Mechanisms for the Production and Suppression of Hydrogen Peroxide at the Hydrogen Electrode in Proton Exchange Membrane Fuel Cells and Water Electrolyzers: Theoretical Considerations

Abstract

Hydrogen peroxide is inevitably produced at the hydrogen electrode in both the proton exchange membrane fuel cell (PEMFC) and the proton exchange membrane water electrolyzer (PEMWE) when platinum-based catalysts are used. This peroxide attacks and degrades the membrane, seriously limiting its lifetime. Here we review some of our previous efforts to suppress peroxide production using PtFe as a hydrogen evolution reaction (HER) catalyst and PtCo as a hydrogen oxidation reaction (HOR) catalyst. The mechanisms, which involve the chemical reaction of adsorbed hydrogen with oxygen, are examined using density functional theory. The onset of excess peroxide production at 0.1 V above the reversible potential has not been adequately explained thus far, and therefore a new mechanism is proposed here. This involves a unique reaction site including hydrogen adsorbed at (110) step edges adjacent to (111) terraces on the Pt surface, as well as on Pt alloys and other metals such as Rh and Ir. This mechanism helps explain the recent finding of the Wadayama group that Ir single crystal surfaces such as Ir(111) and Ir(110) produce little peroxide during the HOR. It also points the way toward the design of new catalysts for the hydrogen electrode that suppress peroxide production while retaining high HOR and HER activity.

1. Introduction

Proton exchange membrane fuel cells (PEMFCs) and water electrolyzers (PEMWEs) are both expected to play important roles in the highly anticipated “hydrogen economy” or “hydrogen society” [1,2]. In order to accelerate their widespread acceptance, however, their costs must be lowered. One strategy that can be effective in lowering costs is to lengthen the lifetime, which decreases the total cost of ownership (TCO) [3]. A key factor in limiting lifetime is degradation due to the undesired production of hydrogen peroxide at the negative (hydrogen) electrode, which can directly attack cell components, including the catalyst support and membrane. The problem arises because oxygen can diffuse easily through the membrane from the positive electrode, and, although it is partly reduced to water, it is also partly reduced to hydrogen peroxide in a two-electron process.

O₂ + 2H⁺ + 2e⁻ → H₂O₂ E⁰ = 0.685 V vs. SHE

(1)

O₂ can also react chemically with adsorbed hydrogen:

O₂ + 2H_ad → H₂O₂

(2)

which could occur either with O₂ already adsorbed (Langmuir–Hinshelwood mechanism) or with O₂ diffusing from the gas or liquid phase (Eley–Rideal mechanism). The generated peroxide can also decompose into two ·OH radicals in the presence of metallic ions, and these radicals are extremely reactive:

H₂O₂ → 2·OH

(3)

Thus far, we have identified two different specific pathways involving either underpotentially deposited hydrogen (H_UPD), which has been proposed to be H within the plane of the top layer of Pt atoms [4] or bridging H [5,6], or overpotentially deposited hydrogen (H_OPD), which has been generally agreed upon to be on-top H [4,5,6]:

Pathway 1—direct H_OPD

O₂ (chemisorbed) + H_OPD → HO_2,ad

(4)

O₂ (physisorbed) + H_OPD → HO_2,ad

(5)

HO_2,ad + H_OPD → H₂O₂ (free)

(6)

Pathway 2—activation of H_UPD

H_UPD → H_OPD

(7)

H_OPD + O₂ (end-on) → HO_2,ad

(8)

HO_2,ad + H_OPD → H₂O₂ (free)

Based on these pathways, it was hypothesized that it would be possible to minimize peroxide production by decreasing the amount of adsorbed H [7]. In earlier work, it had been found that Pt–M alloys, where M = Fe, Co or Ni, adsorb H less strongly than pure Pt [8]. Thus, PtFe was examined and found to suppress peroxide production during the hydrogen evolution reaction (HER) (Figure 1A and Figure S1), based on the lower coverage of both H_UPD and H_OPD, as calculated with density functional theory (DFT). It was found that the presence of O₂ interfered with the HER due to the stronger adsorption of O₂ than H, but this effect is relatively small due to the effective O₂ pressure, which is much less than 1 atm, and the low percentage of 2-electron reduction to H₂O₂. The most serious effect of the presence of O₂ is the membrane degradation caused by hydroxy radicals generated via H₂O₂ decomposition. Thus, the finding that the use of the Pt–Fe HER catalyst was effective in mitigating the degradation of the Nafion membrane and to prolong the cell operation is very important [7].

It was also realized that a lower coverage of H_UPD would provide more adjacent Pt sites for the bridged configuration of adsorbed O₂, which we have proposed to be an obligatory reactant in the 4-electron oxygen reduction reaction (ORR) [9,10]. This topic will also be examined in greater detail in the present work.

Figure 1. Experimental evidence for H₂O₂ production at low potentials in acid solution obtained with a channel-flow double-electrode system, in which O₂ can be reduced at the working electrode, with varying potential, and H₂O₂ is detected at the collecting electrode, which is set at a potential at which H₂O₂ is oxidized. (A) Potential dependence of H₂O₂ oxidation current density for various catalysts measured in O₂-saturated 0.1 M HClO₄ solution at 80 °C, Pt black, commercial Pt/C and PtxAL–PtFe/C, where AL is an abbreviation for “atomic layers,” [7]; (B) potential-dependent H₂O₂ oxidation current density, j(H₂O₂), for commercial Pt/C and Pt_xAL–PtCo/C, measured in 10% air/H₂-saturated 0.1 M HClO₄ at 80 °C; (C) hydrodynamic voltammograms for the ORR in O₂-saturated 0.1 M HClO₄ at Nafion-coated Pt_xAL–PtCo/C and commercial-Pt/C electrodes at 80 °C. The commercial catalyst, c–Pt/C in panel A was TEC10E50E (E-type carbon, 50 wt% Pt), while that in panels B and C was TEC10F30E (F-type carbon, 30 wt% Pt). All data were obtained at a flow rate of 111 cm s⁻¹ and a potential scan rate of 5 mV s⁻¹. The Pt collecting electrode was held at a potential of 1.4 V [7,11].

Figure 1. Experimental evidence for H₂O₂ production at low potentials in acid solution obtained with a channel-flow double-electrode system, in which O₂ can be reduced at the working electrode, with varying potential, and H₂O₂ is detected at the collecting electrode, which is set at a potential at which H₂O₂ is oxidized. (A) Potential dependence of H₂O₂ oxidation current density for various catalysts measured in O₂-saturated 0.1 M HClO₄ solution at 80 °C, Pt black, commercial Pt/C and PtxAL–PtFe/C, where AL is an abbreviation for “atomic layers,” [7]; (B) potential-dependent H₂O₂ oxidation current density, j(H₂O₂), for commercial Pt/C and Pt_xAL–PtCo/C, measured in 10% air/H₂-saturated 0.1 M HClO₄ at 80 °C; (C) hydrodynamic voltammograms for the ORR in O₂-saturated 0.1 M HClO₄ at Nafion-coated Pt_xAL–PtCo/C and commercial-Pt/C electrodes at 80 °C. The commercial catalyst, c–Pt/C in panel A was TEC10E50E (E-type carbon, 50 wt% Pt), while that in panels B and C was TEC10F30E (F-type carbon, 30 wt% Pt). All data were obtained at a flow rate of 111 cm s⁻¹ and a potential scan rate of 5 mV s⁻¹. The Pt collecting electrode was held at a potential of 1.4 V [7,11].

It was subsequently found that a Pt–Co catalyst for the hydrogen oxidation reaction (HOR) suppressed peroxide production at low potentials [11,12] (Figure 1B,C). As seen in Figure 1C, there are two potential regions with different hydrogen peroxide production behavior: one that commences at ca. 0.5 V, and a second with higher slope that commences at ca. 0.1 V. It was thought at that time that H_OPD formed at potentials as high as 0.1 V.

Figure 2 shows the DFT-calculated reaction profile for Pathway 1, reaction 4, in which O₂ is adsorbed at the (110) step on a model Pt(221) surface, which we have found to capture some of the characteristics of a Pt nanoparticle. Adsorbed H_OPD reacts to produce HO_2,ad as the first step in the overall peroxide production. The activation energy for this reaction on the Pt–Co(221) surface was calculated to be 0.60 eV, much larger than that on pure Pt(221), 0.28 eV. However, it was later realized that this pathway should be expected to be most important at potentials below 0.0 V vs. a reversible hydrogen electrode (RHE), where H_OPD, i.e., on-top H on (111) terraces, is predicted to exist on the Pt surface [13].

Figure 3 shows Pt(111) surfaces with full coverage (θ_H = 1) of bridging H in a hexagonal array, which was identified as H_UPD by Ishikawa and coworkers [5,6]. The open Pt atoms within each hexagon are the sites at which on-top H attaches at more negative potentials, as seen on the right (B). The latter were identified as H_OPD. At less negative potentials, at which H_OPD has not yet formed, these sites are open and can accommodate O₂ adsorption, as in Pathway 2, reaction 6, involving end-on (Pauling configuration) O₂ adsorption. This pathway can account for peroxide production in the potential range in which H_UPD exists on Pt(111), as clearly shown by Markovic et al. [14] and as can be observed at higher potentials in Figure 1C (upper panel). Thus, an H_UPD must be activated to the H_OPD state by breaking one Pt–H bond. This process is seen to require an activation energy of ca. 0.2 eV. However, even though the reaction is written in two steps, Equations (7) and (8), as seen in Figure 4, it is actually a concerted process, in which the two steps occur simultaneously. Interestingly, MacNaughton et al. found in situ evidence using X-ray absorption spectroscopy and XPS for a peroxo species on hydrogen-covered Pt(111) in contact with O₂ [15]. Their results indicated that the O–O bond was parallel to the surface, whereas the O–H bond was perpendicular to the Pt surface, similar to the configuration shown in Figure 4 for HO_2,ad.

Corresponding calculations for the PtCo system are still in progress. These are challenging due to the computational difficulty, since a 6 × 6 unit cell would have been required to accommodate the underlying Pt₃Co lattice. In addition, the spin introduced by Co leads to problems with scf convergence. Nevertheless, it can be expected that the activation energy would be significantly smaller due to the initially smaller H_UPD adsorption energy, as previously reported for Pt/Pt₃Co(221) [8].

As already noted above, peroxide production begins to increase at potentials as high as 0.5 V vs. RHE. This could be partly due to the effect of the carbon support, which contains quinone functional groups that are known to catalyze peroxide production [16]. However, this current is much smaller for the PtCo catalyst in this potential range, which would not be expected if the support surface were producing most of the peroxide. Therefore, the diminished peroxide production is most likely due to the lower H_UPD coverage. In previous work, we showed results for a Pt black catalyst [12]. The peroxide production on the latter surface was quite small, but this was due to the low surface area. In fact, it was shown that the peroxide production rate at 0.0 V was highly linear versus the electrochemical surface area, suggesting that the reaction, at least, at this potential, is not structure-sensitive, meaning that essentially every surface Pt atom can be an active site. Even on Pt black, there was a detectable peroxide oxidation current beginning at ca. 0.08 V [12]. As mentioned above, according to the H_OPD/H_UPD model developed by Ishikawa and coworkers, H_OPD should only appear at potentials below 0.0 V, because the H_UPD coverage is still not complete at that potential, reaching only ca. 0.66 [17,18]. Thus, we have sought to devise a new mechanism for peroxide production that involves the formation of the H_ad species that exists on (110) steps, which appears at ca. 0.1 V in the cyclic voltammogram in the absence of O₂ and which is the same as the precursor to the HER [19]. This will be the main subject of the present work.

Our proposed third pathway is thus:

Pathway 3—reduction by H_(110),step

O_2,ads + H_(110),step → HO_2,ads

(9)

HO_2,ads + H_(110),step → H₂O₂ (free)

(10)

This mechanism has two interesting aspects: (1) it accounts for the onset of peroxide production at 0.1 V, where the type of H_OPD that exists on (111) facets would not have significant coverage, and (2) it provides a unique geometry, in which an end-on adsorbed O₂ would be in very close proximity to the active H_(110),step, as shown later. We should also note here that our original rationale to consider H_OPD as the reactant (Pathway 1) was based on the well-known infrared signature (2080 cm⁻¹) of on-top hydrogen that is found in situ at ca. 0.1 V in acid solution on polycrystalline Pt [20,21]. However, as shown by Santana and Ishikawa, this spectral peak is actually not due to the type of on-top H found on (111) terraces (Figure 3). That peak was not found experimentally on Pt(111) single crystal surfaces [22,23] but was found on Pt(110), which includes essentially the same type of site as the (110) step under discussion here. Thus, even though this species is a kind of on-top H, by definition, it should be included as a form of H_UPD, since it appears at potentials above 0.0 V.

This mechanism can also help to explain results of Nakamori et al., in which they used Pt–Ru as a hydrogen catalyst [24], as well as results of non-Pt-based catalysts. Recently, the Wadayama group, following the early work of Neergat et al. [25]. found that Ir single crystal surfaces such as Ir(111) and Ir(110) produced much less peroxide than Pt(111) [26], due to their low affinity for adsorbed hydrogen, according to DFT calculations [27]. It is worth pointing out here that, even on high-quality Pt(111) single crystal surfaces, there is a certain number of (110) steps, which can exert a disproportionate effect. They also reported that Pt(111)–Ir(111) bimetallic surfaces were highly active for the HOR and produced little peroxide [28]. With these results in mind, we have carried out calculations for O₂ adsorption and H adsorption in the present work and we discuss the results in terms of the three pathways proposed above for peroxide production.

2. Results and Discussion

The unique geometry of the proposed active site is shown in Figure 5 for a Pt(221) surface, which includes a three-row-wide (111) terrace and a (110) step. As in our previous work, the configurations of the H_UPD as H_br and the active state of H_ad as H_(110),step are based on the work of Ishikawa and coworkers [5,6,19], which so far provides the model that best accounts for the existence of two distinct adsorbed forms of H on Pt(111) and their respective characteristics. In the present case, we have made the length of the rows three Pt atoms in order to better fit the honeycomb H_UPD (H_br) surface array. On the step, two on-top hydrogen atoms can adsorb at a single Pt atom, creating a “V” [19]. The lower H is very close to an end-on adsorbed O₂ molecule, similar to those pictured in Figure 3. If the latter rotates around its vertical axis, it comes into direct contact with the H_(110),edge and can react, as shown in the figure, with the energy of rotation being negligible, to produce the adsorbed HO₂ intermediate, which can again react with another H_(110),edge to produce free H₂O₂. We have not modeled that step in the present work, under the assumption that the energetics will be similar to that of the first step. The O₂ and H adsorption energies on the (111) terrace of the Pt(221) surface would be very similar to those on the (111) surface itself, the latter having been reported in [7]. To favor the end-on O₂ configuration, the H adsorption must be strong enough, and O₂ adsorption weak enough, so that the end-on O₂ molecule does not convert to the bridging configuration. In the present work, these values were determined to be −0.54 eV for H_br (H_UPD) and −0.45 eV for O_2,br on the (111) terraces of Pt(221). To test our hypothesis, we carried out similar calculations for Rh(221), Ir(221) Pt/Fe/Pt(221), simplified to Pt/Fe(221), as well as Pt/Rh(221) and Pt/Ir(221) (Figure 6, Figure 7 and Figure 8,

Table 1. H_br (H_UPD) and O_2,br adsorption energies (eV) calculated in the present work.

Surface	Hbr (HUPD)	O2,br	10H–O2, Initial	10H–O2, Final
Pt(221)	−0.54	−0.45	0.16	−0.45
Rh(221)	−0.45	−1.46	−0.99	−1.41
Ir(221)	−0.38	−1.26	−1.20	−0.76
Pt/Fe(221)	−0.23	−0.20	0.03	0.31
Pt/Rh(221)	−0.42	−0.53	−0.09	−0.89
Pt/Ir(221)	−0.44	−0.65	0.36	−0.14

and a comparison with other work in Table S1). For the Pt/Fe(221) surface, the H and O₂ adsorption energies were also balanced, and both were much smaller than those on Pt(221). For both of these surfaces, the initial end-on configuration is apparently stable, meaning that a geometrically optimized structure was obtained, even though the differential energy, i.e., just for O₂, was positive, i.e., repulsive interaction with the surface. The reaction (Equation (8)) led to a stable product, HO₂, in the case of Pt but an unstable product in the case of Pt/Fe. The initial state for H_UPD-covered Pt, with the end-on O₂ configuration, could be considered a metastable intermediate in the peroxide production, and thus its energy, 0.16 eV, is a kind of activation energy. This is significantly smaller than that, ca. 0.3 eV for Pathway 1 for Pt (Figure 2) and ca. 0.2 for Pathway 2 (Figure 4).

The Pt/Rh(221) and Pt/Ir(221) surfaces exhibited characteristics that were midway between those of Pt221) and Pt/Fe(221). The end-on O₂ was stable on both H_UPD-covered surfaces, more so than on Pt/Fe(221), on which the O₂ was not attached to the surface. The final HO₂ products for Pt/Fe, Pt/Rh and Pt/Ir were highly unstable, highly stable and slightly stable, respectively. These results by themselves would favor Pt/Rh(221) as a low-peroxide catalyst.

However, another important factor is the electrochemical potential at which the H_(110),edge appears. This value cannot be calculated directly but can be estimated based on the experimental value of 0.1 V vs. RHE for pure Pt, together with the calculated Fermi energy E_F of −5.879 eV, as we recently proposed for PtFeNi HER catalysts [29].

$$U_{MH}=E_{F,Pt,H\left(110\right)}-E_{F,MH}+0.1$$

(11)

Here, U_MH is the electrochemical potential of the H-covered metal surface, whether covered with H_UPD only or H_UPD + H_(110),edge. The calculated values are given in

Table 2. Electrochemical onset potentials U_MH calculated in the present work.

M-nH	Pt(221)	Rh(221)	Ir(221)	Pt/Fe(221)	Pt/Rh(221)	Pt/Ir(221)	10H–O2, Final
M-7H	0.118	−0.294	0.058	−0.032	0.086	0.083	−0.45
M-10H	0.100	−0.295	0.045	−0.014	0.034	0.044	−1.41

Note—7H include H_UPD; 10H include 7H_UPD, 1H_top and 2H_(110),edge.

. These values can only provide a rough indication of the appearance potentials for H_(110),edge, specifically, the values for 10H, but they do help to explain why Pt/Fe(221), even though it is seen to produce a very loosely bonded HO₂ species, does not predict substantial H₂O₂ production, as found in the experiment, since it is predicted to require a potential of −0.014 V, i.e., an advantage of 114 mV compared with Pt(221). Rh(221) is predicted to require a very negative potential, nearly −0.3 V, to form H_(110),edge, but, when covered with a monolayer of Pt, behaves much like pure Pt, but still with an advantage of ca. 66 mV, while Pt/Ir(221) is predicted to have an advantage of 56 mV.

In terms of actual adsorption energies, both Rh(111) [30] and Ir(111) [26] exhibit H adsorption behavior that is similar to that of Pt(111), with H_UPD adsorption being apparent in the cyclic voltammograms. The H adsorption energies for H_UPD on the (111) terraces of Rh(221) and Ir(221) were slightly less than those on Pt(221). The results for Rh(221) are consistent with the experimental measurements of Jerkiewicz and coworkers for polycrystalline Rh [4]. It was noted by that group that the hydrogen adsorption energies for all of the Pt-group metals are similar. The present DFT calculations are also in essential agreement with those of Greeley and Mavrikakis [31], as well as those of Liu et al. for various Ir single crystal surfaces [27]. Thus, even though the H adsorption energies for the (111) terraces are smaller than those on Pt(111) terraces, as already noted, this is not the only factor involved in the decreased H₂O₂ production. The O₂ adsorption must also be considered.

The stronger affinity of Ir surfaces for oxygen in comparison with Pt has been recognized, although it is difficult to find data on molecular O₂ adsorption (see the DFT calculations of Klyukin et al.) [32]. The same situation applies for Rh surfaces. Comelli et al. have stated that molecular O₂ is only found on Rh surfaces that already are partially covered with oxygen atoms [33]. Indeed, until recently, it has been difficult to determine with certainty the molecular O₂ adsorption energy on Pt surfaces, due to the ability of steps to catalyze its dissociation [34,35]. Recent DFT studies rarely include molecular O₂ adsorption; however, our result, −0.45 eV, is in agreement with that of Ou et al. for Pt(111) [36], but slightly smaller than that of Eichler and Hafner, −0.68 eV [37].

The previously mentioned results of Nakamori et al. for Pt–Ru are also consistent with the present approach, since Ru is well-known to have a higher affinity for oxygenated species than pure Pt. So-called “bifunctional” Pt–Ru catalysts were developed by Watanabe and Motoo for methanol oxidation [38] and CO oxidation [39], and, making use of this principle, Ishikawa and coworkers used DFT calculations to explain the mechanism in detail [40].

It is clear why the end-on configuration for O₂ is stable for H-covered Pt(221), i.e., it is surrounded by H_UPD (H_br), and it also clear why it is unstable on Rh(221) and Ir(221), i.e., the bridged configuration O_2,br is much more stable. On Pt/Fe(221), the calculations showed that the end-on O₂ was so weakly adsorbed that it desorbed spontaneously. The very strong O₂ adsorption on the Rh(221) and Ir(221) surfaces also ensures that the product of the reaction (HO_2,ad) will be strongly adsorbed at the step (Figure 6). When these metals are covered with Pt, however, their behavior becomes much more like that of pure Pt.

Also of interest are the recent results of Hayashi et al. in which they report results for Pt–Ir bimetallic surfaces [28]. These authors reported that up to a 1/3 monolayer of Pt on Ir(111) was effective in suppressing peroxide production while providing high HOR activity. This type of surface can be seen to provide the type of active site that we are proposing in Figure 5, i.e., the Pt(110) step, at which the HOR and HER are highly active, together with a lower terrace, Ir(111), on which O₂ adsorption is strong and favors the bridging configuration. We have not modeled this system thus far because it requires a larger atomic model, but we can nevertheless understand essentially how it works. The edge of the Pt islands would behave like the steps in the Pt/Ir(221) system modeled in this work, with a more negative U_MH than pure Pt, and we know that O_2,br adsorbs much more strongly on Ir(111) than on Pt(111) terraces, so that end-on O₂ adsorption would not be expected to be stable.

In order to produce peroxide, the O–O bond must remain intact, and therefore, the O₂ adsorption energy must be small enough that the bond is not broken. Siahrostami et al. have examined this problem from the viewpoint of HO₂ adsorption on Pt-based alloys and other alloys, and have found that there is a volcano relationship between the free energy of HO₂ adsorption and the activity for peroxide production [41]. They found that the Pt–Hg surface was among the most efficient for peroxide production. Other, more recent efforts along these lines have been reviewed [42].

The novel aspect of the present research is that it involves very specific mechanisms and reactive sites that can explain most of the available experimental results, depending upon the potential range. In the range from 0.3 V down to 0.1 V, important for the HOR at high current densities, Pathway 2 is considered most favorable, because it involves H_UPD, which begins to adsorb on Pt and Pt-based catalysts in this potential range. In the range from 0.1 to 0.0 V, important for the HOR at low current densities and open circuit, Pathway 3 is favorable, because it involves the H_(110)edge species, which starts to form at 0.1 on pure Pt. It also explains why Pt-alloys such as Pt–Fe and Pt–Co produce less H₂O₂, since the formation of H_(110)edge is shifted to less positive potentials, as discussed above and as shown by Sheng et al. [43]. For potentials less than 0.0 V, important for the HER, Pathway 1 is favorable, since it involves the presence of H_OPD on (111) terraces, which become prevalent in this range.

At this point, we would like to summarize the three pathways examined in the present work for H₂O₂ production (

Table 3. H₂O₂ production pathways discussed in the present work.

Pathway	H	O2	Potential a	Onset	Remedy b
1	OPD	(110) edge	<0.0 V	sudden	Decr. θH
2	UPD	End-on	<0.5 V	gradual	Decr. θH, incr. Ead,O2
3	(110) edge	End-on	<0.1 V	sudden	Decr. θH, incr. Ead,O2

^a These values are roughly equivalent to the U_MH values for pure Pt and can be shifted negatively by alloying (see Table 2). ^b θ_H can be decreased both by decreasing E_ad,H and making U_MH more negative.

). Thus, the key factors that help to minimize H₂O₂ production while maintaining high HOR or HER activity are as follows:

• The O_2,br adsorption should be stronger than that of H_UPD, so that the 4-electron ORR will be favored vs. the 2-electron process to produce H₂O₂ (Pathways 2 and 3);
• Even if #1 is not satisfied, θ_H can be decreased significantly, both by decreasing the adsorption energy E_ad,H and the appearance potentials U_MHUPD (Pathway 2) and U_MH(110),edge (Pathway 3) via alloying;
• The coverage of H_UPD is also important, because high coverage tends to stabilize the end-on adsorption of O₂ on the (111) terraces. These coverages are dependent upon the U_MHUPD values, which are very close to the U_MH(110),edge values;
• The HOR and HER activities are favored by decreased H adsorption energies at both the (110) step and the (111) terrace, with the effect of the latter being particularly apparent for the HOR [8];
• Based on the above guidelines, PtFe, PtRh and PtIr all are predicted to have high activity for the HOR and HER with low activity for H₂O₂ production. The Pt monolayer on Ir(111) model is also attractive if it is possible to prepare in high area form;
• Of course, the final choice of catalyst also requires consideration of cost, durability and ability to prepare a suitable material. The durability of both Rh and Ir is expected to be superior to that of Fe as an alloying element.

3. DFT Calculations

The calculations were carried out with the DMol³ package (BIOVIA, Materials Studio Visualizer, versions 2021 and 2023) [44]. The geometric optimizations for periodic boundary conditions were carried out with the hardness-conserving semilocal pseudopotential [45] and the PBE functional [46]. The geometry optimizations were carried out with fine-quality settings (convergence criteria 1 × 10⁻⁵ Ha, maximum force 0.002 Ha/Å, maximum displacement 0.005 Å with a numeric quality basis set with polarization functions), and an scf convergence criterion of 1 × 10⁻⁶ Ha, with energy calculations being carried out with an all-electron relativistic basis. To facilitate scf convergence, a minimum kinetic energy was applied to the electrons (thermal smearing), 0.0005 Ha, which is 1/10 of the usual value of 0.005 Ha. These low values were used in order to facilitate scf convergence as well as to avoid unwanted effects of excessive electron energy, as described by Basiuk et al. [47]. In the (221) models, only the coordinates of the bottom layer were fixed.

The H adsorption energy was calculated with Equation (12):

$$∆E_{nH}=\left({\displaystyle \frac{1}{n}}\right)\left[E_{Pt-\left(n/2\right)H_2}-\left(E_{Pt}+E_{\left(n/2\right)H_2}\right)\right]$$

(12)

Here, n is the number of H atoms. This yields the adsorption energy per H atom and can be compared directly with the experimental enthalpy values reported in the literature, for example, those of Jerkiewicz, which are designated as $∆H_{ads}^0\left(H_{chem}\right)$ (see Table S1) [4]. It should be noted that, in previous publications, we have reported H₂ adsorption values, which are larger than those reported in the present work by a factor of two. The earlier approach had been adopted in order to match the convention in reporting the adsorption of molecular O₂:

$$∆E_{O_2}=\left[E_{Pt-O_2}-\left(E_{Pt}+E_{O_2}\right)\right]$$

(13)

The differential adsorption energy of O₂ on a surface already populated with adsorbed H was calculated as follows:

$$∆E_{{nH-O}_2}=\left[E_{Pt-nH-O_2}-\left(E_{Pt-nH}+E_{O_2}\right)\right]$$

(14)

4. Conclusions

The two previous models for peroxide production at the hydrogen electrode in PEMFCs and PEMWEs are reviewed: Pathway 1, in which O₂ is adsorbed at Pt(110) step edges and reacts with H_OPD adsorbed on the lower (111) terrace, and Pathway 2, in which H_UPD is activated thermally to H_OPD and reacts with end-on adsorbed O₂. In order to explain the acceleration of peroxide production below 0.1 V, a third pathway is newly proposed, in which the H_(110),edge species reacts with end-on-adsorbed O₂ on the lower (111) terrace. The new pathway is consistent with published results for pure Pt, Pt-alloy, pure Ir and Pt–Ir bimetallic surfaces. It can also be used to design new catalysts for the HOR and HER that suppress peroxide production and thus prolong the life of the proton-exchange membrane.

The mechanism and unique reactive site can be thought of as further examples of bifunctional catalysis, as already mentioned, or, more accurately, adjacent dual-site catalysis, in which two adjacent catalytic sites perform complementary functions, although in the present case we have sought to inhibit rather than catalyze the peroxide production reaction. Previous examples include the studies of methanol oxidation [38] and CO oxidation on Pt–Ru bimetallic surfaces [39], CO-tolerant HOR catalysts [8], the HER on Pt–Co catalysts [11], the HOR on Pt–Ir bimetallic surfaces [28], and the OER on FeCoS_x surfaces [48]. This emerging field has been reviewed recently [49].