Imaging nanobubble nucleation and hydrogen spillover during electrocatalytic water splitting

Significance

Due to the highly dynamic nature and intrinsic heterogeneity of the electrochemical interface, it is critical to develop new tools so that electrochemical activities can be imaged with high sensitivity and high spatial and temporal resolution. We demonstrate the unique ability to image nucleation and growth of individual hydrogen nanobubbles during electrocatalytic water splitting using superresolution fluorescence microscopy. This method allows us to compare electrocatalytic activity of different electrode materials toward hydrogen evolution reaction and observe in real time the effect of hydrogen spillover from electrode-supported gold nanocatalysts.

Abstract

Nucleation and growth of hydrogen nanobubbles are key initial steps in electrochemical water splitting. These processes remain largely unexplored due to a lack of proper tools to probe the nanobubble’s interfacial structure with sufficient spatial and temporal resolution. We report the use of superresolution microscopy to image transient formation and growth of single hydrogen nanobubbles at the electrode/solution interface during electrocatalytic water splitting. We found hydrogen nanobubbles can be generated even at very early stages in water electrolysis, i.e., ∼500 mV before reaching its thermodynamic reduction potential. The ability to image single nanobubbles on an electrode enabled us to observe in real time the process of hydrogen spillover from ultrathin gold nanocatalysts supported on indium–tin oxide.

As a promising technology for energy storage (1), electrochemical water splitting generates hydrogen and oxygen on the two opposite electrodes, i.e., cathode and anode, respectively. Nanobubbles of hydrogen and oxygen form as intermediate products, which then coalesce and grow into macroscopic gas bubbles escaping from the solution. The ability to fully characterize nanobubble nucleation and growth in electrochemical water splitting can have a major impact on our understanding of the structure–function relationship of the heterogeneous electrode surface and help us design electrocatalysts with improved activity.

Nanobubbles are small pockets of gaseous molecules (2–4) formed in numerous physical and chemical processes ranging from solvent exchange (5), to molecular decomposition (e.g., hydrogen peroxide) (6), to water electrolysis (7). Despite a rich literature on crystalline materials (8–11), our ability to characterize the nucleation and growth of gases at interfaces has been quite limited. The small size, optical transparency, and fast dynamics make interfacial nanobubbles challenging to probe with sufficient spatial and temporal resolution. Stable micrometer- or submicrometer-sized bubbles can be probed by techniques such as atomic force microscopy (12) and dark-field (13) or fluorescence microscopy (14, 15). White and coworkers (16) reported a unique electrochemical method to generate a single nanobubble on the surface of a metal nanoelectrode. Gas molecules are generated on the surface of the nanoelectrode from a gas-evolving redox reaction, such as the reduction of protons (17) and oxidation of H₂O₂ (18) and N₂H₄ (16). When the local concentration reaches a certain limitation, gas molecules nucleate and form a stable nanobubble covering the majority of the electrode surface.

We describe the use of superresolution fluorescence microscopy to image the dynamic nucleation and growth of hydrogen nanobubbles at the electrode/solution interface during electrochemical water splitting. This method is based upon a single-molecule labeling process illustrated in Fig. 1A: In electrochemical water splitting, water molecules are reduced on an indium–tin oxide (ITO) electrode (cathode) generating H₂ molecules at the electrode/solution interface, 2H₂O + 2e⁻ = H₂ (gas) + 2OH⁻; H₂ nanobubbles nucleate when high reducing potentials are reached; fluorescence dye molecules [e.g., Rhodamine 6G (R6G)] can transiently adsorb onto the nanobubble’s gas/solution interface and become momentarily trapped, enabling one to use total-internal reflection fluorescence (TIRF) microscopy to image nanobubbles (19). We found that H₂ nanobubbles can be labeled by single fluorophores, allowing us to use superresolution fluorescence microscopy to reveal their detailed interfacial dynamics. Our results show that hydrogen nanobubbles can be generated on an ITO electrode even at very early stages during water electrolysis, i.e., >500 mV before reaching the thermodynamic reduction potential of water. Moreover, the ability to image single-nanobubble nucleation enables us to observe in real time the process of electrocatalytic hydrogen spillover from ultrathin gold nanocatalysts.

Fig. 1.

(A) A scheme of the experimental setup used for imaging H₂ nanobubbles in electrocatalytic water splitting. H₂ nanobubbles generated on an ITO surface are labeled by single R6G molecules and imaged by TIRF microscopy. (B) A series of TIRF images of a 22.8 × 22.8-μm² area on an ITO electrode taken from a potential scan from 0 V to −2.0 V at 100 mV/s vs. Pt QRE in water containing 1 M Na₂SO₄ and 10 nM R6G. Fluorescence images were recorded at 19.81 frames per second with a 50-ms exposure time. (Scale bar, 5 μm.) (C) A comparison between the rate of nanobubble detection (detections per frame) (blue) and the current–voltage response recorded on the same electrode (black). The red dashed line indicates the baseline of the faradaic current. (D) A scatter plot showing the fluorescence intensity (counts) of individual H₂ nanobubbles in the potential range of interest from −0.5 V to −2.0 V from the same recording. The photon counts of each fluorescent burst on one frame were converted from the total integrated fluorescence signal under the fitted 2D Gaussian function using ThunderSTORM.

Materials and Methods

Chemicals and Materials.

All of the following chemicals and materials were used as received from the manufacturers: R6G perchlorate (Kodak, laser grade), sulforhodamine G (SRG) (Aldrich Chemical Co.; dye content ∼60%), poly(methyl methacrylate) (PMMA) (M_r ∼ 996,000; Aldrich), acetone (Fisher Chemical, 99.8%), sodium sulfate (Na₂SO₄; J. T. Baker, 101.8%), sulfuric acid (H₂SO₄, 5 mol/L; EMD Millipore Corporation), hydrazine (N₂H₄; Sigma-Aldrich, anhydrous 98%), isopropyl alcohol (IPA) (Fisher Chemical, 99.9%), 3-aminopropyltriethoxysilane (APTES) (Sigma, >98%), 5-nm–diameter Pt nanoparticles (NPs) (citrate capped, dispersed in 2 mM citrate; NanoComposix, Inc.), and ITO-coated microscope coverslips (SPI Supplies, sheet resistance 15–30 Ω/square). Deionized water (>18 MΩ·cm) was obtained through a Barnstead Nanopure water purification system and used for all aqueous solutions.

TIRF Microscopy.

Single-molecule TIRF imaging experiments were performed on a custom-modified Olympus IX70 inverted microscope configured for TIRF, using an Olympus Apo N 60× 1.49 NA objective and a 532-nm green laser (CrystaLaser) emitting at 10 mW (2.5 kW/cm²). An additional 1.5× magnification on the microscope was used. The fluorescence images were optically filtered with an ET590/50m emission filter (Chroma Technology) and acquired on an Andor iXon+ EMCCD camera cooled to −85 °C. Images were recorded by using an exposure time of 50 ms (frame rate: 19.81 Hz) or 10 ms (frame rate: 98.1 Hz). An amplifier gain of 300 as well as a preamplifier gain of 5.1 was used. The voltage function was generated by a 273A potentiostat (Princeton Applied Research) and applied across the working electrode (ITO) and the Pt quasi-reference electrode (QRE). A thin polydimethylsiloxane (PDMS) film with a 2-mm–diameter hole was attached to the surface of the ITO to define the area of the working electrode. A PCI-6251 (National Instruments) data acquisition card and a BNC-2090 breakout box were used to interface the potentiostat and the PC and to digitize the current–voltage signal.

Image Analysis.

Single-molecule fluorescence images were analyzed using the ThunderSTORM plug-in in ImageJ (20). Each fluorescent single-molecule spot is described by a point spread function (PSF), which is fitted with a 2D Gaussian function using maximum-likelihood estimation to achieve subdiffraction localization of single molecules. The photon counts of each fluorescent burst on one frame were converted from the total integrated fluorescence signal counts under the fitted 2D Gaussian function, using ThunderSTORM.

Single-Molecule Tracking.

Tracking of fluorescent puncta was performed using custom software (Insight3) (21). Briefly, this entailed the following key steps. Fluorescent puncta were detected above a user-set threshold (that was set to approximately eight times the SD of the background signal), fitted with a 2D Gaussian function to find centroid positions, and linked to puncta in adjacent frames which had displaced by 350 nm or less to form trajectories. Since the goal was to study moving particles, trajectories containing fewer than three consecutive localizations were discarded.

Scanning Electron Microscopy.

Scanning electron microscopy (SEM) imaging of gold nanoplates was performed on an FEI XL830 Dual Beam Focused-Ion Beam system.

Results and Discussion

Fig. 1B displays eight TIRF images taken from Movie S1, showing R6G-labeled H₂ nanobubbles on ITO as we scanned the electrode potential from 0 V to −2.0 V vs. a Pt QRE. Nanobubbles are seen as individual short fluorescence blinking events due to R6G molecules transiently binding onto and detaching from the nanobubble’s gas/water interface. The fluorophore–nanobubble interaction is at high frequency even in dye solutions of nanomolar concentrations: We estimate that a 10-nM R6G solution produces ∼700 collisions per second with a hemispherical 50-nm–radius nanobubble supported on a large electrode (SI Appendix, Fig. S1). Due to fast diffusion of dye molecules in and out of the evanescent field, R6G molecules cannot be individually resolved when no potential was applied or the potential was too low to reduce water (i.e., more positive than −0.75 V vs. Pt QRE) as shown by the dark background in Fig. 1B. Further increasing the potential in the negative direction allows for fluorescence imaging and counting of individual H₂ nanobubbles.

We can readily see more frequent nanobubble nucleation at higher negative potentials from Fig. 1B, which is due to an exponentially increasing rate of water reduction (22). We plotted the potential-dependent frequency of nanobubble detection, defined as the counted number of nanobubbles per frame (detections per frame), and the results are displayed in Fig. 1C. Also shown is the potential-dependent electrochemical current signal (black trace). A direct comparison like this reveals some striking differences between the fluorescence-based rate of nanobubble nucleation and the rate of the electrochemical reduction, i.e., faradaic current.

First, the faradaic current is not detectable until approximately −0.9 V and exhibits a slow increase due to low electrocatalytic activity of the ITO surface for water reduction. On the other hand, the optical signal starts to rise before −0.75 V (Fig. 1 B and C) and shows a significantly faster increase with voltage. In electrocatalytic water splitting, thermodynamically, it takes 1.23 V for water to be electrolyzed independent of the solution pH. In the electrolyte solution used here (deionized water, 1 M Na₂SO₄, 10 nM R6G), the most probable anodic reaction on the Pt QRE is water oxidation (2H₂O = 4H⁺ + 4e⁻ + O₂ (g)). The thermodynamic potential required to reduce water on the ITO (2H₂O + 2e⁻ = 2OH⁻ + H₂ (g)) is −1.23 V vs. Pt QRE. Our results have shown that H₂ nanobubbles start to be generated at the electrode/solution interface at a potential >500 mV more positive than the thermodynamic reduction potential of water. The potential drop in the ITO film is estimated to be around 10–20 µV, which is too small to cause any significant effect on the applied potential. Although it is often thought that nanobubble nucleation on an electrode surface requires a large current density to reach a high critical gas concentration (16), our results show H₂ nanobubbles can readily form on a macroscopic ITO electrode even when there is nearly no detectable faradaic current. We believe their formation is facilitated by the presence of active electrocatalytic sites on the surface of the ITO electrode. Despite the nearly unobservable faradaic signal on the electrode, the local reduction rate for water at these catalytic sites may be quite high, facilitating nanobubble nucleation. Identifying such active sites on a large heterogeneous electrode may require the combined use of high-resolution scanning probe methods. For example, Unwin’s group reported the presence of very active sites where oxygen evolution reaction can be observed before detecting general electrocatalytic current using scanning electrochemical cell microscopy (SECCM) (23).

Unlike the slowly increasing current signal, the detection frequency of H₂ nanobubbles exhibits a much faster increase before −1.4 V followed by a quick drop with potential (Fig. 1C). This unique peak-shape response indicates a competition between nanobubble nucleation and size growth as the potential is changed as illustrated in SI Appendix, Fig. S2. When a new H₂ molecule is formed on the electrode in the presence of surface nanobubbles, it may (i) diffuse laterally and join other H₂s to nucleate into a new nanobubble, (ii) diffuse laterally and join a nearby nanobubble, and (iii) diffuse into the bulk solution. The initial fast increase in nanobubble frequency indicates nanobubble nucleation dominates as more H₂s are generated on a relatively “clean” electrode. When the surface density of nanobubbles becomes sufficiently high, the rate of H₂ molecules finding a nearby nanobubble starts to increase, leading to nanobubble growth. This causes a further reduction in nanobubble–nanobubble spacing. Further increasing the rate of water reduction at higher negative potentials causes nanobubbles to coalesce.

The use of higher fluorophore concentrations allows us to image more frequent nanobubble generation as expected. SI Appendix, Fig. S3 displays the detection frequency of H₂ nanobubbles as a function of the R6G concentration. Here, H₂ nanobubbles were imaged on an ITO electrode at −1.0 V constant potential vs. Pt QRE in water containing 1 M Na₂SO₄ and R6G of various concentrations. It can be seen that more nanobubbles can be labeled and detected in solutions containing more fluorophores as expected. Although we expect that more nanobubbles can be seen at higher fluorophore concentrations, the fluorescent background also increases, preventing accurate counting of surface nanobubbles. We found 10 nM to be an optimized concentration for R6G fluorophores in most conditions.

Fluorescence-based nanobubble imaging can also be readily achieved with other fluorophores, e.g., resorufin, sulforhodamine G (SRG), and rhodamine B. SI Appendix, Fig. S4 displays imaging results of H₂ nanobubbles, using SRG. Unlike R6G, SRG is negatively charged at neutral pH. The fact that we can use both negatively and positively charged fluorophores to image H₂ nanobubbles on ITO suggests that electrostatic interactions (e.g., fluorophore–nanobubble or fluorophore–electrode interaction) may not be a governing factor in nanobubble labeling. Interestingly, it takes roughly 20 times more R6G molecules to label the same number of H₂ nanobubbles on the electrode at the same conditions (i.e., same electrode, potential, supporting electrolyte, etc.). This result suggests that SRG is more effective in labeling nanobubbles. Understanding the exact reasons for this difference requires a more systematic study using a series of dye molecules.

Although we focus on imaging H₂ nanobubbles in electrocatalytic water splitting in this study, the approach of dye labeling and fluorescence imaging is applicable to nanobubbles generated by other means, including freely diffusing nanobubbles obtained from gas-supersaturated solutions [i.e., mixing water with gas-saturated isopropanol (SI Appendix, Fig. S5) and interfacial nanobubbles generated by catalytic hydrazine decomposition (SI Appendix, Fig. S6)].

Fig. 1D is a scatter plot showing how fluorescence intensity of individual nanobubbles changes with potential as the electrode potential was scanned from 0 V to −2.0 V vs. Pt. A comparison of the fluorescence intensity of single nanobubbles and that of single R6G molecules immobilized on the ITO electrode surface (SI Appendix, Fig. S7) reveals that nanobubble intensities are all weaker than surface-immobilized single R6G molecules. Their weaker and somewhat uniform intensities at each potential make us believe that H₂ nanobubbles are labeled by single R6G molecules (the single-molecule characteristics are further discussed in the following paragraph). Their lower fluorescence intensity is due to the extra distance the R6G fluorophores have from the electrode surface when they are trapped on the nanobubble surface. In TIRF-based fluorescence imaging, the excitation light intensity decays exponentially with distance into the bulk solution, leading to lower fluorescence intensity for fluorophores located farther away from the electrode surface. In addition to their lower intensity, we found a clear decreasing trend of their average fluorescence intensity when electrode potential was increased in the negative direction. As is revisited and discussed in greater detail in the following three paragraphs, this decreasing intensity reveals key information about the nanobubble size and its potential dependence.

Fig. 2A displays intensity–time plots of randomly selected H₂ nanobubbles recorded at −1.0 V constant potential in solutions containing 1 M (Fig. 2A, Upper traces) or 1 μM Na₂SO₄ (Fig. 2A, Lower traces). More nanobubble plots are shown in SI Appendix, Fig. S8. Interestingly, we found that most nanobubbles (∼97%) show abrupt intensity increase indicating adsorption of a single fluorophore, followed by a subsecond period of near constant intensity and a sudden decay to the baseline. Because the average on time of fluorescent signal is much shorter than the time before photobleaching of surface-adsorbed R6G under similar conditions (∼0.021 s vs. ∼2.5 s; SI Appendix, Fig. S9) and electrochemically generated nanobubbles are quite stable when electrode potential is maintained (17), we believe the quick decay in fluorescence intensity is due to the departure of the fluorophore from the nanobubble surface. Although rare (∼3%), we occasionally observe nanobubbles labeled by multiple fluorophores as judged by their multilevel plots (Fig. 2B). The high single-molecule coverage suggests that molecule–molecule repulsions may prevent additional fluorophores from being adsorbed on the nanobubble surface.

Fig. 2.

Hydrogen nanobubbles are labeled by single fluorophores. (A) Fluorescence intensity−time traces showing five randomly selected nanobubbles detected at −1 V constant potential in water containing 1 nM R6G and 1 M (Upper traces) or 1 μM Na₂SO₄ (Lower traces). Traces were obtained by averaging the total intensity of a 6 × 6 pixel area around the center spot. (Scale bars, 50 fluorescence counts and 2 s.) (B) Fluorescence images and corresponding intensity traces (1 nM R6G and 1 μM Na₂SO₄) showing two representative two-step detections. Arrows mark the time points of the fluorescence images. (Scale bars, 100 counts and 0.1 s.) Each pixel size is 177.8 × 177.8 nm². Images were recorded at 98.1 frames per second with a 10-ms exposure.

Knowing H₂ nanobubbles are labeled by single fluorophores, we now examine their potential-dependent fluorescence response (Fig. 1D). The average fluorescence intensity per nanobubble decreases from ∼270 counts at −1.1 V to ∼70 counts at −2.0 V. Assuming nanobubbles are hemispherical (24), one can estimate their size from the known exponential decay of the evanescent wave intensity with distance (SI Appendix, Fig. S7 and Eqs. S3 and S4), giving average bubble radii increasing from 42 nm at −1.1 V to 90 nm at −2.0 V, suggesting that nanobubbles are larger at higher potentials. We believe the nanobubble’s larger size is caused by an exponentially increasing electro-reduction rate, which leads to the following: (i) H₂ will be “pumped” into the existing nanobubbles at a higher rate and (ii) more nanobubbles will nucleate, leading to faster coalescence.

We now show the ability to probe their size growth/shrinkage in real time during a voltage scan at a higher scan rate of 1.0 V/s. Fig. 3A displays five intensity-potential plots (marked in the images in Fig. 3B) showing changes in an individual nanobubble’s fluorescence intensity (from one trapped fluorophore) with voltage. Nanobubbles 1 and 2 were identified at around −1.0 V to −1.2 V and had an initial fluorescence intensity of ∼130 counts. Their intensities dropped toward the baseline before reaching −2.0 V, which must be caused by the nanobubble’s size growth at higher negative potentials “lifting up” the trapped fluorophores from the ITO. Nanobubble 3 shows both a decay stage (size growth) in the forward scan and an increase stage (size reduction) in the reverse scan. Nanobubbles 4 and 5 show only intensity increases (size reduction) with reducing potential. The size reduction is caused by a decrease in the electrochemical kinetics of water reduction.

Fig. 3.

Potential-driven nanobubble growth and shrinkage. (A) Fluorescence intensity−time traces of five nanobubbles seen in a potential scan (1 V/s) in water containing 1 M Na₂SO₄ and 10 nM R6G. (B) TIRF images taken at three potentials showing the locations of the five nanobubbles in A. (Scale bar, 5 μm.) Images were recorded at 19.8 frames per second with a 50-ms exposure time.

We then imaged the dynamic nucleation of H₂ nanobubbles from the hydrogen evolution reaction (HER) on ultrathin (∼30 nm) micrometer-sized gold nanoplates (25). Fig. 4A displays four TIRF images taken from a voltage scan (+0.5 V to −1.8 V, Movie S2) showing electrocatalytic HER and nanobubble formation in water containing 25 mM H₂SO₄, 1 M Na₂SO₄, and 5 nM R6G. Fig. 4B is an SEM image showing the microstructures of individual nanoplates at four locations in Fig. 4A, of which two nanoplates are seen partially stacked together at spot 3. Additional SEM images of other nanoplates are shown in SI Appendix, Fig. S10. Fig. 4C compares the frequencies of nanobubble nucleation from an area around a gold nanoplate (red box) and the bare ITO (yellow box). Nanobubbles are seen on both gold and ITO starting at −0.5 V and their detection frequencies in the two areas both increase with potential, indicating effective proton reduction. Interestingly, the detection frequencies are comparable in gold and ITO below −1.5 V. More frequent nanobubble formation can be seen in areas around gold nanoplates (spots 1 and 3) than on ITO at higher negative potentials (Fig. 4 A and C), which confirms the higher electrocatalytic activity of gold nanoplates toward proton reduction. It is interesting to see that the gold nanoplates at spots 2 and 4 did not show significant nanobubble activity, indicating they are inactive during this experiment possibly due to surface ligands or ineffective electric contacts with the ITO surface.

Fig. 4.

Imaging electrocatalytic HER and hydrogen spillover. (A) TIRF images (22.8 × 22.8 μm²) showing H₂ nanobubbles on a gold-nanoplate–modified ITO electrode in water containing 25 mM H₂SO₄, 1 M Na₂SO₄, and 5 nM R6G. The electrode potential was scanned at 100 mV/s from +0.5 V to −1.8 V vs. Pt QRE. The gold nanoplates exhibit plasmon luminescence even in the absence of an applied potential. The gold plates are highlighted with false color based on B, an SEM image of the same electrode area marked in A (blue box) showing the microstructures of the nanoplates. Spot 3 has two gold nanoplates partially stacked together. (C) A comparison of the rates of nanobubble detection from the two (5 × 5 μm²) boxes (red and yellow) marked in A. The gold nanoplates did not show significantly higher activity until passing −1.5 V. (D and E) Scatter plots showing the accumulated spatial distribution of H₂ nanobubbles in two potential windows: −1.2 V to −1.5 V (D) and −1.5 V to −1.8 V (E). Each colored dot represents one detected nanobubble. Images were recorded at 19.81 frames per second and a 50-ms exposure.

It is exciting to see a very large amount of nanobubbles generated on the ITO surface within a large distance of at least 3 μm around the gold nanoplates from spots 1 and 3. This can be clearly seen from the two images in Fig. 4A (third and fourth panels) and the accumulation plots shown in Fig. 4 D and E. We used the method of single-molecule tracking to examine the possible lateral motions of H₂ nanobubbles on the ITO electrode. Our results show that H₂ nanobubbles are remarkably stationary once generated on the ITO electrode (SI Appendix, Fig. S11). Our results confirm that the additional nanobubbles observed around the gold nanoplates are indeed nucleated on the ITO surface outside the gold. We believe they are formed due to the well-known hydrogen spillover effect (26–28), where H atoms generated on a catalyst surface migrate out onto the support substrate. As schematically illustrated in SI Appendix, Fig. S12, we envision that protons are electrochemically reduced to form H atoms, some of which combine into H₂ molecules and nucleate into nanobubbles on the surface of the gold nanoplates. When the surface coverage of H on gold reaches a certain extent, i.e., saturation, they start to migrate out onto the nearby ITO surface, where they combine into H₂ molecules, which then nucleate into H₂ nanobubbles. Although most previous studies on hydrogen spillover were conducted in the gas phase (29) and very few electrochemical studies were reported (30, 31), our research has provided real-time imaging evidence suggesting that hydrogen spillover can indeed take place in water electrolysis on ITO-supported gold nanocatalysts.

Using superresolution fluorescence microscopy, we can identify the exact location of each nanobubble on the electrode. This allows us to plot the locations of all of the detected nanobubbles in a certain potential window as shown in Fig. 4 D and E. By comparing the spatial distribution of H₂ nanobubbles in two potential windows −1.2 V to −1.5 V and −1.5 V to −1.8 V, we found that the effect of hydrogen spillover is most apparent at higher potentials. Nanobubbles are somewhat randomly generated on the gold-modified ITO electrode at lower negative potentials. At higher negative potentials, however, numerous additional nanobubbles are seen near the gold edges, indicating effective hydrogen spillover. This result suggests that the effect of hydrogen spillover may require that the catalyst surfaces reach a critical surface coverage of atomic H.

Conclusions

In summary, we have demonstrated that H₂ nanobubbles generated in electrochemical water splitting can be transiently labeled by single fluorophore molecules, allowing us to image their dynamics at the electrode/solution interface. We discovered that hydrogen nanobubbles can form even at very early stages in water electrolysis, i.e., >500 mV before reaching its thermodynamic reduction potential. The ability to use superresolution fluorescence microscopy to image single-nanobubble nucleation enabled us to observe the process of hydrogen spillover in electrocatalytic water splitting in real time from ultrathin gold nanocatalysts supported on an ITO electrode. Admittedly, the present method requires the use of optically transparent electrodes. Besides ITO, the technique can also be used on other electrodes, such as glass or quartz coated by thin films of fluorine-doped tin oxide (FTO), carbon, metals, and other conductive materials. This work opens up more research capacities for understanding dynamic nucleation of gases at various solid/liquid interfaces, especially at the electrode/solution interface.