Immobilization method dictates the photoelectrochemical glycerol oxidation performance of semiconductor catalyst assemblies

Summary

Photoelectrochemistry provides a direct route to convert sunlight into valuable chemicals. However, high-performance photoelectrode architectures often require complicated synthesis steps and expensive instrumentation. Although physical immobilization (e.g., spray coating) of catalyst particles on semiconductors seems to be a simple and universal approach, it is rarely implemented to prepare photoelectrodes. Here, we highlight challenges associated with such deposition strategies and demonstrate the importance of rational photoelectrode design. Specifically, we introduce metal interlayers (Pd, Au, or Ni) between n-type Si and a spray-coated PdAu catalyst. We show that these interlayers are essential for generating photopotential at the interface, while the PdAu catalyst governs C3 selectivity during glycerol oxidation. We also demonstrate that the interlayer determines key performance metrics (e.g., photopotential, photocurrent density, and stability). With this approach, the best-performing PdAu/Au/Si and PdAu/Ni/Si photoelectrodes can deliver high reaction rates (>100 mA cm⁻²) with sufficient stability even at higher illumination intensities (50 suns).

Graphical abstract

Highlights

• •Simple, scalable photoelectrode architecture for high-flux PEC glycerol valorization
• •Metal interlayer dictates photopotential, charge-transfer behavior, and stability
• •Spray-coated PdAu governs catalytic activity and selectivity
• •Stable PEC glycerol oxidation reaches over 100 mA cm⁻² under 50 suns

Balog et al. introduce a simple, scalable photoelectrode design for glycerol oxidation that enables stable operation at high current densities. The hybrid architecture is based on n-type Si, with an electrodeposited metal interlayer dictating photopotential and charge-transfer kinetics and a spray-coated PdAu catalyst layer governing C3 selectivity.

Introduction

Electrochemical (EC) valorization of organic compounds is gaining momentum in the production of valuable chemicals from less profitable or otherwise useless industrial by-products.^1,2,3 Although only electricity is needed to carry out EC processes, proper control over the product selectivity is yet to be achieved.^4,5 This is extremely important, as the separation of the broad pool of products might endanger the financial viability of the EC process.² As an example, the preparation of C3 products from glycerol oxidation can be highly profitable, as they serve as raw materials for several important polymers and cosmetics.⁶ Performing the oxidation process at an acceptable current density (i.e., at high anodic potentials) without breaking the C–C bonds, however, is a challenge⁷ and often requires noble-metal catalysts.^8,9 Photoelectrochemical (PEC) methods can provide a solution, since the reaction rate only slightly changes the product distribution.¹⁰ Additionally, the PEC scenario requires less electricity input and a lower amount of catalyst compared to dark EC processes.¹¹ Both are beneficial and reduce operational costs, potentially making PEC glycerol oxidation more cost effective compared to its purely EC and non-EC counterparts.¹²

Among different semiconductors (SCs), Si absorbs a large portion of the solar spectrum due to its narrow band gap.^13,14 As an indirect SC, its low absorption coefficient necessitates the use of a significant material thickness (>50 μm) to ensure full harvesting of the incident light.¹⁴ One key property for such SCs to be suitable as photoelectrodes is the lifetime of the light-generated charge carriers, which must be long enough to drive the desired redox reaction at the SC/electrolyte interface.^15,16 As a result of rapid recombination of light-generated electrons and holes (short lifetime), Si displays poor PEC performance without a catalyst.¹⁷ To further boost the photocurrent density and reach a sufficiently high photopotential value, additional interlayers can also be used.^18,19 Metal/Si heterojunctions can be employed to improve charge separation and transfer, leading to efficient reaction at the surface.²⁰ A suitable built-in electric field (and photopotential under illumination) can only be obtained when the work function of the metal is larger (i.e., more negative on the energy scale) than the Fermi level of n-type Si.^11,18,21 As an added benefit, these layers can also function as protective coatings for Si, which is otherwise susceptible to corrosion in contact with the electrolyte.²² Because the metal layer can simultaneously serve as the charge-extraction layer, the catalyst, and the protective coating, the requirements for these distinct roles may become intertwined and difficult to optimize simultaneously.²³ The main role of a catalyst is to reduce the overpotential by enhancing the rate of the EC reaction,²⁴ and it does not necessarily meet the requirements to build up a photopotential. Similarly, the catalyst is often employed as a particulate layer, where it cannot protect against corrosion or ensure efficient charge extraction.²⁵

In contrast to purely EC scenarios, spray coating (or other physical immobilization methods) is rarely applied for catalyst deposition on photoelectrodes. It is solely employed to deposit the SC scaffolds of the photoelectrode assemblies on the conductive substrates.^26,27 Physical catalyst immobilization strategies, however, offer the benefit that pre-synthesized metallic species (e.g., mono-, bi-, or multi-metallics) with precisely controlled size, morphology, and composition can be anchored on the photoelectrode surfaces. By separating the synthesis from the immobilization step, greater flexibility in the different processes can be achieved. The main drawbacks of physical immobilization are the difficulty of achieving thin catalyst layers with homogeneous coverage and the weak physical attachment of the catalyst to the SC surface (without post-annealing).²⁸ These issues are inherently remedied by EC deposition techniques, as a more uniform catalyst layer with a stronger interaction with the SC surface can be prepared.²⁹ In addition, surface defects (i.e., recombination centers) can be selectively eliminated by EC deposition, due to their increased EC activity and, therefore, higher rate of catalyst deposition at these sites.^30,31 The same EC deposition protocol, however, can lead to diverse morphologies on different surfaces, which can affect the overall PEC behavior and may also hinder future scale-up efforts.^32,33 Therefore, ensuring identical composition and morphology of (multi-metallic) catalysts on different SCs can be more easily achieved by spray coating of pre-synthesized or commercially available materials. In this case, achieving satisfactory PEC activity and stability is challenging. Combining EC deposition and spray coating can ensure the effective separation of the intertwined roles that metals can adopt in these architectures, enabling investigation of the different interfaces.

In this study, we demonstrate a photoelectrode design based on a physically immobilized PdAu catalyst operating under concentrated solar illumination (10 suns). We surmise that the use of metal interlayers (Pd, Au, or Ni) underneath the spray-coated PdAu catalyst is required for photopotential buildup. We show that the metal interlayer determines (1) the magnitude of the photopotential, (2) the achievable photocurrent density, and (3) the stability (at 25 mA cm⁻²), but it does not affect the product distribution (which is dictated by the catalyst). Finally, we test the most stable photoelectrodes (namely, PdAu/Au/Si and PdAu/Ni/Si) under 50-sun illumination to demonstrate that these electrodes can be used for long-term operation at reaction rates higher than 100 mA cm⁻².

Results and discussion

Characterization of the synthesized PdAu catalyst

Bimetallic PdAu particles were prepared in a 1:1 nominal ratio by simple wet-chemical synthesis. These particles were spray coated on an n-type Si wafer from a 15 wt % Sustainion ionomer containing suspension. Note that self-standing catalyst layers cannot be deposited on the Si surface without the use of an ionomer, as complete loss of the catalyst can be observed after immersing the ionomer-free PdAu-coated electrodes in the electrolyte. X-ray diffraction (XRD) measurements were carried out to investigate the structural characteristics of the synthesized PdAu catalyst. The obtained diffraction patterns verified that a polycrystalline bimetallic catalyst was produced (Figure S1). The XRD pattern was analyzed using the Le Bail fitting method, from which the average crystalline domain size was determined to be 9.5 nm. Transmission electron microscopy (TEM) measurements were performed to assess the average particle size of PdAu, which was 9.6 ± 2.6 nm (Figure S2), in excellent agreement with the domain sizes derived from the XRD fitting. The exact composition of the bimetallic catalyst was determined from energy-dispersive X-ray spectroscopy (EDX) and X-ray photoelectron spectroscopy (XPS) measurements. The atomic percentages of Pd and Au determined by EDX were 44% and 56%, respectively (first row of Table S1), which is in close agreement with the expected ratio from the synthesis procedure. In contrast, XPS results (Figure S3) show a Pd-rich surface, where a 57% Pd and a 43% Au amount were determined (first row of Table S2). No additional oxidized forms of the metals were detected on the surface of the PdAu/Si sample (Figure S3). Only slight shifts in the respective binding energies (to lower values) compared to their metallic counterparts were found (Table S3), caused by alloy formation.^34,35

EC and PEC oxidation of glycerol on spray-coated PdAu catalyst

In the first set of experiments, dark oxidation and photo-oxidation of glycerol were compared to scrutinize the catalytic performance of the spray-coated PdAu layers. Concentrated solar illumination was used in all PEC experiments, as it is the only way to achieve reaction rates comparable to those of the purely EC process.¹¹ Compared to illumination at 1-sun intensity, this can introduce additional challenges (e.g., accelerated photocorrosion or other degradation processes attributable to the elevated local temperature), making it important to examine both new and previously studied systems under such conditions. For the preparation of the dark electrode, the PdAu catalyst was spray coated directly on the indium (In)-covered printed circuit board (PCB) contact (to which the Si was soldered in the case of the photoelectrode). Rather surprisingly, linear voltammograms showed similar onset potentials of glycerol oxidation both in PEC and EC scenarios (Figure 1A). Ideally, the onset potential of the PEC process should be more negative than that of its dark EC counterpart containing the same catalyst, due to the photopotential generated in the SC near the interface upon illumination. It is noteworthy that a photocurrent is observed only when the photoelectrode surface is exposed to light (i.e., no current is detected under dark conditions, as shown in Figure 1A). This behavior arises from the fact that, in the dark, an n-type SC acts as an insulator (at potentials more positive than the flat band) and becomes conductive only under light exposure. In addition to exhibiting identical onset potentials, the dark electrode showed a higher achievable current density (Figure 1A). This trend is also reflected in the dynamic chronopotentiometric experiments at 25 mA cm⁻², where a lower (i.e., more negative) potential response was measured for the PdAu/In (Figure 1B). This indicates the absence of a photopotential and a more sluggish oxidation process in the case of the photoelectrode (a steeper current increase was observed on PdAu/In, as shown in Figure 1A). In conclusion, the PdAu deposited on the Si surface by spray coating does not, by itself, exhibit satisfactory PEC performance. UV photoelectron spectroscopy (UPS) measurements (Figure S4) also confirmed that the photopotential cannot build up in the PdAu-covered Si, as there are negligible differences between the Fermi levels of bare and PdAu-containing Si (see Figure S5 and the related discussion in supplemental information).

Figure 1.

Dark and photo-oxidation of glycerol on spray-coated PdAu catalyst

(A) Linear (photo)voltammograms recorded at a sweep rate of 10 mV s⁻¹. The photoelectrode was illuminated with 10-sun intensity. The electrolyte was 0.5 M glycerol containing 1 M CsOH.

(B) Dynamic chronopotentiometric measurements at 25 mA cm⁻² using −10 μA cm⁻² current pulses for 5 s every 50 s of photo-oxidation.

(C) Product distribution in the electrolyte after the dynamic chronopotentiometric experiments at 25 mA cm⁻². Error bars represent the standard deviation from the average, originating from three measurements.

As noble-metal catalysts (such as PdAu) display poor stability in glycerol oxidation as a result of surface poisoning (due to strongly adsorbed intermediates) and/or surface oxidation,^36,37 periodic pulses were applied during the current-controlled experiments (Figure 1B). This protocol promotes the removal of accumulated intermediates and restores the initial oxidation state of the catalyst.^37,38 The photoelectrode also showed lower stability compared to its dark counterpart (Figure 1B). Based on these observations, glycerol oxidation in the dark surpasses the PEC process in all aspects except the achievable product selectivity (Figure 1C). While nearly 100% total Faradaic efficiency (FE) toward profitable glycerol oxidation products was determined after the PEC reaction (with predominantly C3 selectivity), only ∼70% was identified in the dark EC process. The remaining roughly 30% FE can be ascribed to the formation of undesirable CO₂ (i.e., product of complete mineralization of glycerol), which is difficult to quantify in the highly alkaline electrolyte.^39,40,41 The significant difference in product distribution between PEC and EC glycerol oxidation can be explained by the difference in surface potential in the two scenarios. While the surface potential of the photoelectrode remains constant, pinned to the valence-band position of the SC,^11,37 it varies with the applied bias in the case of the dark electrode.

Effect of metal interlayers between Si and PdAu on PEC performance metrics

To address the drawbacks of the PdAu/Si system described above, three different metal layers (Pd, Au, or Ni) were pre-deposited on the surface of Si, resulting in a Schottky-type electrical contact rather than simple physisorption.¹⁸ Subsequently, the PdAu catalyst was spray coated on top of these metal-coated Si electrodes from a suspension. Therefore, its attachment to the metal interlayer is dominated by physical adhesion and ionomer-assisted binding rather than direct chemical bonding. The rationale behind this composite architecture was the hypothesis that the metal layers can ensure the buildup of photopotential (owing to their favorable—i.e., more negative on the vacuum scale—Fermi level alignment with respect to Si, as shown in Figure S6), while the spray-coated PdAu provides the catalytic activity and selectivity. The PEC glycerol oxidation performance of the different metal layers containing photoelectrodes was probed by photovoltammetry and compared with PdAu deposited directly on the Si surface (Figure 2A). Lower onset potentials and higher photocurrent densities were achieved (up to a ∼2.5-fold improvement compared to PdAu/Si, as shown in Figures 2A and 2B) with all metal sublayer-containing electrodes. The latter observation (i.e., the higher reaction rate) can be surprising, as, in principle, fewer photons can reach the photoabsorber through the multi-layered structure. To quantify the parasitic light absorption/reflection caused by the metal sublayer, an amount of Ni equivalent to that used on Si was deposited on a transparent fluorine-doped tin oxide (FTO) electrode, and its transmittance was measured. Although most of the light loss originates from the spray-coated PdAu catalyst, the measurement revealed an additional ∼20% loss of incident light due to the presence of the Ni interlayer (Figure S7). Consequently, the intermediate metal layers reduce the amount of light reaching the SC, thereby lowering its effective light absorption. Nevertheless, the photocurrent remains higher relative to PdAu/Si, most likely due to enhanced charge-carrier separation (i.e., suppressed recombination). The type of the metal interlayer between the Si and the spray-coated PdAu catalyst had an impact on both the onset potential (and thus the photopotential) and the photocurrent density. PdAu-coated Au/Si showed the best performance in both aspects (with an onset of −0.7 V vs. Ag/AgCl and a maximum current density of ∼110 mA cm⁻² at +0.2 V vs. Ag/AgCl), followed by the Ni- and Pd-layer-containing photoelectrodes with the same onset of −0.6 V (vs. Ag/AgCl) and a maximum current density of ∼80 and ∼70 mA cm⁻² (at +0.2 V vs. Ag/AgCl), respectively. To demonstrate that the presence of the metal interlayer only influences the onset potential of the photoelectrodes (by enabling photopotential buildup), the EC activity of PdAu was compared after spray coating on the In contact, Ni sheet, and Ni-coated FTO substrates. To eliminate the impact of the slight variations in current density, the linear sweep voltammetry curves were normalized, clearly indicating that under dark conditions the onset potential is independent of the substrate (Figure S8A). In contrast, the differences in the onset potential for the photoelectrodes persist even after normalization (Figure S8B). Accordingly, the potential values measured at 25 mA cm⁻² followed the sequence of the onset values (Figures 2C and 2D). PdAu/Si and PdAu/Pd/Si photoelectrodes, however, exhibited unstable behavior even after a few minutes of photo-oxidation, while there was no change in the performance of photoelectrodes with Ni and Au interlayers during a 30-min current-controlled experiment (Figure 2C).

Figure 2.

Effect of metal interlayers between PdAu and Si on the PEC performance under 10-sun illumination

(A) Photovoltammograms for PdAu-coated Si, Pd/Si, Ni/Si, and Au/Si electrodes recorded at a sweep rate of 10 mV s⁻¹. The electrolyte was 0.5 M glycerol containing 1 M CsOH.

(B) Maximum photocurrent densities derived from the photovoltammograms at +0.2 V (vs. Ag/AgCl). Error bars represent the standard deviation from the average, originating from at least three measurements.

(C) Dynamic chronopotentiosmetric experiments at 25 mA cm⁻² using −10 μA cm⁻² current pulses for 5 s every 50 s of photo-oxidation.

(D) Potential values derived from the chronopotentiometric measurements after 50 s of photo-oxidation at 25 mA cm⁻².

Error bars represent the standard deviation from the average, originating from at least three measurements.

Scanning electron microscopy images were recorded to examine the surface coverage and visualize the morphology of the different PdAu-coated photoelectrodes (Figures S9 and S10). The electrodes are indistinguishable from each other based on the lower-magnification images; however, incomplete coverage of the PdAu catalyst can be observed in all cases (Figure S9). On the high-magnification images, the underlying metal layers (consisting of fine grains) are visible on the uncovered parts of the respective electrodes (insets in Figure S10). As a result of the imperfect catalyst coverage, the exact Pd/Au ratio on the surface of the Pd and Au interlayers containing electrodes determined by EDX (Table S1) and XPS (Table S2) is skewed toward the respective metal used as the interlayer. This also suggests that the interlayers are in direct contact with the electrolyte during the PEC experiments, which can be one of the reasons for the differences observed between the photoelectrodes. To investigate the participation of the interlayers as a catalyst in the glycerol oxidation process, the PEC performance of Pd/Si, Ni/Si, and Au/Si was compared with their PdAu catalyst-containing counterparts (Figure S11). Although Pd/Si showed some PEC activity, the adsorbed intermediates and/or surface oxidation (Figure S11A) rapidly deteriorated its performance. The more negative onset potential and the higher photocurrent density of PdAu/Ni/Si compared to Ni/Si also underline that the exposed Ni interlayer in the architecture cannot be the reason for the PEC activity, as Ni/Si shows a large overpotential in glycerol oxidation (Figure S11B). In contrast, Au/Si showed remarkable activity toward glycerol oxidation, even outperforming its PdAu/Au/Si counterpart (Figure S11C). Thus, the performance achieved with PdAu/Au/Si may reflect contributions from both the PdAu catalyst and the exposed Au interlayer.

Based on the reaction mechanism established under alkaline conditions (Figure S12),^11,37 the product distributions determined after dynamic chronopotentiometric experiments at 25 mA cm⁻² for all PdAu-containing photoelectrodes were compared with those obtained on Au/Si and Ni/Si (Figures 3A, S13, and S14). Pd/Si was omitted from this comparison, as the aforementioned current density cannot be achieved with this electrode (Figure S11A). PdAu-containing electrodes showed similar product distribution both in the absence and presence of the metal interlayers. This indicates that the spray-coated PdAu catalyst is responsible for the selectivity. Mainly C3 products (glycerate, tartronate, and lactate) were obtained on these electrodes, with a total FE above 70% (Figure 3B). In contrast, a drastic decrease in the C3 selectivity was observed on Au/Si, which was even more pronounced on the Ni/Si electrode (Figure 3B). In this latter case, C1 formate became the main product with an FE of nearly 70%. These prominent differences in the selectivity further strengthen the negligible catalytic effect of the underlying metal layers in the presence of PdAu.

Figure 3.

Product distribution in the electrolyte after PEC oxidation of glycerol

(A) FEs of the detected products after dynamic chronopotentiometric experiments at 25 mA cm⁻² for PdAu-, PdAu/Au-, PdAu/Ni-, PdAu/Pd-, Au-, and Ni-catalyst-coated Si photoelectrodes under 10-sun illumination in 0.5 M glycerol containing 1 M CsOH. Error bars represent the standard deviation from the average, originating from at least three measurements.

(B) FEs of C3, C2, and C1 products on different catalysts derived from (A).

To better understand the trend in the photocurrent densities (shown in Figures 2A and 2B), electrochemical impedance spectroscopy (EIS) measurements were carried out at open circuit potential (OCP) under 10-sun illumination (Figure 4). Two semicircles were observed in the Nyquist plot of all the PdAu-containing photoelectrodes (Figure 4A). The high-frequency semicircle is related to the charge transfer from the Si to the metal interlayers (or to the electrolyte in the case of PdAu/Si), while the second low-frequency semicircle indicates the further charge transfer to the spray-coated PdAu catalyst.^42,43 To determine the charge-transfer resistance (R_ct) values for the two processes (Figure 4B), the Nyquist plots consisting of two semicircles were fitted using a series R_s–(R_ct,1||Q₁)–(R_ct,2||Q₂) equivalent circuit (Figure S15), where R_s represents the series resistance, R_ct the charge-transfer resistances, and Q the constant phase elements. Lower R_ct values were determined in the presence of the metal interlayers, a trend reflected in the obtained maximum photocurrents, where a lower R_ct results in a higher photocurrent.

Figure 4.

EIS measurements at OCP under 10-sun illumination

(A) Nyquist plots for PdAu-coated Si, Pd/Si, Ni/Si, and Au/Si electrodes; the inset visualizes the semicircles at high frequencies. The electrolyte was 0.5 M glycerol containing 1 M CsOH.

(B) R_ct values derived from the Nyquist plots by fitting the two semicircles of each electrode (at high and low frequencies) with a series R_s–(R_ct,1||Q₁)–(R_ct,2||Q₂) equivalent circuit.

To prove that the two semicircles (and, therefore, the determined R_ct values) are related to the aforementioned charge-transfer processes, additional EIS measurements were performed at −0.5 V (vs. Ag/AgCl), where a significant band bending is expected (the flat-band potential of Si is −0.82 V [vs. Ag/AgCl]³⁷). The magnitude of the high-frequency semicircles remained nearly unchanged compared to the ones measured at OCP (Figure S16), signaling that the charge transfer from the Si to the metal interlayers is not (or only slightly) affected by the band bending (Figure S17A). This shows that the light-generated holes can efficiently transfer to these metal layers even at OCP. However, they rapidly recombine with the electrons from the conduction band (as shown schematically in Figure S18A), preventing them from reaching PdAu. This is reflected in the large second semicircles observed at low frequencies without external bias (Figures S16 and S17B). In contrast, the recombination can be suppressed by applying −0.5 V (vs. Ag/AgCl) to the photoelectrodes, due to the increased band bending (by introducing an energy barrier for electrons to leave the Si photoabsorber, as shown in Figure S18B). Consequently, the light-generated holes can reach the PdAu catalyst through the metal layers, indicated by the drastically decreased low-frequency semicircles (Figure S16) and the related R_ct values (Figure S17B).

The kinetics of charge-carrier transfer and recombination were further studied using intensity-modulated photocurrent spectroscopy (IMPS). IMPS spectra were recorded in the potential range between −0.8 and +0.2 V (vs. Ag/AgCl) only for the photoelectrodes with metal interlayers (Figure S19). The instability of PdAu/Si due to corrosion processes occurring in the absence of protective metal coatings inhibited its evaluation. Between −0.8 and −0.5 V (vs. Ag/AgCl), recombination dominates the response of the photoelectrodes, indicated by the almost perfect circles in the IMPS spectra (Figures S19A–S19C). As the potential was increased, the diameter of the semicircles in the upper quadrants decreased, signaling suppressed recombination and increased charge extraction (Figures S19D and S19F). Although the semicircles in the upper quadrants can still be observed at the most positive potentials, their negligible size (compared to the semicircles in the lower quadrants) suggests that the charge transfer is the dominating process. The determined charge transfer (k_tr) and surface recombination (k_sr) rate constants (from IMPS) were plotted as a function of the applied potential (Figure 5). The highest k_tr was observed for PdAu/Au/Si, slightly lower for PdAu/Ni/Si, and about half as much for PdAu/Pd/Si (Figure 5A). In contrast, the k_sr of PdAu/Au/Si was slightly lower than that of PdAu/Ni/Si, while the lowest value was obtained for PdAu/Pd/Si (Figure 5B). The lower k_tr and higher k_sr of PdAu/Ni/Si compared to PdAu/Au/Si result in lower photocurrents. Also, the effect of the lower k_tr of PdAu/Pd/Si is mitigated by the lower k_sr; thus, the difference in the photocurrent density (compared to the other two electrodes) is lower than what would be expected from the k_tr values alone (especially at the most negative potentials). Therefore, the observed trend in the photocurrent densities can be explained by the branching ratio between these two processes.

Figure 5.

Rate constants for PdAu-coated Pd/Si, Ni/Si, and Au/Si electrodes at different potentials, derived from IMPS measurements in 0.5 M glycerol containing 1 M CsOH

(A) Charge-transfer rate constants as a function of potential.

(B) Recombination rate constants as a function of potential.

PEC stability at exceptionally high reaction rates

PdAu/Au/Si and PdAu/Ni/Si showed the best stability under 10-sun illumination (Figure 2C) at an applied current density of 25 mA cm⁻². To assess the stability of these photoelectrodes at higher reaction rates, additional PEC experiments were performed under both 10 and 50 suns (Figure 6). With increasing illumination intensity, a dark current was observed. Furthermore, the achieved photocurrent densities with PdAu/Au/Si (Figure 6A) and PdAu/Ni/Si (Figure 6B) were only up to 2.5 times higher at 50 suns. Both phenomena are likely associated with the strong heating effect of the 50-sun illumination intensity (causing a ∼20°C temperature increase in 20 cm³ of electrolyte during a single photovoltammogram), which affects the overall reaction kinetics. In addition, at higher illumination intensities, the increased generation rate of charge carriers may lead to their accumulation and enhanced recombination, further contributing to the observed nonlinear photocurrent response. Potentiostatic measurements were performed at −0.08 V (vs. Ag/AgCl) with chopped illumination to prevent rapid temperature rise of the electrolyte and allow sufficient time for the desorption of adsorbed intermediate products from the PdAu surface.³⁷ Similarly to the photovoltammograms, the photocurrents approximately doubled when the light intensity was increased from 10 to 50 suns (Figures 6C and 6D). In addition, both photoelectrodes exhibited stable performance during a 10-min measurement, regardless of the illumination intensity. Based on these results, increasing the light intensity is a good strategy to achieve higher reaction rates; however, proper temperature control is required to minimize the heating effect of the illumination.

Figure 6.

Comparison of PEC performance between PdAu/Au/Si and PdAu/Ni/Si electrodes under 10- and 50-sun illumination

(A and B) Photovoltammograms recorded at a sweep rate of 10 mV s⁻¹ for (A) PdAu/Au/Si and (B) PdAu/Ni/Si.

(C and D) Potentiostatic measurements at −0.08 V (vs. Ag/AgCl) for (C) PdAu/Au/Si and (D) PdAu/Ni/Si using chopped light illumination.

Overall, an efficient photoelectrode design was developed for the PEC oxidation of glycerol, which allows the use of physically immobilized catalysts that can offer stable operation even under concentrated solar illumination. Specifically, different metal/Si heterojunctions (namely Pd/Si, Au/Si, and Ni/Si) were created via spontaneous galvanic displacement or electrodeposition, on which the PdAu catalyst (with a ratio of 1:1) was deposited by spray coating. The metal interlayers beneath the PdAu allowed the buildup of a photopotential, while the PdAu catalyst was responsible for the catalytic properties (i.e., activity and selectivity). The formation of C3 products with an FE of nearly 70% was observed in all PdAu-containing electrodes regardless of the type of metal interlayer. However, the value of the photopotential, the achievable photocurrent density, and the stability of the photoelectrodes were impacted by the metal layer used. The variation in the photopotential was attributed to the Fermi-level differences between bare and metal-interlayer-coated Si photoelectrodes, as measured by UPS. The trend in the photocurrent density (PdAu/Au/Si > PdAu/Ni/Si > PdAu/Pd/Si > PdAu/Si) was explained by EIS and IMPS measurements. The best-performing PdAu/Au/Si and PdAu/Ni/Si showed stable PEC behavior at 25 mA cm⁻² under 10-sun illumination, and this stability was maintained at even higher light intensities where the photocurrent exceeded 100 mA cm⁻² for both electrodes.

While the results presented here demonstrate the promise of the approach, a comprehensive assessment of its limitations and future optimization strategies is still required. A key limitation of the present work is the lack of long-term stability testing under well-controlled thermal conditions. Future studies should therefore focus on the design of flow PEC cells with precisely controlled temperature, including regulation of the inlet electrolyte temperature, as well as on optimization of the electrolyte flow rate to maintain a nearly constant local temperature even under concentrated illumination, thereby improving long-term stability. Importantly, the simplicity of the photoelectrode fabrication may facilitate straightforward scale-up for future applications. In addition, enhancing C3 selectivity will likely require systematic optimization of the hydroxide concentration in the electrolyte and the ionomer content within the PdAu catalyst layer. These considerations provide a clear pathway toward the development of stable and scalable PEC systems for selective glycerol valorization.

Methods

Materials

All chemicals were purchased from commercial suppliers (Sigma-Aldrich and VWR International), were of at least analytical grade, and were used without further purification. MilliQ grade (ρ = 18.2 MΩ cm) ultrapure deionized (DI) water was used to prepare all the solutions.

Electrode preparation

A single-side polished n-type Si wafer (Silicon Materials; dopant phosphorus, resistivity 0.3–0.5 Ω cm, thickness 525 μm) was used as the photoactive material, which was contacted on a copper-based PCB using molten In. Prior to contact preparation, 1 × 1-cm pieces were cut from the Si wafers and were sonicated in acetone, isopropanol, and DI water for 3 min. In the next step, the native oxide layer from the surface was removed by immersing the Si pieces into buffered oxide etchant (BOE; consisting of 10 g NH₄F·HF, 15 cm³ H₂O, and 2.5 cm³ concentrated HF) solution for 1 min and rinsed with water. During the contact preparation, both the backside of Si and the PCB were coated by molten In on a hotplate at 180°C (at which temperature the In becomes spreadable). The two In-covered parts were then pressed together and cooled down to room temperature (by removing the assembly from the hotplate). After the contact was prepared, the whole electrode assembly was immersed again in the BOE solution for 10 s to remove the oxide that may have formed during the high-temperature contact preparation. In the case of metal interlayer-containing electrodes, different metals were deposited on the surface of Si. Ni layer was electrodeposited from stirred 0.1 M NiSO₄ and 0.1 M H₃BO₃ containing aqueous solution using a two-step potentiostatic method: −1.4 V (vs. Ag/AgCl) was applied for 1 s, then −1.0 V (vs. Ag/AgCl) was used until 25 mC cm⁻² charge density was passed. Au and Pd layers were electroless deposited by immersing the Si electrodes in the precursor containing solution for 5 min. The water-based Au solution contained 0.1 M K₂SO₄, 1 mM H₂SO₄, 1 mM KCl, and 1 mM HAuCl₄, while the Pd-containing aqueous solution consisted of 0.1 M K₂SO₄, 2 mM HCl, 1 mM KCl, and 1 mM PdCl₂. As the last step, the PdAu catalyst was spray coated onto the surface of the bare and metal interlayer-containing Si electrodes at 100°C from an ethanol-based 1 mg cm⁻³ suspension together with 15 wt % ionomer (Sustainion XB-7). The spray coating was performed with a hand-held airbrush with compressed-air carrier gas, and the catalyst loading was 141 ± 12 μg cm⁻².

Catalyst synthesis

For the synthesis of PdAu bimetallic particles, 0.0177 g of PdCl₂ was dissolved in 10 cm³ of 0.02 M HCl solution and stirred at 350 rpm for 1 h. Thereafter, 26 cm³ DI water and 4 cm³ 25 mM HAuCl₄ were added to the Pd stock solution (Pd/Au molar ratio 50:50). To form the PdAu particles, 10 cm³ 0.1 M NaBH₄ (PdAu/NaBH₄ ratio 1:5) was added dropwise into the solution containing Pd and Au ions (with stirring at 900 rpm). The obtained PdAu suspension was stirred at 350 rpm for 1 h. In the cleaning step, the sedimented PdAu particles were washed four times with 40 cm³ DI water and two times with 20 cm³ ethanol. Finally, the particles were redispersed in 10 cm³ ethanol using an Elmasonic P 30H (Elma Schmidbauer) ultrasonic bath (30 min, 37 kHz, and 100% power) and an UP200St (Hielscher Ultrasonics) ultrasonic homogenizer (10 min, 26 kHz, 60% amplitude, and 100% pulse). The mass concentration of the samples was determined by evaporating 1 cm³ of PdAu suspension.

Characterization methods

The surface morphology and the coverage were studied by a Thermo Fisher Scientific Apreo C scanning electron microscope equipped with an Everhart-Thornley detector. The microscope was operated at 25 pA current and 10 kV acceleration voltage. EDX measurements were performed at a current of 0.2 nA and an acceleration voltage of 20 kV using a Bruker X-ray detector.

The crystal structure of the synthesized PdAu catalyst was analyzed by XRD using a Rigaku MiniFlex II desktop diffractometer equipped with a Co K_α (λ = 1.79 Å) X-ray source. Diffraction patterns were recorded in the 30°–70° 2θ range with a scanning speed of 0.4°/min. The divergence slit, receiving slit, and anti-scattering slit were set to 1.25°, 0.3 mm, and 1.25°, respectively. For the XRD measurement, the PdAu catalyst was spray coated onto a glass slide, and the amorphous nature of this substrate caused a broad background increase in the 30°–40° 2θ range. The XRD pattern was fitted using the Le Bail method in GSAS-II software to obtain the average crystalline domain size. A CIF file from the Crystallography Open Database (COD #1510274) was used for the fitting. The final fit resulted in a χ² value of 1.7.

The particle size of the PdAu catalyst was determined by TEM using a JEOL JEM-1400Plus instrument (JEOL, Tokyo, Japan) operating at an accelerating voltage of 120 kV. Samples were prepared by depositing 10-μL aliquots of the suspension on 200-mesh copper grids coated with a carbon-supported film and allowing them to air dry.

The surface composition of the spray-coated PdAu catalyst was examined by XPS using a SPECS instrument equipped with a Phoibos 150 MCD 9 hemispherical analyzer. The analyzer was operated in fixed transmission mode with 40 eV pass energy for acquiring survey scans. For the high-resolution scans, 20 eV pass energy was used. Al K_α radiation (hν = 1,486.6 eV) was used as an excitation source and operated at 150-W power. Ten scans were averaged to obtain a single high-resolution spectrum. For spectrum evaluation, the CasaXPS commercial software package was used. Further details regarding the XPS fitting can be found in supplemental methods.

The Fermi level of Si and its change due to the presence of the metal interlayers were determined by UPS with He(I) excitation (21.22 eV). External bias of 10 V was applied to the samples to accelerate secondary electrons to the analyzer. To remove the surface carbon layers from the samples, gentle Ar⁺ bombardment was carried out with 500 eV of energy for 2 min. The recorded UPS spectra were corrected for additional He(I) satellites. The work function of the samples was determined from the second derivative of the secondary electron cutoff at large binding energies, and the Fermi level was extracted from fitting the first derivative of the small binding energy region to a Gaussian function. The valence band of Si was determined from fitting the onset of the small binding energy region.

The light absorption/reflection properties of the Ni interlayer with and without the PdAu catalyst were measured with an Agilent 8453 UV-visible diode-array spectrophotometer in the 300- to 1,000-nm wavelength range.

EC experiments and product analysis

All experiments were performed in a one-compartment glass cell using a Metrohm Autolab PGSTAT302 type potentiostat/galvanostat in a standard three-electrode setup. Si-based photoelectrodes were employed as the working electrodes, a Pt sheet as the counter electrode, and 1 M NaOH containing Hg/HgO as the reference electrode. For the sake of simplicity, the potential values were shifted to the Ag/AgCl (3 M NaCl) scale during the data evaluation. Therefore, the potentials mentioned in the text and shown in the figures can be converted to the reversible hydrogen electrode (RHE) scale using the following equation: E_RHE = E_Ag/AgCl + 0.210 V + 0.059 × pH (where pH is 14). Before all experiments, 0.5 M glycerol containing 1 M CsOH electrolyte was carefully de-aerated by argon gas, and argon bubbling was carried out continuously during the reactions to exclude the dissolution of oxygen from the air and to ensure adequate convection. During the PEC measurements, a solar simulator (Sciencetech SL-50A-WS, AM1.5) was employed as the light source with a power density of 1,000 mW cm⁻², while the light-chopping frequency was 0.20 Hz to track the photocurrent and dark current simultaneously. Linear (photo)voltammograms were recorded with a sweep rate of 10 mV s⁻¹. For the PEC measurements at 5,000 mW cm⁻², a custom-designed high-flux solar simulator purchased from Sciencetech was employed.

EIS spectra were measured at OCP and at −0.5 V (vs. Ag/AgCl) with added 10 mV RMS perturbation in the frequency range between 100 kHz and 0.1 Hz (OCP) or 1 Hz (−0.5 V), recording 10 points per frequency decade.

IMPS measurements were carried out using a Metrohm Autolab PGSTAT302N type potentiostat/galvanostat equipped with an FRA32 module and a light-emitting diode (LED) driver kit. Spectra were recorded between −0.8 and +0.2 V (vs. Ag/AgCl) at 100-mV steps in the frequency range of 20 kHz to 0.1 Hz. 200 mW cm⁻² white light illumination was used in these experiments with added 20 mW cm⁻² sinusoidal modulation (10%). Normalization of the measured signal was carried out by determining the number of the incident photons employing a calibrated silicon photodiode and assuming that the LED is monochromatic at its wavelength maximum (450 nm). From this, the corresponding maximum photocurrent was calculated using the assumption that one photon generates one electron-hole pair.

The product distribution after the PEC experiments was analyzed by a Bruker Avance Neo 500 NMR spectrometer and a Shimadzu Prominence LC-20AD liquid chromatograph (for high-performance liquid chromatography [HPLC]) equipped with a ReproGel H (9.0 μm, 300 × 8 mm) column and an SPD-M20A diode array detector. During the HPLC measurements, 10 mM sulfuric acid eluent was used with a flow rate of 0.8 cm³ min⁻¹ and column temperature of 45°C. Before the HPLC analysis, 1 cm³ of undiluted electrolyte and 250 μL of 2.0 M sulfuric acid were mixed, and 20 μL of sample was injected into the column from this mixture. For NMR measurements, 45 μL from the sample prepared for HPLC, 405 μL of H₂O, and 50 μL of phenol and DMSO containing D₂O were mixed in the NMR tubes. The one-dimensional ¹H spectrum was measured using a solvent presaturation method to suppress the water peak. After the concentrations of the different products were determined and the electrolyte volume considered, the amounts of substance (in mol) were calculated. Multiplying these by the number of electrons required for the formation of each product (Table S4) and by the Faraday constant, the charge associated with the formation of the respective product was obtained. FE values were then calculated as the charge passed for the formation of each product divided by the total charge passed. The selectivity of each product was determined as the molar fraction of the given compound relative to the total amount of quantified organic products. The yield of each product was calculated as the ratio of the molar amount of the product to the initial moles of glycerol present in the electrolyte.

The error bars in the figures represent the standard deviation from the average (mean) originating from at least three separate measurements performed with at least three different electrodes.

Resource availability

Lead contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Dr. Csaba Janáky (janaky@chem.u-szeged.hu).

Materials availability

There are restrictions to the availability of PdAu catalyst due to the lack of an external centralized repository for its distribution and our need to maintain stock. We are glad to share the catalyst suspension with reasonable compensation by requesters for its processing and shipping.

Data and code availability

• •Data reported in this paper will be shared by the lead contact upon request.
• •This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

This project has received funding under the European Union’s Horizon Europe research and innovation program from the European Research Council (ERC, grant agreement no. 101043617). Project no. RRF-2.3.1-21-2022-00009, titled National Laboratory for Renewable Energy, has been implemented with support provided by the Recovery and Resilience Facility of the European Union within the framework of the program Széchenyi Plan Plus. The authors thank Dr. Viktória Hornok (University of Szeged) for performing TEM measurements and Dr. Szilárd Sáringer (University of Szeged) for his valuable assistance in the characterization of the PdAu suspension used in this work.

Author contributions

Conceptualization, Á.B., G.F.S., E.C., and C.J.; data curation, Á.B.; methodology, Á.B., G.F.S., N.V., E.C., and C.J.; formal analysis, Á.B., G.F.S., D.F., and C.H.; investigation, Á.B., G.F.S., D.F., C.H., and N.V.; visualization, Á.B. and G.F.S.; writing – original draft, Á.B.; writing – review & editing, Á.B., S.G.F., and C.J.; funding acquisition, C.J.; resources, E.C. and C.J.; supervision, E.C. and C.J.

Declaration of interests

The authors declare no competing interests.

Published: August 1, 2026