably, the low concentration of the defective sites limits the active site density. Therefore, increasing the defect density likely holds key to increasing the metal loading in this class of catalysts.
The following observations are made in an attempt to correlate the SWV and the delta-mu results. The lowpotential Fe 2+/3+ redox couple is characteristic of the original FeN 4 porphyrin moiety, whereas the high-potential Fe 2+/3+ redox couple is characteristic of the FeN 4 cluster found in the microporous carbon defect sites. At a pyrolysis temperature of 600 • C, both of these sites seem to coexist and the higher ORR activity obtained for heat treatment temperatures ≥600 • C is due to the high-potential Fe 2+/3+ redox couple seated in the defective pockets. Based on the above experimental results, the following reaction scheme is proposed for ORR in dilute alkaline media on heat-treated FeTPP/C catalyst. Similar set of reactions can be developed, mutatis mutandis, for dilute acidic media [22,71,72,86]. Equation While the Fe II valence state favors a square-planar tetracoordinate environment, the high potential of this redox reaction causes the OH − species to poison the active site at the axial position. This poisonous OH − species prevents direct molecular adsorption of O 2 on the active site. Further, the adsorbed OH − species mediates the 2e − outer-sphere electron transfer reduction of solvated O 2 molecule as shown in the reaction schemes (3)-(5) although this outer-sphere mechanism appears to be extremely weak in this class of pyrolyzed catalysts as explained previously. It should be noted that the case of outer-sphere mechanism does not arise in acidic media for reasons explained previously. Finally, the 4e − electrocatalytic inner-sphere electron transfer mechanism is shown in Figure 10, where the molecular O 2 displaces the OH − species and chemisorbs directly on the Fe II active site. Based on our experimental results, the electrocatalytic process in Figure 10 is shown to take place via a redox mechanism involving Fe II/III couple [78]. Once molecular O 2 adsorbs on the Fe II active site, the reaction proceeds to the ferrous-hydroperoxyl adduct via the superoxo and the ferric-hydroperoxyl states. The ferrous-hydroperoxyl adduct is very critical since its stability determines the product distribution. active site will lead to its desorption into the bulk electrolyte. As mentioned previously, the stable acid-base adduct formed in alkaline media stabilizes the ferrous-hydroperoxyl adduct. This ensures that the catalytic cycle shown in Figure 10 regenerates the active site via the formation of ferrichydroxyl species. However, in acidic media the analogous ferrous-hydrogen peroxide adduct Fe II -(OHOH) is unstable and causes desorption and/or disproportionation of the peroxide intermediate (H 2 O 2 ) into the bulk electrolyte. This leads to higher peroxide yield in acidic media as shown previously.

Conclusions
The case of ORR in alkaline media shows interesting mechanistic differences compared to acidic media. In acidic media, the presence of OH ad due to water activation blocks the direct adsorption of molecular O 2 on the active site.
In alkaline media, the presence of OH ad due to specific adsorption of hydroxide species not only blocks the O 2 adsorption but also promotes outer-sphere electron transfer mechanism to yield the 2e − peroxide intermediate as the product. The interaction between the O 2 ·(H 2 O) n cluster and the surface hydroxyl species causes certain nonspecificity to the identity of the underlying electrode metal. This nonspecificity is a result of kinetic differences in alkaline media that opens the gate to use a wide-range of nonnoble metals and their oxides as electrode materials, while in acidic media the absence of this outer-sphere mechanism necessitates the use of highly catalytic surfaces based on Pt.
On noble metals such as Pt, this nonspecificity in alkaline media is observed only in the oxide formation region, and depending on the extent of OH ad coverage both inner-and outer-sphere electron transfer mechanisms coexist in alkaline media. On nonnoble metals such as the first-row transition elements that are completely passivated by a layer oxide film upon immersion into the aqueous electrolyte, this outersphere electron transfer process is likely to be the dominant mechanism. While the outer-sphere mechanism primarily seems to yield peroxide intermediate as the final product, this could be beneficial in cases like Ru/C catalyst, where the peroxide intermediate generated in the outersphere process could be readsorbed on the oxide-free Ru site to reduce further to form hydroxide as the final product. However, in cases like nonnoble metal surfaces, where the electrode surface is incapable of reducing peroxide intermediate any further, only 2e − reduction products are likely to predominate in alkaline media. So, it is important to promote the inner-sphere electron transfer mechanism in alkaline media by facilitating the direct molecular  O 2 adsorption and promoting complete 4e − reduction process.
On catalyst surfaces such as the Fe-N 4 systems, the operation of redox mechanism allows the direct molecular O 2 adsorption on the Fe 2+ -N 4 active site and prevents the precipitous outer-sphere reaction of solvated O 2 molecule with the OH ad -covered metal site. This is due to the OH ad species acting as a labile ligand in the Fe-N 4 -based catalysts that helps to promote the 4e − inner-sphere electron transfer mechanism by efficiently reducing O 2 to OH − in alkaline media. Further, in the class of Fe-N 4 -based catalysts, it is important to perform ORR studies with a thin film of catalyst surface (≤25 μm) rather than thick catalyst films that are typically used in the literature, where the peroxide formed is completely quenched within the catalyst layer and hence not being detected at the ring electrode. This is especially true while Koutecky-Levich plots for calculating the number of electrons transferred since such calculations were derived for smooth electrode surfaces. Also, it is noted that if a catalyst is considered to perform 4e − oxygen reduction, then it is imperative to show the reduction of externally added hydrogen peroxide in the absence of O 2 . Finally, in conclusion, the presence of the outer-sphere electron transfer mechanism in alkaline media is a kinetic effect. The only thermodynamic advantage of performing ORR at high pH is the improved stability of the hydroperoxide anion intermediate on the active site that enables complete 4e − transfer. So, in order to stabilize the H 2 O 2 intermediate in acidic media, the use of dimeric active sites that simultaneously bind both oxygen atoms are required.