In-situ plasmonic tracking oxygen evolution reveals multistage oxygen diffusion and accumulating inhibition

Understanding mass transfer processes concomitant with electrochemical conversion for gas evolution reactions at the electrode-electrolyte interface plays a key role in advancing renewable energy storage and conversion. However, due to the complicated diffusion behavior of gas at the dynamic catalytic interfaces, it is still a great challenge to accurately portray mass transfer of gas during electrocatalysis process. Here, we track the diffusion of dissolved oxygen on Cu nanostructured plasmonic interface, which reveals multistage oxygen diffusion behaviors, including premature oxygen accumulation, spontaneous diffusion and accelerated oxygen dissipation. This work uncovers an accumulating inhibition effect on oxygen evolution arising from interfacial dissolved oxygen. With these knowledges, we develop a programmable potential scan strategy to eliminate interfacial gas products, which alleviates the concentration polarization, releases accessible actives sites and promotes electrocatalytic performance. Our findings provide a direct observation of the interfacial mass transfer processes that governs the kinetics of gas-involved multiphases catalysis.

. Second, the as prepared Cu seeds on ITO electrode (abbreviated to Cu seeds/ITO electrode) was served as the working electrode for subsequent growth of Cu NPs in the same electrochemical cell. The growth of Cu NPs was performed by a cyclic voltammetry during a dynamic potential range from -0.10 to -0.30 V (vs. SCE) with different deposition cycles, at a scan rate of 0.05 V s -1 . Then, the as-prepared electrode 3 (abbreviated to Cu NPs/ITO electrode) was removed from the deposition solution and rinsed copiously with deionized water. Last, the Cu NPs/ITO electrode was dried under a gentle stream of nitrogen before its utilization for gas evolution reaction.

Simulations
Full-wave electromagnetic simulation was performed on a commercial software (LUMERICAL, FDTD Solution) and finite element models (FEM) simulation was carried out by COMSOL Multiphysics 5.3a software (COMSOL Inc.) on a high-performance desktop workstation (16 G RAM). For FDTD simulations, the perfectly matched layers were used along the propagation directions and the spatial grids' size in all axes was set to 2.0 nm. Consider the proposed model, the complex refractive index of Cu2O, CuO and Cu(OH)2 were taken from literatures. [3][4] Both magnetic-and electric-field enhancement maps under different excitation wavelength were calculated in FDTD using frequency domain field. The extinction spectra of proposed configurations were conducted using a regular plane-wave with a pulse length of 2.6 fs. The power monitor is located in a plane 14 nm away from the metal film. In order to eliminate the contribution from random dispersed particles induced over-evaluation on the gap distance, a gap distance 26 nm for Cu NPs/ITO (150 cycles) was adopt in simulation (Supplementary Figure 8). We adopted two strategies to approximately estimate the influence of oxygen accumulation on the optical properties based on the individual unit cell dimer-onfilm structure shown in Supplementary Figure 8.
(1) As oxygen evolution, the refractive index of the electrolyte (1.33) located at interfacial region could gradually decrease due to the accumulation of oxygen molecules (refractive index: 1.00). 5 Because the limitation in precise identification of absolute oxygen centration close to the electrode surface, it is very hard and impossible to acquire the dynamic hybrid refractive index of interfacial electrolyte (including solved oxygen) during dynamic potential scan. To qualitatively evaluate the influence of oxygen accumulation, we rationally set the refractive index of environment from 1.33 to 1.06 in simulating the dynamic environment in potential range from + 0.20 V to + 0.90 V then to + 0.67 V (Region III corresponding to 130 s to 220 s in time scale); To further evaluate subsequent oxygen spontaneous and accelerated diffusion processes, the refractive index of environment was 4 rationally set from 1.06 to 1.20 in potential range from + 0.67 V to -0. 35  (2) We then considered an extreme condition, where oxygen molecules are accumulated on the electrode surface to form an oxygen shell around Cu NPs, this enables to exclude surrounding aqueous electrolyte solution and to decrease the refractive index of environment. To qualitatively evaluate the influence of oxygen accumulation, we rationally set the thickness of the oxygen shell from 0 nm to 120 nm to simulate the dynamic environment in the potential range from + 0.20 V to + 0.90 V then to + 0.67 V (Region III corresponding to 130 s to 220 s in time scale); To further evaluate subsequent oxygen spontaneous and accelerated diffusion processes, the thickness of oxygen layer was rationally set from 120 nm to 85 nm in the potential range from + 0.67 V to -0. Despite all these efforts, it should be noted that the dynamic of oxygen diffusion at the EEIs is far more complicated than what we proposed here. Our simulation results only provide a qualitative evaluation of interfacial processes, a more precise description of the dynamic interfacial electrochemical reactions should consider electric field interference, build more complicated models and adopt density functional methods or molecular dynamics simulations.

Electrode
In classical nucleation theory, nucleation of a new solid phase on a heterogeneous interface has to overcome a free energy barrier related to the thermodynamic costs for forming a critical cluster of atoms. 6 During electrodeposition, this nucleation energy barrier can be effectively regulated by changing the local electrochemical supersaturation at the electrode-electrolyte interface through adjusting the applied overpotential on the reduction reaction. 7 The driving 5 force induced by reaction, charge transfer and diffusion as well as crystallization overpotential work together on the electrodeposition process. 8 Considering the difficulty in identification of each overpotential contribution, here we referred to nucleation (ηn) and growth (ηg) overpotential for explicit describing Cu electrodeposition (Supplementary Figure 1). During initial nucleation stage, an essential prerequisite for nucleation of Cu embryos was a sufficient overpotential, which enabled to surmount the large free energy barrier (ΔGn). After nucleation occurs, attributed to the lowering free energy barrier (ΔGg) for the growth of Cu seeds, a lower overpotential was competent in subsequent growth process. Because of the lower free energy barrier, it was more thermodynamic favorable to introduce a Cu adatom on a Cu nuclei interface than re-nucleation on a non-homologous heterogeneous interface. Taken together the difference of the free energy barrier in both nucleation and growth, an electrochemical seed-mediated To construct the plasmonic-enhanced system, the copper seeds were firstly deposited on ITO electrode at a fixed polarization nucleation pulse (Figure 1 Figure 2). Hence, it provided the discrete preferential nucleation sites for subsequent copper deposition, avoiding the random heterogeneous nucleation. For achieving high copper seeds coverage, the cathodic polarization potential was optimized without the disturbance from side reaction, such as hydrogen evolution reaction. The detailed electrochemical and time-of-flight secondary ion mass spectrometry analysis was shown in Supplementary Figure 3-4. Supplementary Figure 3 indicated that the amount of charge during the nucleation increased drastically accompanied by the Cu coverages reduction beyond -0.60 V vs. SCE. It should be noted that the immoderate enhancement of polarization overpotential resulted in the co-electrolysis of Cu electrodeposition and hydrogen evolution that required extra amount charge to complete the nucleation step. Specifically, hydrogen evolution competed with the Cu nucleation and restrained the diffusion of the Cu 2+ to the electrode surface, as indicated by the decreased diffusion coefficient of Cu 2+ from 2.20×10 6 cm 2 s -1 at -6 0.60 V to 1.98 ×10 6 cm 2 s -1 at -0.90 V (Supplementary Figure 4). In order to eliminate the coelectrocatalysis of hydrogen evolution and increase the efficiency of Cu electrodeposition, a polarization potential at -0.6 V vs. SCE was chosen for the initial Cu nucleation and followed by a relative lower polarization potential window from -0.10 to -0.30 V vs. SCE for subsequent growth of the presence Cu nuclei (Supplementary Figure 5).
The diffusion coefficient of Cu 2+ during electrochemical deposition condition was acquired from Cottrell equation. In brief, the bare ITO electrode was submerged in the solution containing 1 mM CuSO4 and 0.1 M NaClO4 with a Pt wire counter electrode and a SCE reference electrode (Supplementary Figure 4). The current transients were recorded for electrodeposition of Cu NPs by applying a potential step from 0.20 V to various termination potential (− 0.50, − 0.60, − 0.70, − 0.80 and − 0.90 V vs. SCE). The current decay was linear with t -1/2 , and it indicated a planar diffusion regime emerged caused by the overlapping of growing hemispherical diffusion layers, which provided mass transport for nanocrystal growth in the process of copper nucleation. [9][10] The diffusion coefficient could be obtained from the slop of i versus t -1/2 plot based on the Cottrell equation: 11 where n is the number of transferred electrons (n = 2), F is the Faraday's constant (96485 C mol -1 ), DCu2+ and CCu2+ is the diffusion coefficient and the bulk concentration of Cu 2+ (1 mM), A is the geometric area of the bare ITO electrode 1.5 cm 2 (width 7.5 mm and length 20 mm).
The calculated diffusion coefficient for copper electrodeposition on a bare ITO electrode under cathodic polarization potential (− 0.60 V) was 2.2×10 -6 cm 2 s -1 , which was on the same magnitude compared to the diffusion coefficient of Cu 2+ in water (5.0 ×10 -6 cm 2 s -1 ) 12 and the polarographic diffusion coefficient of copper ions (7.2 ×10 -6 cm 2 s -1 ). 13 Note that no remarkable initial current spine caused by the charging of the double electric layer was observed (Supplementary Figure 3, a-e).
It could be concluded that the growth rate of copper nuclei was extremely rapid. The rising portion of the transient current was not experimentally accessible under our conditions, thus the estimation of the nucleation mechanism (i.e., instantaneous or progressive) of copper nanoparticles was not possible from the transient current curves. 14 However, it was reasonable 7 to conjecture that the electrochemical reduction process of Cu 2+ described here experienced an instantaneous nucleation process on a bare ITO electrodes based on the fact that no noteworthy charging phase was observed. 15 The electrochemical formed copper seeds exhibited higher coverage and favorable mono-dispersity than progressive nucleation, which could avoid seriously coupling of the diffusion layer of the adjacent copper nanoparticles and mitigate deteriorative inhomogeneity of the particle size. was assigned to the reduction of Cu 2+ to Cu. The other two anodic peak II and III was ascribed to the oxidation of Cu to Cu + and Cu to Cu 2+ , respectively. The dependence of the peak currents of peak I on the square root of the scan rate, revealing a diffusion-controlled kinetic process during reduction of Cu 2+ . Meanwhile, the dependence of the peak currents of peak II (e) and III (f) on scan rate. It indicated that both the anodic processes were involved in surface reaction-controlled manner.

Supplementary Note 3：Characterization of Electrochemical Deposited Cu NPs
Electrochemical seed-mediated growth of Cu NPs was adopted to fabricated plasmonicenhanced system for single-wavelength monitoring interfacial reactions at the electrodeelectrolyte interface. Owing to the enlarged nanoparticle size, the extinction cross section increased gradually and tend to saturate after deposition 150 cycles ( 0.21 nm could be indexed to the Cu (111) plane of face-centered cubic Cu, which indicated that the major exposed planes were Cu (111) of face-centered cubic Cu in accordance with the above XRD results. In addition, high-angle annular dark-field imaging and energy-dispersive X-ray spectroscopy (EDX) analysis revealed Cu NPs possessed a nanosphere morphology with a diameter of ca. 160 nm which was agreed well with the AFM measurements (164 nm) ( Figure   1). Due to the relative lower cathodic polarization window from -0.10 to -0.30 V vs. SCE applied for the Cu nuclei growth, it could explain the growth of the Cu NPs were mainly To quantitatively describe the gap distribution between adjacent Cu nanoparticles, we provided a statistical analysis on the average particle-to-particle distances (Supplementary Figure 9). According to the definition of gap, only the two adjacent NPs with a minimum distance comparable to surrounding particles are considered as a candidate dimer for statistical analysis. From the statistical analysis we found a decrease of the gap from 479 nm to 55 nm with deposition cycles from 5 to 150 cycles accomplished with a gradual growth of particle size, and a slightly increase of gap from 55 nm to 97 nm with deposition cycles from 150 to 300 cycles caused by the Ostwald ripen-like effect. 17 In order to eliminate the contribution from random dispersed particles induced over-evaluation on the gap distance, we adopted half of the Owing to the increase of the symmetry of trimer, more possible configurations are present in the ensemble system, which result in the extreme difficulty in distinguish their contribution to optical characteristics.
Here, a seed-mediated electrochemical deposition strategy instead of electron beam evaporation or other photolithography techniques, was used to fabricate the particles-on-film structure. Due to the intrinsic limitation of non-template electrochemical method, the formation of dispersed particles and trimers is not avoidable. The electrodeposited structures, such as dimer and timer on the film, could also attribute to the optical properties of the ensemble system.
We compared the distribution and amount of isolated dimer and isolated trimer (per μm 2 ) on Cu NPs/ITO electrode (150 cycles). Statistical analysis indicated that the isolated dimer was 19 more widely distributed, and its cover density (per μm 2 ) was ca. 4.5 times higher than that of Cu O + H + 2 → 2 ( ) + 2 (4-3) The influence of the sweep rate on both anodic and cathodic polarization was shown in Supplementary Figure 11, b. The cathodic peaks C1 was assigned to the electrochemical reduction of CuO to Cu2O. The other peak C2 was caused by the simultaneous reduction of both Cu2O and CuO to Cu. 19 During kinetic analysis, a linear relationship between anodic peak currents and the scan rate indicated a surface reaction controlled kinetic in anodic processes 22 (Supplementary Figure 11, b, Peak A1, A2 and A3). Meanwhile, the cathodic processes (Peak C1 and C2) exhibited a linear function as the square root of the scan rates, indicating a diffusion-controlled mass transport in these processes (Supplementary Figure 11, c).
Interestingly, a gradually emerging cathodic peak located at − 0.45 V was found at relative high scan rate (> 0.20 V s -1 ) which was assigned to the reduction from Cu(OH)2 to Cu2O. Taken   We further performed rotating-ring disc electrode (RRDE) measurements to collect oxygen at various polarization potentials to explore more insights into OER. We found that at collection potential − 0. Here we aimed to use electrochemical steady-state extinction spectroscopy to monitor

Supplementary Note 5：Principle of the Single-Wavelength Monitoring Strategy
The plasmonic properties of metal nanoparticles can be described by solving the Maxwell equation using classical electrodynamic Mie theory. [25][26] According to Mie theory, the analytical solution to the scattered fields generated by a plane wave incident on a homogeneous plasmonic nanoparticle leaded to the extinction, scattering and absorption cross-section 27-28 : the general expressions for the scattering and extinction cross sections from the Mie theory can be expressed as where k is the incoming wavevector and L are integers corresponding to the diploe, quadrupole and hexapole contributions etc. Here, the aL and bL composed of Riccati-Bessel functions are given by: where m=nn/nm, nn and nm is the refractive index of the nanoparticle and surrounding medium, respectively. And x = Rkm, R is the radius of the nanoparticle, km is the wavenumber of incident light. When the size of the plasmonic nanoparticles is relative smaller than the wavelength of the incident light, taking into account the contribution from dipole term, the extinction crosssection of the plasmonic particle can be given by: where c is the speed of the light and R is the radius of the nanoparticle, V is the volume of the nanoparticle, ω is the frequency of the incident light and εm is the dielectric constant of the surrounding medium, ε is the dielectric constant of the nanoparticle, ε = ε1 + iε2. ε1 and iε2 are real and complex parts of the material dielectric constant, respectively. 27 Based on this equation, the optical properties of the plasmonic nanoparticles can be engineered by tuning their size and dielectric properties of the surrounding and material of the nanoparticle. 29 The extinction crosssection will increase when the medium's dielectric function decrease. The dependence of the LSPR extinction cross-section on the surrounding dielectric environment enables the ultrasensitive detection of the minor changes in surrounding medium. Note that the plasmonic 27 extinction cross-section is proportion to a negative half power of dielectric function of surrounding medium. Conventionally, the plasmonic sensing application is to detect changes in the bulk refractive index of surrounding environment by monitoring the LSPR peak wavelength shifts through spectroscopic extinction measurements (eq. 5-5 and 5-6).
= (5-5) where ωmax is the LSPR peak frequency and ωp is the plasma frequency of the bulk metal. The λmax and λp corresponding to the wavelength of LSPR peak and the plasma frequency of the bulk metal. Here, the λ = 2πc/ω and the relationship between dielectric constant and index of refraction εm = n 2 . Hence, the LSPR peak wavelength shift is approximately linear with changes in refractive index of the surrounding medium over small ranges of n. 28 For collection full extinction spectra, the monochromator and grating as well as multichannel photodetector are essential to identify the plasmonic peak shifts. In order to facilitate the in-situ spectroscopic monitoring interfacial electrochemical reactions, the single-wavelength extinction monitoring strategy was developed with superior sensitivity for electrochemical reactions on the EEIs and eliminated the necessity for intricate optical apparatuses. Interestingly, comparing the interfacial composition at + 0.15 V (passive region) and + 0.67 V (peak b), negligible change was observed, thus excluding the possible fluctuation of extinction intensity in peak b induced by interfacial phase transitions. Therefore, the corresponding extinction intensity dynamics in peak b was attributed to the oxygen multi-diffusion behavior. 41 In oxygen accelerated region (-1.10 V), the reduction of oxygen resulted in remarkable adsorption of hydroxyl groups on the surface, that confirmed by the obvious increase of the intensity of Cu(OH)2 and OHin O 1s peak (531.9 eV). In summary, the above quasi in-situ XPS analysis reveals detailed phase transitions under various polarization and provides direct chemical information to corroborate our electrochemical experiments. (1) In peak a, a gradual OH/O ratio increase companied by slight decrease was observed during potential range -1.10 V to -0.10 V, which was of the same evolution trend as predict by electrochemical measurements. The dynamic of extinction intensity in peak a could be attributed to interfacial Cu oxides transitions.
(2) In passive region (potential from -0.10 V to + 0.20 V), the OH/O ratio maintains a steady state, which provides another direct molecular evidence to confirm the interfacial stability. where E was the potential at distance z from the electrode-electrolyte interface (z was vertical coordinate), d was the height of the oxygen diffusion boundary from the electrode surface.
According to the Nernst Equation 38 , the potential E could be expressed as: where T is absolute temperature, R is the gas constant, F is the Faraday's constant and n was the electron transfer number (n = 4). Moreover, the formal potential EOH -/O2°' = EOH -/O2° + β, where EOH -/O2° was the standard reduction potential and β was the contribution from the activity coefficients for both OHand O2. Based on the diffusion equations, the diffusive fluxes for OHand O2 were described by Fick's Law, assuming without obvious gas bubble evolution. 39 The concentration gradient closed to the electrode-electrolyte interface prompted these species diffusion across the diffusion layer. Hence, the diffusive fluxes of OHand O2 were expressed as: = − Combining equations 7-11, 7-12 and 7-2, the potential at the electrode interface, E(z = 0) and the potential in diffusion layer, E(z = d) were obtained:  (7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17) In order to understand the influence from saturation of oxygen during anodic polarization on the gas evolution reaction, the linear scan voltammetry (LSV) was precisely modeled and the bulk mass transport losses (ηm), kinetic losses (ηk), electrolyte ohmic losses (ηo) and concentration overpotential (ηc) were considered. Under practical electrochemical measurements, the total current density was composed by the sum of convention, migration, diffusion and electric double layer charging. In general, it was expressed by the sum of the Faradaic and non-Faradic portion for the total current density: During the electrochemical measurements, the record current included both the Faradaic and non-Faradic current (iF + inF). However, during in-situ extinction measurements, the fluctuation of extinction intensity directly represented the oxygen concentration fluctuation at the electrode-electrolyte interface. Therefore, the correlation between the extinction intensity and the Faradaic current intensity could be established, so we could use the extinction intensity to denote the Faradaic current intensity (iF). In an anodic polarization, oxygen evolution at the 48 electrode surface was proceeding during persistent polarization, the relative measured current density was given by: = nF ( , − , ) = + (7-21) = nF ( , − , ) = (7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22) where ie and iext were the measured current during electrochemical and in-situ electrochemical extinction measurements, respectively, d was the thickness of the diffusion layer, CO2, b was the bulk concentration, DO2 was the diffusion coefficient of the dissolved oxygen in the electrolyte, Considering CO2, s e > CO2, s ext , hence, the ηc obtained in electrochemical procedures was larger than that measured during in-situ extinction measurement, which gave the distinct evidence for the difference in the measured concentration polarization potential and relative onset potential. Most importantly, the above analysis demonstrated that in-situ extinction measurement provided direct information related to Faradaic process without the interference from the non-Faradaic components, which possessed the capacity to precisely determine the relevant parameters in gas evolution reactions such as concentration overpotential and onset potential etc.