Reactive oxygen species in iridium-based OER catalysts

Exceptional reactivity of electrophilic oxygen species in highly OER-active IrIII/IV oxyhydroxides is evidenced by room temperature CO oxidation.


CO exposition of iridium oxides in flow-through reactor
CO oxidation experiments over iridium oxide powders were first performed in a glasslined steel, U-shaped tube reactor with an inner diameter of 5 mm. 25 mg of iridium oxide powder were diluted in 250 mg inert SiC powder with a particle diameter of 250 µm -355 µm. Prior to CO exposure, the samples were dried in a He flow of 100 mL min −1 (controlled by Bronkhorst mass flow controllers) for 1 h at room temperature. The subsequent switch from inert He to a 1 % CO in He flow (100 mL min −1 ) was realized via a 6-port switch valve (Valvo, Vici), which excluded dead volumes. The effluent gas was monitored on line with an X-Stream, Emerson/Rosemont gas analyzer to quantify CO and CO 2 concentrations. The catalyst temperature was recorded by an analog connection to the gas analyzer. To ensure gas purity, the CO line was equipped with a carbonyl remover (consisting of a tube filled with inert SiC heated to 573 K) as well as a CO 2 -trap (crushed KOH filled cartridge). The He line had water as well as oxygen filters.
The iridium oxide samples were an X-ray amorphous Ir III/IV oxyhydroxide (IrO x , 99.99 % metals basis, AlfaAesar Premion ® ) and a crystalline rutile-type IrO 2 (99.9 % metals basis, Sigma Aldrich). Details on the materials' properties can be found elsewhere. for IrO x and rutile-type IrO 2 . The left graph shows both the CO and CO 2 traces for the two samples. The right graph shows only the CO 2 traces zoomed in and includes a second CO dosage after having exposed the sample to a He flow for 20 min after the first CO dosage. During the second CO dosage, starting at 183 min, the CO 2 concentration is two orders of magnitude smaller than during the first CO dosage, starting at 0 min. Figure S1 shows that upon initial CO introduction into the reactor loaded with IrO x , a strong CO 2 signal is measured that declines to a value close to the baseline within 50 min. In a second dosage of CO, after 20 min of pure He exposition, the recorded CO 2 signal of IrO x is smaller by two orders of magnitude, which corroborates the assumption that the majority of reactive oxygen surface species was consumed during the first CO exposure. For the rutile-type IrO 2 sample, no CO 2 is registered upon the exposition of the sample to CO ( Figure S1). were dosed each for 20 min and NEXAFS was measured in between without air contact.

CO exposition IrO
For the ozone generation, a commercially available ozone generator TC-1KC was used.
Oxygen was passed through Teflon tubing at a rate of 1 L min −1 . The effluent gas contained a mixture of ≈1 % ozone and 99 % molecular oxygen and was dosed into the measurement compartment via a leak valve.
In the XPS measurements, a pass energy (PE) of 10 eV and an exit slit setting of 111 µm was used, which led to an approximate resolution of 0.  Over the course of the entire experiment of the IrO x sample (top), the shape of the CO 2 profile for the IrO x is considerably different to the traces of H 2 O and O 2 , confirming that the observed formation of CO 2 is not simply a reaction of CO with residual gas-phase oxygen or water. The figures in the lower part show a comparison of the active (IrO x , left) and inactive (rutile-type IrO 2 , right) samples. In both experiments, when introducing the sample at t = 0 min into the chamber filled with CO, also the H 2 O and the O 2 traces increase since more surface area of the setup (i.e. load lock in which sample was stored) was exposed, which led to a slightly different background signal of the QMS traces. Nevertheless, it is evident that the change in the CO 2 trace of the IrO x is not just a change in background signal, but the actual formation of CO 2 .  Figure S3: Ir 4f spectra (left) and O 1s spectra (right) of IrO x before and after CO and after ozone exposure. While the Ir 4f spectrum is affected neither by CO nor by ozone, the intensity of the O 1s spectrum at low binding energies slightly decreases after CO exposure. The intensity increase in the O 1s spectrum at ≈531 eV after ozone exposure is most likely due to carbonate formation at the surface (compare to C 1s spectrum in Figure S4 (left)). This carbonate formation at the surface might mask the replenishment of O I− in the top layer of IrO x .  Figure S4: C 1s spectra (left) and S 2p spectra (right) of IrO x before and after CO and after ozone exposure. After the CO exposure, mainly the intensity at around 284.5 eV corresponding to graphitic carbon increases, a phenomenon commonly observed in near-ambient-pressure XPS chambers due to the "high" base pressure (10 −6 Pa) of such systems when compared to UHV-only systems (10 −8 Pa). 4 This phenomenon also occurs when keeping the sample in vacuum without dosing CO as shown in Figure S7. After the ozone exposure a slight increase of intensity at around 286 eV may be seen due to oxidation of carbon forming carbonates (as also indicated in the O 1s spectrum Figure S3) during the presence of ozone. The S 2p spectra show that the amount of S on the sample is slightly increasing during the course of the experiment. Since similar concentrations of C and S are observed before and after ozone, it can be excluded that the decrease/increase of the 529 eV feature after CO/after ozone is simply the result of carbon or sulfur species covering the O I− species.  Figure S5: O K-edge of the IrO x sample measured in total electron yield (TEY) mode before and after CO exposure. Like in AEY mode, the spectra witness a relative decrease of the 529 eV feature compared to the 530 eV feature upon CO dosage. total electron yield (TEY) (right) mode before and after ozone exposure. The spectra witness a relative increase of the 529 eV feature compared to the 530 eV feature upon ozone exposure. The additional contribution in the AEY spectra between 548 eV and 558 eV is due to the S 2p XPS peak interferring with the AEY spectrum. After opening the connecting valve, the gas volume in the microcalorimeter was flushed out by a He stream of initially 1 mL min −1 and subsequently 5 mL min −1 ( Figure S11).
In the effluent gas, we found no residual CO but only CO 2 (see Figure S11). This finding suggests that the vast majority of CO introduced to the measurement volume In the experiment, the next dosing step was only performed once the thermosignal base line was reached (for higher amounts of reacted CO not shown). The low signal-to-noise ratio for the first dosing steps is due to the very small thermosignals registered.  In the experiment, the next dosing step was only performed once the thermosignal base line was reached (for higher amounts of reacted CO not shown). The low signal-to-noise ratio for the first dosing steps is due to the very small thermosignals registered.
was consumed by reactive oxygen species of IrO x . Therefore, in the calculation of the differential heats of reaction, the entire dosed amount of CO was taken into account.
There might be, of course, some residual, unreacted CO molecules that were possibly not detected by the microGC. Unfortunately, due to the measurement setup, it was not possible to measure the composition of the gas phase after each CO pulse in order to determine the exact number of molecules that reacted in each step. The reported value for the reaction enthalpy is therefore given with a considerable error bar and has to be viewed as a lower limit for the reaction enthalpy per molecule. The large standard deviation of this measurement shows that the equipment was operated at its limits since it is usually used only for monitoring adsorption and not reaction events. to ensure both initial and final state effects were captured. 11 The resultant relative O 1s BEs were shifted to absolute BEs using a reference calculation on a (4x4x4) supercell of IrO 2 . We took the measured O 1s BE of the reference system as 530.0 eV. Previously, 1 we have verified that the relative BEs in IrO 2 were converged to better than 0.1 eV with our computational setup.

O K-edge spectra
Oxygen K-edge spectra were computed in two ways, which proved to be nearly identical, see Figure S13. In the first we approximated the X-ray absorption process using a oneelectron Fermi's golden rule expression as implemented in the XSpectra package. 12,13 Here we neglected the core-hole and used the ground state results from the USPP calculations. We previously found this approach to give good agreement with experiment for rutile-type IrO 2 when the computed spectra were convoluted using a Lorentzian with an energy dependent linewidth, Γ(E) = Γ 0 + Γ(E), to account for lifetime broadening. 1 In an effort to ensure this one-electron approximation remains valid for the defect structures giving rise to the O I− state, we also employed a resolvent-based Bethe-Salpeter Equation (BSE) approach to capture the screened core-hole potential and electron-hole dynamics. 14,15 These calculations were carried out by combining the Kohn-Sham wavefunctions from QE (Quantum Espresso) with the NIST core-level BSE solver (NBSE) through the OCEAN package. 16,17 For these calculations, however, we employed norm conserving pseudopotentials generated with the FHI98PP package. 18 The exchange and correlation was treated with the local density approximation (LDA) based on Perdew and Wang's parameterization of Ceperely and Alder's data. 19 We found the results were converged using a kinetic energy cutoff of 100 Ry (400 Ry) for wavefunctions (charge density) with ground and final state k-point mesh equivalent to those used in the total energy calculations. Methfessel-Paxton smearing 20 was used with a width of 0.002 Ry during the self-consistent field calculations. The screening calculations were performed including bands to capture states over 100 eV above the Fermi energy. All spectra were broadened with a Lorentzian with a constant width of 0.2 eV to account for lifetime broadening. For both the BSE and one electron calculations the spectra of the symmetry unique oxygen atoms were aligned relative to one another using their corresponding ∆SCF binding energy. Gaussian broadening (FWHM=0.6 eV) was also introduced to account for instrument resolution and phonon broadening. Figure S13 shows the BSE and one-electron O K-edge spectra of IrO 2 (left) and the IrO x -type oxide discussed earlier (right). 1 Both calculations are shown without Gaussian broadening to highlight their differences. The spectrum calculated with the one-electron approximation can be seen to capture all the features present in the BSE approach.
While there are slight differences, the remarkable agreement is a consequence of the fact that the 1s core hole is energetically isolated from other states on the atom, which tends to diminish the importance of core-hole dynamics. As a result, the primary differences in the spectra are due to the two treatments of lifetime broadening and core-hole potential, e.g. linear response within the random phase approximation for the BSE calculation compared to the DFT ground state without a core hole.

Heats of reaction
By assuming that the reactant CO and product CO 2 are gas-phase species, as indicated by gas-phase analysis, the heat of reaction can be computed as: where E IrOx is the total energy of the IrO x system missing one oxygen, E IrOx−O* is the total energy of the parent IrO x -O* system, and E CO 2 (E CO ) is the total energy of an isolated CO 2 (CO) molecule computed at the Γ-point in a 30Å x 30Å x 30Å box. The E CO 2 − E CO term is a constant that sets the energy of the oxygen atom in the product.
Thus, equation 1 could be rewritten:

Reactions on (110)
We considered the reaction of CO (g) with O on the (110) CUS, bridge, and lattice sites.
In this case, we used the fully oxidized surface, as shown in Figure S14. However, we employed a (1x1) cell for computational efficiency.
Reaction with the CUS oxygen had a barrier of 158 kJ mol −1 and a heat of reaction of -245 kJ mol −1 . The high barrier may be tied to the symmetry constraint imposed on the incoming CO, e.g. the CO axis was aligned with the Ir-O CUS axis. Because we saw no spectroscopic evidence for CUS oxygen we did not explore this further.
Conversely, CO (g) reaction with the bridging oxygen had a barrier of 15 kJ mol −1 and heat of 87 kJ mol −1 . We could not identify a minimum energy path for reaction with the in-plane oxygen.
Thus, only the O I− like species was found to react with gas-phase CO.

Reactions on (113)
We considered the reaction of CO with oxygen on a partially reduced (113) surface, Figures S17 and S18. The surface has two types of under coordinated Ir atoms that can adsorb gas-phase CO, one is coordinated by 3 oxygen atoms and one by 4 oxygen atoms.
Of these, a 4-fold site binds CO the strongest, at 237 kJ mol −1 . Thus, we explored the reaction of CO with surface oxygen from this site. The carbon side of the CO adsorbed on 4-fold site is 3Å from an O I− site, where the nature of the site was determined by BSE calculation, see Figure S19. We saw that after CO adsorption the (113) surface could distort, thereby forming an Ir-Ir bond, see Figure   S20. The barrier for this distortion was only 10 kJ mol −1 . It is not clear if the distortion is related to CO adsorption or a property of the clean surface. When CO is present the systems total energy is lowered by 97 kJ mol −1 through this Ir-Ir bond formation.
From this state the barrier for CO ads reaction was 50 kJ mol −1 . The product CO 2 was adsorbed by 163 kJ mol −1 .