A discussion on the possible involvement of singlet oxygen in oxygen electrocatalysis

The dioxygen molecule has two bound states, singlet and triplet, which are different in energy, lifetime, and reactivity. In the context of oxygen electrocatalysis as applied to fuel cells and water splitting the involved O2 is typically considered to be exclusively in its triplet ground state. However, applying spin-conservation rules for the transformation between triplet O2 and singlet OH−/H2O reaction intermediates predicts an additional free energy barrier associated with the required spin flip. As a result, for conditions under which both can form, the formation of triplet dioxygen from the singlet OH−/H2O might be slower than the formation of singlet O2. Correspondingly, singlet O2 might be more active than triplet O2 in the oxygen reduction reaction. Here, we discuss the possible existence and influence of singlet oxygen in oxygen electrocatalysis. Some perspectives for studying singlet oxygen in oxygen electrocatalysis are also provided.

Oxygen electrocatalytic reactions, including the oxygen evolution reaction (OER) and the oxygen reduction reaction (ORR), have attracted great attentions in recent years [1][2][3]. Equation (1) shows the OER and ORR in acidic environment and equation (2) shows them in alkaline OER and ORR are, respectively, rate determining for water electrolysis and hydrogen fuel cells. In recent years, new insights into the reaction mechanisms have been developed for the design of better oxygen electrocatalysts. For example, (a) some highly active catalysts have been identified as pre-catalysts due to surface reconstruction under the OER condition. The observed high OER activity is given by the reconstructed surface [4][5][6]. (b) The observation of lattice oxygen involvement leads to the setup of a lattice oxygen mechanism (LOM) in addition to the conventional adsorbate evolution mechanism (AEM). LOM routes enable the direct O-O coupling, which breaks the adsorption-energy scaling relation between adsorbed OH and OOH intermediates, and thus better OER performance [7,8]. (c) There is a close relationship between the spin state of active metal cations and catalytic activity [9][10][11][12]. On one hand, the adsorption free energies of key intermediates are highly sensitive to the spin state of cations at the surface [10]. On the other hand, charge transport may be more efficient in catalysts with spin channels [11,12].
From an orbital point of view, the oxygen molecules have two spin states (figure 1(a)), the triplet state ( 3 O 2 , ground state) and singlet state ( 1 O 2 , metastable state). These figures are approximate 'cartoons' that do not adequately represent the required antisymmetric wavefunctions [13,14]. In the triplet ground state ( 3 Σ g − ), there are two unpaired and spin-parallel electrons in π * orbitals, which makes the molecule paramagnetic. A reduced electronic repulsion stabilizes this open-shell orbital configuration due to quantum spin exchange interactions (QSEI) [15]. In singlet oxygen, all the energy levels are closed shells, where the frontier electrons can share the same space (orbital) and instead, the Coulomb attractions are optimized (see the Fermi heap in the center of the 1 ∆ g O 2 molecule in figure 1(b) right). According to the difference in the occupancy of π * orbitals, singlet oxygen can be further divided into two types, 1 ∆ g and 1 Σ g + (figure 1(a)). The 1 Σ g + singlet state oxygen is 0.65 eV higher in energy than the 1 ∆ g state) and can be quickly (in microseconds or nanoseconds) converted to the low-energy 1 ∆ g state. The 1 Σ g + singlet state oxygen is always shorter-lived than the 1 ∆ g singlet state oxygen [16]. Therefore, unless otherwise stated, all the singlet oxygen mentioned in the following part refers to the 1 ∆ g state oxygen. Compared to the ground state triplet oxygen, excited singlet oxygen 1 ∆ g is still 0.98 eV higher in energy and is metastable, which can exist for about 1 h in the gas phase [17]. It can be easily transformed into the triplet oxygen in solvents within milliseconds and microseconds [17]. The reactivity of singlet oxygen is different from that of triplet oxygen [18]. Singlet oxygen is regarded as one of the 'reactive oxygen species' [19].
As the energy of triplet oxygen is 0.98 eV lower than that of singlet oxygen, in most cases, it is considered that only the triplet O 2 is involved in oxygen electrocatalysis. The reversible potential of the four-electron transferred triplet oxygen generation is 1.23 V, while the reversible potential of singlet oxygen evolution can be calculated to be 1.23 + 0.98/4 = 1.475 V based on the Gibbs free energy ∆G = −nEF. Since the potential gap between anode and cathode could be up to ∼2 V in water electrolysis, it is thus energetically possible to produce singlet oxygen during the OER. According to the spin-conservation [20], the reaction rate might slow down if the spin polarization (i.e. the number of spins 'up' minus the number of spins 'down') of the products differs from that of the reactants [21]. For instance, if electron transfer must wait for an incoherent step, a slow spin relaxation mechanism occurs via a collision of paths that are against the foreseen spin-rules for optimal electrocatalysis [22,23]. In this regard, singlet oxygen is kinetically less challenging, more desirable for OER and ORR in reaction kinetics, because electron transfers are easily all coherent in the transformation between singlet oxygen and singlet OH − /H 2 O. Considering the highly active feature of singlet oxygen, attentions need to be paid to detecting its existence and exploring its possible functions in oxygen electrocatalysis.
Some experiments have shown that not all oxygen generated by electrochemical reactions is in the triplet state. Recently, the generation of singlet oxygen was detected in Li (Na)-oxygen batteries [24,25] and Li-ion batteries [26]. Specifically, 4% of the oxygen generated in the cathode of a Li-oxygen battery is singlet oxygen, which increases to 6% in the presence of trace water [24]. The singlet oxygen not only decays to triplet oxygen but also reacts with electrolyte to form reactive intermediates, such as ROOH, R * , and ROO * [24]. The active singlet oxygen has been demonstrated as an oxidant to react with the carbon-contained electrolytes. The parasitic reaction products (i.e. Li 2 CO 3 , Li acetate, and Li formate) severely limit the cycle life of the Li-oxygen batteries [24]. Generation of singlet oxygen in Na-oxygen batteries has also been shown to negatively affect stability [25]. In a study of Li-ion batteries Wandt et al reported direct generation of singlet oxygen from the lattice of layered LiNi x Co y Mn 2 O 2 when charging it to a high state of charge [26]. The highly active singlet oxygen was proved to be responsible for the electrolyte degradation in this Li-ion battery. One possible route for the release of singlet oxygen in metal-air and Li-ion batteries, as proposed by [27]. On the basis of the Marcus theory analysis, they predicted that the reaction rate constant for the production of 1 O 2 is three orders of magnitude higher than that of the 3 O 2 , suggesting nonnegligible possibilities of singlet oxygen production. Furthermore, the production of singlet oxygen has also been found in photocatalytic water splitting [28,29], and its generation is correlated to the oxidation of superoxide anion radical ( • O 2 − ), which needs additional energy inputs (compared to the direct generation of triplet oxygen) [30]. Generally, a minimum energy of 237.24 kJ mol −1 is needed for triplet OER [31]. As the singlet oxygen is 94.29 kJ mol −1 higher in energy than that of triplet oxygen, when the input energy is higher than 331.54 kJ mol −1 , the production of singlet oxygen is thermodynamically possible in electrocatalytic water splitting.
While the formation of singlet O 2 is energetically possible at a high overpotential employed for OER, no direct observation of singlet oxygen has been reported to date. However, some studies experimentally proved that some of the reaction intermediates have singlet character. By studying the Co 4 O 4 cubane OER catalyst, Nocera et al proposed that the in situ formed antiferromagnetic (AFM) coupled Co(IV) 2 dimer may drive singlet O-O bond formation by conveying singlet character to the oxygen radicals to promote direct coupling and bond formation with no spin barrier (figure 2) [32]. The preferable production of triplet oxygen on ferromagnetic (FM) catalysts has been successfully demonstrated by Garcés-Pineda et al [33]. The magnetic field favors the parallel alignment of oxygen radicals on the FM catalysts during the formation of the triplet O-O bond. In several of our co-author's works, it was observed that catalysts with dominant FM interactions, between the active metal centers with unpaired electrons, facilitate the 3 O 2 formation. These works strongly indicate that OER activity is spin-dependent. A hypothesis may be made that the catalysts with dominant AFM interactions (cooperative interactions) can favor the formation of 1 O 2 to avoid the spin barrier.
The electrochemical triplet ORR activity can be related to magnetic property of the electrocatalysts [34]. Jose et al performed comprehensive density functional theory (DFT) simulation to unveil the correlation between the ORR activity and magnetic property of the Pt 3 Co materials (figure 3) [34] Due to the QSEI character of the open-shell FM and AFM Pt 3 Co catalysts, both FM and AFM Pt 3 Co catalysts have milder chemisorption energies than the non-magnetic Pt 3 Co alloy. This milder chemisorption energy is beneficial for the electron delocalization and thus improving the ORR activity.
For electrochemical ORR, there is no experimental report involving singlet oxygen to date. This is because the lifetime of singlet oxygen is milliseconds or even microseconds in solution. It is difficult to experimentally compare the influences of triplet and singlet oxygen to the ORR performance. In contrast, theoretical studies have predicted that singlet oxygen may be easier to be reduced as compared to triplet  oxygen. Using DFT calculations, Nazmutdinov et al have theoretically analyzed the spin effects in ORR [35]. The rate-determining step (RDS) for ORR is calculated to be the first electron transfer to form the adsorbed O 2 − on the surface of Pt. The energy of Coulomb repulsion (U) of electrons on the two different oxygen orbitals was compared. During the reduction of triplet oxygen, the first electron is transferred from an electrode to a half-occupied π * orbital (U > 0); while in the case of singlet oxygen, the first electron is transferred to an empty π * orbital (U = 0). Therefore, relatively, a more negative potential is needed for the first electron transfer from electrode to triplet oxygen. It was therefore raised that singlet oxygen should facilitate the RDS and reduce the overpotential of ORR, versus triplet O 2 .
Inspired by above pioneering works, we propose a possible mechanism for triplet and singlet oxygen generation by considering the spin direction of the four transferred electrons in OER. Our discussions are based on the recognized AEM [3,36,37] and LOM [7,8] routes. AEM path has two subtypes, namely, the Eley-Rideal (ER)-type ( figure 4(a)) and the Langmuir-Hinshelwood (LH)-type ( figure 4(b)) [37]. For OER in base, each of the four steps involves the orbital overlap between the lone pair on the OH − and an empty orbital on the active site. Due to the strong spatial overlap between metal e g and O 2p orbitals (for example, 3d z 2 and 2p z orbitals) in octahedral site, the oxygen intermediates are adsorbed on the metal site through the formation of σ-like bond between O 2p and metal e g during the entire process [38]. The transferred spin direction is then dominated by the partially occupied metal t 2g orbitals via exchange interactions through a space-separated π-type tunneling between t 2g and p x /p y orbitals [15]. According to previous study [11], it is more favorable to transfer electrons with opposite spin direction to that of the unpaired t 2g electron. In this way, the spin direction of the 1st transferred electron should be antiparallel to that of the unpaired t 2g electron.
Through the ER-type route, an unoccupied orbital on the adsorbed oxygen is favorable for the coupling of the next OH − . Thus, the 2nd transferred electron should be the one in the same orbital with opposite spin to that of the 1st transferred electron ( figure 4(a)). Subsequently, with the transition of adsorbed O to adsorbed OOH, the electron in the fully occupied π * orbitals could transfer in the 3rd elementary step. Similar to the 1st transferred electron, the spin direction of the 3rd transferred electron is also dominated by the spin direction of unpaired t 2g electrons, which means the 1st and 3rd transferred electrons possess the same spin direction. In the last step, which is the spin selection step, if the spin state of the 4th transferred electron is the same as that in the 3rd step, 3 O 2 will form; otherwise, 1 O 2 will form. In this regard, catalysts with unpaired t 2g electrons as well as chiral molecules with spin selectivity should be favorable for the transfer of electrons with a specific spin direction, thus facilitating the release of 3 O 2 [39].
The LH-type route occurs when the distance between two neighboring metal sites is relatively short ( figure 4(b)). For the successful formation of a σ O-O bond between the two adjacently adsorbed OH − , the spin direction of the 1st and 2nd electrons should be opposite, indicating an AFM configuration of metal sites is preferable. The 2nd electron transfer is the spin selection step. The spin directions of the 3rd and 4th transferred electrons are also dominated by the AFM configuration. In the LH-type route, the generation of 1 O 2 might be more favorable ( figure 4(b)).
LOM may occur when the O p-band of lattice oxygen is higher than the d-band center of metal [40]. Oxygen molecules generated through LOM path have two possible origins: (a) both two oxygen atoms of the produced dioxygen molecule are from the lattice of the oxides; and (b) one from lattice oxygen while the other from the electrolyte (figures 5(a) and (b)). In the former case ( figure 5(a)), to form the σ O-O bond, each of the two adjacent oxygens needs a half-filled 2p orbital and the two unpaired electrons need to have opposite spins. Thus, two spin-antiparallel electrons are extracted in the 1st and 2nd steps. The spin direction of the 3rd transferred electron should be antiparallel to the unpaired t 2g electron of the center metal site for optimizing the quantum spin-exchange interactions with the metal site. The last step determines whether the generated oxygen molecule is a singlet or triplet state, which is similar to the case of the ER-type route. However, if the 1st and 2nd transferred electrons possess the same spin direction, it is difficult for the σ O-O bond formation. Under this circumstance, the latter case will take place ( figure 5(b)), wherein one lattice oxygen couples an adsorbed OH [7], instead of another lattice oxygen, to form an O-O bond. In this one-lattice-oxygen-participated route, similar to the case of the ER-type route, the generation of 3 O 2 or 1 O 2 is also dependent on the spin direction of the 4th transferred electron. Thus, from the spin-selective point of view, OER via the LOM route could also selectively produce 3 O 2 or 1 O 2 depending on the spin direction of 4th electron transferred.
Likewise, the electrocatalytic triplet and singlet O 2 reduction reactions (denoted as 3 O 2 RR and 1 O 2 RR) are compared by considering the spin direction of the four transferred electrons. The ORR mechanism can be regarded as a reverse OER route [41]. Here, the reverse ER-type mechanism is selected as an example to show the electron transfer in the 3 O 2 RR (figure 6(a)) and 1 O 2 RR ( figure 6(b)). In the case of 3 O 2 RR, due to the two unpaired and spin-parallel electrons in π * orbitals of 3 O 2 , the 1st and 2nd transferred electrons should be anti-parallel to the existing electrons. Assuming the electrons in the π * orbitals of 3 O 2 are spin up, then the 1st and 2nd electrons should be spin down. After accepting the 1st and 2nd electrons, the two π * orbitals of adsorbed OOH intermediates are fully occupied, so that the 3rd electron could only fill the higher-energy σ * orbital of adsorbed OOH intermediates. The electron filling in the higher-energy anti-bonding makes the O-O bond unstable, so that the adsorbed OOH transforms into adsorbed O intermediates and OH − . The 3rd electron should fill the 2p orbital of adsorbed O intermediates, while its spin is most probably anti-parallel to the unpaired t 2g electron. In the end, the last spin-down electron transfers to pair with the 3rd electron. As for 1 O 2 RR, because of the existing of an empty π * orbital, the transfer of the 1st electron faces no Coulomb repulsion. The spin direction of the 1st electron is most probably anti-parallel to the unpaired t 2g electron, which is under the same determining factor as mentioned in the OER part. Then the 2nd to the last electron transfer should be the same to corresponding steps of 3 O 2 RR.
Summarizing, we have proposed several pathways for singlet oxygen generation during OER within both the AEM and LOM reaction mechanisms. LH-type route on AFM-configuration catalysts most likely favor (a) The detection of singlet oxygen during OER. Singlet oxygen in the gas phase with a pressure of 104.3 kPa at 296 K is relatively long-lived (72 min) [42], but contact with a liquid reduces its lifetime to milliseconds or even microseconds. The transition of 1 O 2 to the ground state gives the emission of photons at 1270 or 633 nm, which provides an optical method to detect the singlet oxygen which has been used in lithium ion batteries [26]. The 1270 nm radiation (phosphorescence) is caused by the spin/parity-forbidden radiative transition from 1 O 2 to 3 O 2 [43], and the 633 nm emission (luminescence) is due to the formation of a 1 O 2 dimer followed by the spin-allowed transition to the 3 O 2 [26]. For quantifying the singlet oxygen, the 633 nm emission is more sensitive than 1270 nm radiation, even though the yield of 1270 nm is generally much higher than that of 633 nm emission, which depends on the square of the steady-state 1 O 2 concentration [26]. Another route is to use chemical probes. For example, the fluorophore and trap probes will become fluorescing or electron paramagnetic resonance active, respectively, upon reacting with the singlet oxygen [17]. Considering the possible oxidation of chemical probes under the OER conditions, detecting the 1270 nm and 633 nm emissions originated from the transition of singlet oxygen to triplet oxygen may be more promising. One should keep in mind that these two radiative emissions' yields are very small [26], which brings challenges to the detection of singlet oxygen. (b) ORR with singlet oxygen. Compared to reducing triplet oxygen, reducing singlet oxygen is expected to facilitate the first electron transfer in ORR. However, the singlet oxygen has very limited lifetime when contacting a liquid like water. It is thus challenging to provide a singlet oxygen saturated electrolyte for ORR. According to previous report [17], the lifetime of singlet oxygen in D 2 O electrolyte is relatively longer (∼67 µs) than in H 2 O (∼3.5 µs). It may be possible to use a properly designed electrochemical cell with D 2 O electrolyte to investigate 1 O 2 reduction. In addition, an impinging jet flow cell with a rotating valve for accurately controlling oxygen gas dosing might be another possible tool to test the ORR with singlet oxygen (figure 7) [44]. In order to enable 1 O 2 participated ORR before its decay in the electrolyte, the pipeline length divided by the electrolyte flow rate should be less than 67 µs. The 1 O 2 RR and 3 O 2 RR activity can be evaluated by cyclic voltammetry measurements. Conventionally, ORR kinetic activity is measured by rotating disk electrode (RDE), which allows us to conduct mass-transport-correction to obtain the ORR kinetic current given by the Koutecky-Levich equation [45]. However, the recommended setup (figure 7) for studying 1 [46]. Properly using this strategy, the reaction mechanisms and free energy change for singlet and triplet oxygen production on a given catalyst can be systematically predicted. It may also facilitate the design of appropriate catalysts for generation of either 3 O 2 or 1 O 2 . Here, the spin-polarized DFT calculations is performed on Pt (111) to compare the relative adsorption energy of singlet and triplet oxygen (figure 8). The adsorption energy of the 1st step for 1 O 2 RR is approximately 1.1 eV lower than that of the 3 O 2 RR, indicating better thermodynamics for singlet oxygen capture. Figure 7. Sketch of the impinging jet flow system for the study of singlet oxygen reduction reaction. WE, RE, and CE refer to the working electrode, reference electrode, and counter electrode, respectively. For a fair comparison, the pipeline's length for 1 O2 and 3 O2-saturated electrolyte should keep identical, so does the electrolyte's speed. Adapted from [44], Copyright (1998), with permission from Elsevier. (d) The opportunities and challenges brought by the highly active 1 O 2 . Considering the distinct properties of singlet oxygen, strategies to make full use of its advantage and to avoid its negative effects can be proposed. On one hand, the generation of 1 O 2 with high reactivity may facilitate some parasitic oxidation reactions (for example, nitrogen oxidation reaction [47], electrochemical degradation of polluted water [48] et al) by oxidizing reaction intermediates. On the other hand, since singlet oxygen is highly active, there are possible interactions between singlet oxygen and catalysts or additive agents (i.e. conductive carbon and binder). The influence by such interactions on the stability of the catalyst surfaces probably need to be considered. For example, active 1 O 2 may directly lead to the degradation of carbon-based materials [49]. The electrode stability may be also influenced by 1 O 2 through indirectly reacting with additive agents such as binder.