In situ mapping of activity distribution and oxygen evolution reaction in vanadium flow batteries

Understanding spatial distribution difference and reaction kinetics of the electrode is vital for enhancing the electrochemical reaction efficiency. Here, we report a total internal reflection imaging sensor without background current interference to map local current distribution of the electrode in a vanadium redox flow battery during cyclic voltammetry (CV), enabling mapping of the activity and reversibility distribution with the spatial resolution of a single fiber. Three graphite felts with different activity are compared to verify its feasibility. In long-term cyclic voltammetry, the oxygen evolution reaction is proved to enhance activity distribution, and homogeneity of the electrode and its bubble kinetics with periodic fluctuation is consistent with the cyclic voltammetry curve, enabling the onset oxygen evolution/reduction potential determination. Higher activity and irreversibility distribution of the electrode is found in favor of the oxygen evolution reaction. This sensor has potential to detect in situ, among other processes, electrochemical reactions in flow batteries, water splitting, electrocatalysis and electrochemical corrosion.


Comparison of the TIRi sensor and SPRi sensor after cyclic voltammetry.
As the module of the TIR sensor is the prism, it is still stable in the positive electrolyte of 0.1 M VO 2+ and 2 M H2SO4 after cyclic voltammetry (CV) as shown in Supplementary Figure 1

Photograph and SEM comparison of GF, TGF and PGF electrodes.
Supplementary Figure 2

Calibration of the TIRi sensor.
To calibrate the

Images of the intensity and intensity variation in one cycle.
Supplementary Figure 4(a) shows seven TIRi images captured by the CCD camera at different potentials in one cycle. Supplementary Figure 4(b) shows six images of intensity variation by subtracting the initial image at 0V from all the subsequent images in one cycle. In detail, the intensity of the image keeps stable between 0 V and 0.06 V for no electrochemical reactions. Subsequently, the intensity begins to decrease sharply at 0.66 V for the oxidation of the VO 2+ ions leading to the increasing refractive index of the electrolyte. After maintaining stable between 0.66 V and 1.56 V, the intensity starts to recover into the almost initial intensity because of the reduction of the VO2 + ions. The visible intensity variation images confirm the TIRi sensor to detect the electrochemical reaction of the VFB's electrode.
where 0 and 0 are the concentrations of the oxidized and reduced species at the initial time, n 5 is the electron number of the redox reaction, F is the Faraday constant, DO and DR are the diffusion coefficients of the oxidized and reduced species, ( ′ ) is the current density. Similar to the reported work 2 , the response of the TIR sensor in view of its detection depth is where ( ) is the reflected light intensity, and are the refractive index changes per unit concentration of the oxidized and reduced species respectively. R is the intensity change per unit refractive index. According to equations (S2), (S3) and (S4), the intensity can be derived: the intensity variation is shown as follows: Rewrite the equation (S6) as: Hence, the current density can be derived as: Actually, the value of "nF/b" is a constant c (nF/b = c). For simplicity, the relative current density distribution is calculated herein by supposing nF/b =1 in this work. We can get the current density distributions by setting nF/b with the different values, but the ratio of current density between different positions remains the same. By the following calibration experiment, the absolute current density distribution remains the same under the condition of nF/b with different values. The derived CV curves of A, B and C in Supplementary Figure 5(d) show the relative local current densities along with the potential. In order to verify the current density measurements by deconvolution of the intensity variation, we adopt the calibration experiment to provide the current density with a physical unit. A graphite plate with smooth interface is used as the uniform and well-tunable electrode for calibrating the light intensity and the current density measurements. As shown in Supplementary Figure 6 Figure 5. Hence, the calculated relative current density is not absolute value. Compared to measuring the diffusion coefficient, we obtain the current density with a physical unit through the calibration experiment. According to the above quantitative relationship between the current density (ipa(EW)) recorded by EW and the relative current density (ipa(TIR)) obtained by the TIRi sensor ( ( ) = 0.0044 ( ) ), the current density with a physical unit recorded by the TIRi sensor can be achieved. The relationship between the current density (ipa(EW)) recorded by EW and the relative current density (ipa(TIR)) obtained by the TIRi sensor.

Simulation of the cyclic voltammetry process.
To prove the semi-infinite diffusion model and the convolution relationship in Supplementary For further studying the contributions of the detection signal, we have studied the effects of diffusion effect and different electrode reaction rates on concentration variation of the electrolyte, which are described in the following three aspects.
(1). The diffusion effect between fiber arrays on concentration variation in vertical direction.
As shown in Supplementary Figure 12 (2). The diffusion effect between fiber distribution on concentration variation in horizontal direction.
As shown in Supplementary Figure 13  To further verify the above description, the mappings of (a) the peak oxidation current densities

Region of interests (ROI) for current density curves.
As shown in Supplementary Figure 17(a), the size of an image captured by CCD is 1460 × 1920 pixels. In order to reduce the intensity noise, the intensities of each 5×5 pixels are averaged into the 23 intensity of one point as shown in the enlarged views of Supplementary Figure 17(a, b). We label each point as region of interest (ROI).  Fig. 3b.

The electrochemical activity distribution of the thin surface layer of graphite fibers.
The electroactive area of a bulk electrode (graphite felt) is the surface areas of graphite felt fibers that participate in electrochemical redox reaction (as the electrochemical reaction sites). Traditional electrochemical techniques to measure it are cyclic voltammetry, chronovoltammetry and so on. For example, in the cyclic voltammetry process, the scan rate is changed to obtain different CV curves.
According to the relationship of the peak current in proportion to the square root of the scan rate, the electroactive area of a bulk electrode is obtained. Besides, in the chronovoltammetry process, the electroactive area of a bulk electrode can also be obtained by the Cottrell equation 1 .
As shown in Supplementary Figure 19 and provides the local current density, which is affected by the fibers in the diffusion layer (thickness: the diffusion length ~1 mm), including local current and diffusion current. For example, the reactant and product concentrations at the point G within the penetration depth is affected by electrochemical reaction from the local fibers and the diffusion process within the diffusion length (semi-sphere C in Supplementary Figure 19(b)). The current density at the point G has the same situation. Therefore, the image reflects the electrochemical activity distribution of the thin surface layer of graphite fibers.

Parallel measurements of current density distribution by the TIRi sensor.
In order to provide statistical data and stability of the sensor in multiple measurements, we have

Different potential windows in the long-term cyclic voltammetry.
The purpose of selecting different potential windows is: (1). To determine the onset potential by directly counting the number of bubbles, it is necessary to start the oxygen evolution reaction under a narrow potential window (such as: 0 V -1.7 V). So the 33 bubble generation is more moderate, which is convenient for statistical calculation of bubbles.
(2). To study the relationship between the oxygen evolution reaction and the electrochemical activation effect of the electrode, the oxygen evolution reaction in a large potential window is required to generate a large number of oxygen bubbles, resulting in a higher degree of electrode activation.
Therefore, this work selects two different potential windows to achieve the above purpose.
As shown in Supplementary Figure 27

Comparison of the bubble distributions from the local bubble dynamics and the onset potentials of oxygen evolution/reduction reaction at different cycles.
In order to confirm the bubble dynamics, we compare the bubble distributions from the local bubble dynamics, and the onset potentials of oxygen evolution/reduction reaction at different cycles.
(1). Three types of the local bubble dynamics.
As shown in Supplementary Figure 29 (2). The onset potentials of oxygen evolution/reduction reaction at different cycles.
In order to further verify the feasibility of visually determining the onset potential of the oxygen evolution/reduction reaction by the counts of bubbles, we have plotted the kinetic curves of the counts of bubbles in different cycles (e.g. 5th, 10th, 15th, 20th, 25th, 35th, 45th) during long-term CV as shown in Supplementary Figure 30 14. Bubble generation in the region with high activity. Figure 31, the region labelled by four dashed rectangles is the region R with comparatively high |ipa|. According to Figure 4(c) or Supplementary Video 1, the bubbles from the oxygen evolution reaction appear on the region R. It means that the region R with high activity is more likely to generate bubbles. As the onset oxygen evolution potential is among the potential scan range (0 V -1.7 V, Figure 4(b)) and the counts of bubbles in the 1st cycle ( Figure   4(a)), the oxygen evolution reaction begins at the 1st cycle.

As shown in Supplementary
Supplementary Figure 31. The mapping of the peak oxidation current densities |ipa| of the TGF in the 1st cycle.
Four dashed rectangles: the region R with comparatively high |ipa|.

The influence of the bubble on the local electrochemical reactions.
In the long-term CV, bubbles would block the local electrochemical reaction (abbreviated as: BBR), as can be seen from the first few cycles (1st-5th cycles in Supplementary Figure 32) in longterm CV. In detail, the peak oxidation/reduction currents of the CV curves in the first few cycles gradually become smaller in Supplementary Figure 32(b, c, d), because the bubble generation at the beginning is in the growth stage, and thus the BBR effect is greater than the electrochemical activation effect (owing to the generation of the oxygen-containing functional group on the electrode).
As the cycles continue, the counts of bubbles become saturated and reach a stable equilibrium state as shown in Figure 4(a). It means that the BBR effect reaches the limit. At the same time, the electrochemical activation effect gradually increases and also reaches the limit. However, compared with the BBR effect, the electrochemical activation effect dominates at this time, so that the net effect makes the peak oxidation/reduction currents of the CV curve increase in Supplementary

Region of interest for activity and reversibility distribution.
To exclude the fiber contact regions with low reflected intensity for its absence from the measurement range with high sensitivity, we choose the range of 1.54 -3.41 mA cm -2 for |ipa|, 1.

Bubble generation on the electrode (TGF + PGF) after the long-term CV.
Supplementary Figure 41 shows the bubble generation on the electrode (TGF + PGF) after the long-term CV. Almost all bubbles grow on the PGF instead of the TGF owing to the PGF with higher activity rather than the TGF. It indicates that bubbles are more like to generate on the electrode with higher activity.

Bubble generation process in the first five cycles at different scan rates.
Supplementary Figure 42 shows the intensity images captured by the CCD camera in the first five CV cycles at different scan rates. It is faster to generate bubbles when the scan rate is lower as shown in Supplementary Figure 42. The CV curves of the TGF recorded by the EM at different scan rates are plotted in Supplementary Figure 43(a). It is found that the oxidation and reduction current densities of the vanadium ions raise sharply along with the increasing scan rate while the current density of the OER has a very small increase along with the increasing scan rate. The counts of bubbles generated by the OER in the first five cycles at different scan rates are plotted in Supplementary Figure 43(b). When the scan rate is small, the oxidation and reduction current 48 density of the vanadium ions is small while the bubble generation of the OER is fast and vice versa.
This attributes to the enough current density of the OER and its longer duration time at lower scan rate.

Reflectivity comparison of p and s polarized light for the TIRi sensor in theory.
Supplementary Figure 46 shows the angular spectrum of p and s polarized light for the TIRi sensor in theory. When the incident angle is larger than the angle of the total internal reflection TIR , the total internal reflection occurs and the reflectivity is 1 for p and s polarized light. If the incident angle is smaller than TIR , the reflectivity sharply decreases and the decrease rate of the p polarized light is larger than that of the s polarized light. It means that the reflectivity of the p polarized light is more sensitive than that of s polarized light at the incident angle near TIR . Hence, the p polarized light with higher sensitivity is chosen in this work as the incident light of the TIRi sensor.