Tuning the Selectivity of Glycerol Electrooxidation via Catalyst Composition and Operating Conditions

Abstract

The electrooxidation of glycerol presents a promising route for upgrading a biodiesel byproduct into high‐value chemicals while lowering the energy cost of electrochemical hydrogen production, but its mechanistic understanding remains incomplete. This work investigates the influence of temperature and catalyst composition on glycerol electrooxidation using three distinct electrocatalysts: first‐row transition‐metal NiCo oxide, noble metal AuIn alloy, and hybrid PdNi alloy. Electrochemical, in situ UV–vis reflectance and in situ Raman measurements reveal catalyst‐specific oxidation mechanisms: an oxyhydroxide‐mediated path for NiCo, an adsorbed hydroxide‐mediated path for AuIn which suffers from inhibition, and a dual mechanism for PdNi. Oxyhydroxide mechanisms favor formic acid, whereas adsorbed hydroxide pathways selectively promote C₃ products such as glyceric acid and dihydroxyacetone. Flow cell electrolysis across a range of temperatures and flow rates relevant to industrial application shows that elevated temperature enhances reaction rates and mitigates catalyst inhibition, a relevant effect for the hybrid catalyst. These findings provide key mechanistic insights and practical guidelines for optimizing continuous‐flow glycerol electrooxidation systems.

Temperature has a high impact on glycerol oxidation processes both in the macroscopic scale, increasing reaction rates and reducing cell potentials, and in the microscopic scale, affecting the reaction mechanism to improve electrode stability and modify product distribution. We studied these effects over three distinct catalysts bridging the gap between laboratory and industrial scales by using in situ characterization and flow cell electrolysis.

1. Introduction

In recent decades, glycerol production has been on an upward trend due to the growth of the biodiesel industry [1, 2]. However, this increase in production has not been matched with industry demand, causing glycerol to become a low‐value product and be considered a residue [3, 4]. Consequently, there has been an incentive to develop processes that use glycerol as a feedstock and incorporate it into the circular economy [5]. One of these processes is the substitution of the oxygen evolution reaction (OER) with the glycerol oxidation reaction (GOR) in industrial electrochemical processes such as hydrogen production or carbon dioxide reduction [6, 7, 8]. The use of GOR results in a lower cell voltage because glycerol has a lower oxidation potential than water, thereby decreasing the overall energy cost of the process [9, 10, 11].

Additionally, glycerol can be oxidized into a variety of high‐added‐value products in the complex chain of reaction pathways shown in Scheme 1 [12, 13]. Glycerol oxidation starts with alcohol oxidation, producing either glyceraldehyde or dihydroxyacetone, depending on which OH group is oxidized. Both compounds can undergo a C─C bond cleavage to produce glycolic acid, glycolaldehyde, and formic acid. However, they also have their own reaction pathways. Glyceraldehyde can undergo further oxidation to form glyceric acid, which in turn is oxidized to formic acid [14]. Meanwhile, dihydroxyacetone can undergo dehydration followed by a Cannizzaro rearrangement to produce lactic acid, which can be further oxidized to yield acetic acid and formic acid [15]. Therefore, the products can be separated into four groups: C3_DHA (dihydroxyacetone, lactic acid, and acetic acid), C3_GLAD (glyceraldehyde and glyceric acid), C2 (glycolic acid and glycolaldehyde), and C1 (formic acid). Finally, formic acid can be oxidized into CO₂; however, it is considered a nonvaluable product and is usually left out of product characterization.

SCHEME 1.

Scheme of different glycerol oxidation reaction pathways and product categorization. Product categories are C3_GLAD (glyceraldehyde (GLAD) and glyceric acid (GLA)), C3_DHA (dihydroxyacetone (DHA), lactic acid (LA), and acetic acid (AC)), C2 (glycolaldehyde (GOAD) and glycolic acid (AHA)), and C1 (formic acid (FA)).

Various materials have been studied for glycerol oxidation, where two broad categories can be discerned. On the one hand, noble metals such as Au, Pt, or Pd tend to favor the production of less oxidized products like glyceric acid or dihydroxyacetone and operate at low potentials [16, 17, 18, 19]. However, these elements are scarce, which increases the cost of manufacturing catalysts. They can also suffer from inhibition due to surface oxidation. On the other hand, first‐row transition metals, such as Ni, Co, or Cu, tend to favor the production of more oxidized products such as formic acid, but suffer from high operational potentials [20, 21, 22, 23, 24, 25]. However, since these metals are more abundant, their manufacturing cost is considerably lower. To achieve low manufacturing cost and reduced operational potential, hybrid catalysts combining both families of materials—such as PtNi, PtCo, or PdNi—have been developed. These catalysts maintain or even improve the catalyst's electrochemical performance while reducing the manufacturing costs [26, 27, 28, 29, 30].

The different performance of the noble metal catalysts and first‐row transition metals is explained by differences in their glycerol oxidation mechanisms. Noble metal catalysts oxidize glycerol via an adsorbed hydroxide mechanism, where they adsorb hydroxide species onto their surface, which subsequently oxidize glycerol into the corresponding ketone, aldehyde, or carboxylic acid as follows [31, 32]:

$$\mathrm{M}\hspace{0.17em}+\hspace{0.17em}{\mathrm{OH}}^{-}\hspace{0.17em}\to \hspace{0.17em}\mathrm{M}\hspace{0.17em}-\hspace{0.17em}{\mathrm{OH}}_{\mathrm{ad}}\hspace{0.17em}+\hspace{0.17em}{\mathrm{e}}^{-}$$ (I)

$$\begin{array}{c}{\mathrm{RCH}}_2\mathrm{OH}\hspace{0.17em}+\hspace{0.17em}\mathrm{M}\hspace{0.17em}-\hspace{0.17em}{\mathrm{OH}}_{\mathrm{ad}}\hspace{0.17em}+\hspace{0.17em}{\mathrm{OH}}^{-}\hspace{0.17em}\hspace{0.17em}\\ \to \hspace{0.17em}\mathrm{RCHO}\hspace{0.17em}+\hspace{0.17em}2\hspace{0.17em}{\mathrm{H}}_2\mathrm{O}\hspace{0.17em}+\hspace{0.17em}\mathrm{M}\hspace{0.17em}+\hspace{0.17em}{\mathrm{e}}^{-}\end{array}$$ (II)

$$\mathrm{RCHO}\hspace{0.17em}+\hspace{0.17em}3\hspace{0.17em}{\mathrm{OH}}^{-}\hspace{0.17em}\to \hspace{0.17em}{\mathrm{RCOO}}^{-}\hspace{0.17em}+\hspace{0.17em}2\hspace{0.17em}{\mathrm{H}}_2\mathrm{O}\hspace{0.17em}+\hspace{0.17em}2\hspace{0.17em}{\mathrm{e}}^{-}\hspace{0.17em}$$ (III)

Meanwhile, first‐row transition metals are not able to generate this adsorbed hydroxide layer and thus need to be oxidized into their oxyhydroxide state (M³⁺), which then reacts with glycerol as follows [20, 33]:

$$\mathrm{M}{\left(\mathrm{OH}\right)}_2\hspace{0.17em}+\hspace{0.17em}{\mathrm{OH}}^{-}\hspace{0.17em}\to \hspace{0.17em}\mathrm{MOOH}\hspace{0.17em}+\hspace{0.17em}{\mathrm{H}}_2\mathrm{O}\hspace{0.17em}+\hspace{0.17em}{\mathrm{e}}^{-}$$ (IV)

$$\begin{array}{c}\begin{array}{l}\mathrm{MOOH}\hspace{0.17em}+\hspace{0.17em}{\mathrm{RCH}}_2\mathrm{OH}\hspace{0.17em}+\hspace{0.17em}{\mathrm{OH}}^{-}\hspace{0.17em}\\ \to \hspace{0.17em}\mathrm{M}{\left(\mathrm{OH}\right)}_2\hspace{0.17em}+\hspace{0.17em}{\mathrm{H}}_2\mathrm{O}\hspace{0.17em}+\hspace{0.17em}\mathrm{RCHO}\hspace{0.17em}+\hspace{0.17em}{\mathrm{e}}^{-}\end{array}\end{array}$$ (V)

$$\mathrm{RCHO}\hspace{0.17em}+\hspace{0.17em}3\hspace{0.17em}{\mathrm{OH}}^{-}\hspace{0.17em}\to \hspace{0.17em}{\mathrm{RCOO}}^{-}\hspace{0.17em}+\hspace{0.17em}2\hspace{0.17em}{\mathrm{H}}_2\mathrm{O}\hspace{0.17em}+\hspace{0.17em}2\hspace{0.17em}{\mathrm{e}}^{-}$$ (VI)

Although both mechanisms involve the formation of intermediate species, since they are formed at different applied biases, it is not surprising that they produce different types of products. However, we should not assume that the oxidation mechanism is the only factor in determining the reaction selectivity, since mass transport and intermediate product adsorption also affect reaction pathways [34, 35, 36]. This means that studying the behavior of an electrocatalyst only under static conditions can lead to misleading results as it can lead to an underestimation of Faradaic efficiencies as well as an inaccurate analysis of product distribution, as the reaction products remain in the solution throughout the electrolysis and can be further oxidized to carbon dioxide, diminishing the overall Faradaic efficiency [37, 38]. Therefore, it is essential to evaluate the electrode performance under continuous operating conditions, which more accurately reflect industrial applications [39]. In addition to the intrinsic selectivity of the electrocatalyst, product distribution is also affected by operational parameters such as reactor size, flow rate, temperature, and current density [40, 41]. Temperature is a particularly relevant parameter, as it enhances reaction kinetics, enabling higher current densities, increasing process output, and lowering cell voltages, thereby reducing the energy cost of the process [42].

Because of that, in this work, we investigate how the two glycerol oxidation mechanisms can be enhanced by modulating the operational conditions, mainly focusing on the electrolyte temperature and flow rate. To study the two aforementioned mechanisms, we considered three electrocatalyst families: a nickel–cobalt oxide representing first‐row transition metal catalysts, which performs the oxyhydroxide mechanism; a gold indium alloy as the noble metal catalyst, which performs the adsorbed hydroxide mechanism; and a palladium nickel alloy as a hybrid catalyst that combines the properties of the two previous families. This approach enables a comparative analysis of their electrochemical behavior and performance in flow‐cell electrolysis, as well as the interplay between thermal effects and the glycerol oxidation mechanisms.

2. Experimental Section

2.1. Electrode Synthesis

For the electrode synthesis, the following materials were used: NiSO₄·6H₂O (Merck, 99%), CoSO₄·6H₂O (Alfa‐Aesar, 98%), thiourea (Merck, 99%), KCN (Merck, pa), D‐glucose (Merck, 99.5%), H₂SO₄ (Panreac, 95%), In₂(SO₄)₃ (Merck, 98%), KOH (Merck, 99.9%), K[Au(CN)₂] (Plating Decor Recubrimientos, S.L, 99%), Ni(NO₃)₂ (Probus, 97%), PdCl₂ (Merck, 99.9%), NaCl (Panreac, p.a.), and HCl (Panreac, b.p.). The electrodes were deposited over two distinct materials, a flat iron sheet (Ossian Hull cell panels, Ossian Lagerqvist AB, Järfälla, Sweden) for voltammetric measurements and a Ni‐foam porous substrate support (Recemat Ni‐4753, 1.6 mm thick) for electrolysis. Prior to deposition, the supports were first cleaned with a diluted HCl solution, rinsed with water and ethanol, and dried. Unless specified otherwise, a titanium iridium oxide mesh was used as the counter electrode.

To synthesize the NiCo electrodes, a bath consisting of 0.033 M NiSO₄·6H₂O, 0.066 M CoSO₄·6H₂O, and 0.1 M thiourea was prepared. A NiCo hydroxide was electrodeposited by applying −1 V versus Ag/AgCl (3.5 M KCl) for 15 min. Then, the NiCo hydroxide electrodes were thermally treated at 300 °C for 2 h to obtain a NiCo oxide electrode.

To synthesize the AuIn electrodes, a bath consisting of 0.5 M KCN, 0.1 M D‐glucose, 0.025 M In³⁺, 1.3 M KOH, and 0.025 M K[Au(CN)₂] was prepared following a previously reported method in a two‐electrode cell configuration [37]. The bath was maintained at 70 °C, a current density of −140 mA cm⁻² was applied for 70 s for the iron sheet, and a current density of −80 mA cm⁻² was applied for 120 s for the Ni‐foam support.

To synthesize the PdNi electrodes, a bath consisting of 0.1 M Ni(NO₃)₂, 1.7 mM PdCl₂, 0.48 M NaCl, and 0.144 M HCl was prepared following a previously reported method in a two‐electrode cell configuration [29]. Using two graphite plates connected in parallel as the counter electrode and with constant stirring, the electrode was left for 60 s at open‐circuit potential. Then, a current density of −30 mA cm⁻² was applied for 120 s for the iron sheet, and a current density of −50 mA cm⁻² was applied for 120 s for the Ni‐foam support.

2.2. Structural Characterization

The morphology of the electrodes was determined via field emission scanning electron microscopy (FESEM) using a JEOL‐7100, and the crystalline structure was analyzed via X‐ray diffraction (XRD) on a PANalytical X’pert PRO diffractometer with monochromatized Cu Kα radiation (λ = 1.5406 Å), operating at 45 kV and 40 mA in a Bragg–Brentano configuration.

2.3. Electrochemical Measurements

Electrochemical measurements were done in a single‐compartment cell, in a three‐electrode cell configuration. The electrolyte was deoxygenated under an argon atmosphere, with the flat iron electrodes as the working electrode, a graphite counter electrode, and a reversible hydrogen reference electrode. Capacitance measurements were recorded in a 1 M KOH (Merck, 99.9%) electrolyte at five scan rates ranging from 10 to 50 mV s⁻¹. Cyclic voltammograms were recorded in a 1 M KOH (Merck, 99.9%) electrolyte in the absence and presence of 0.1 M glycerol (Sigma‐Aldrich, 99%), at three different temperatures (25, 40, and 60 °C) and at 50 mV s⁻¹. Some cyclic voltammograms were normalized by the electrochemically active surface area (ECSA) calculated using a C_s value of 0.04 mF cm⁻² [43, 44].

2.4. UV–vis Reflectance In Situ Measurements

In situ UV–vis reflectance measurements at 500 nm recorded simultaneously with the cyclic voltammograms used the same three‐electrode cell configuration from Section 2.2 in a custom‐made Teflon cell with two fused‐silica windows at 90° from each other. The incidence angle on the electrode surface was therefore 45°. A Hg‐Xe lamp (150 W) was used as the light source, and the wavelength was selected with a monochromator (Model 9055, Sciencetech). The beam was focused on the electrode surface with a lens. The reflected beam was directed to a photomultiplier tube (PMT Hamamatsu R955) powered by a high‐voltage power supply (Brandernburg 475R). The photocurrent generated at the PMT was transformed to a voltage by a 1 kΩ resistor connected to the PMT's output and fed into the auxiliary input of the potentiostat, allowing the simultaneous recording of the reflectance dependence on the electrode potential. Reflectance data were normalized by dividing by the reflectance value at the starting potential of the cyclic voltammogram, which coincided in most cases with the maximum reflectance.

2.5. Raman Spectroscopy In Situ Measurements

In situ Raman spectroscopy measurements were carried out on a Renishaw Raman spectrometer with a 1200 L/mm grating using a Leica L63x waterproof immersion objective and a 785‐nm laser at 5% power. The measurements were done in a homemade electrochemical cell with a Ag/AgCl (3.5 M KCl) leak‐free and a Pt wire as the reference and counter electrode, respectively. The spectra were collected at steady state under different applied potentials ranging from 0.3 to 1.7 and back to 0.3 V versus RHE with a 0.1 V step between applied potentials.

2.6. Flow Cell Electrolysis

Flow cell electrolysis was done in a stack setup as described in Figure 1a. Both the anode and the cathode had the same geometric area (10 cm²); the anode consisted of the selected electrocatalyst supported over Ni foam, and the cathode the bare Ni foam support. The two compartments were separated with a Nafion 117 membrane, and a 1 M KOH and 1 M glycerol electrolyte were pumped into both compartments at flow rates of 5.7 and 17.1 mL min⁻¹. For each set of operational parameters, two electrolysis experiments were carried out, and the anolyte was collected at 30 and 60 min, obtaining a total of 4 aliquots per operational parameter. A two potentiostat system was used; a first potentiostat was connected in a three‐electrode configuration to an Ag/AgCl (3.5 M KCl) leak‐free reference electrode and measured the open‐circuit potential; meanwhile, the second potentiostat was connected in a two‐electrode configuration and applied a current density of 90 mA cm⁻² (0.9 A) for 1 h as described in Figure 1b. Due to the PdNi electrode inhibition in some electrolysis, a potential of −0.4 V_AgCl was applied for 60 s before starting the electrolysis to ensure that the material was in its metallic state.

FIGURE 1.

(a) Flow cell internal stack setup. Blue pieces represent the cathodic compartment, red pieces represent the anodic compartment, and the yellow piece represents the Nafion 117 membrane. Green arrows represent the 1 M KOH and 1 M glycerol electrolyte flow path. (b) Flow cell external setup. Light blue square represents the two‐electrode setup, and dark blue square represents the three‐electrode setup.

2.7. Product Characterization

The flow cell electrolysis products were determined using a combination of high‐performance liquid chromatography (HPLC) and proton nuclear magnetic resonance (H‐NMR) techniques. HPLC was conducted on a HPLC WATERS 1525 instrument with a carboxylic acid column (Bio‐Rad‐Aminex HPX 87H, 300 mm × 7.8 mm) at a temperature of 35 °C with a 20‐µL sample loop. Diluted sulfuric acid solution (5 mM) was used as the mobile phase at a flow rate of 0.6 mL min⁻¹. The products were identified using a WATERS 2998 UV–vis spectrometer set at 210 and 270 nm. H‐NMR analysis was conducted on a Bruker 400 using a water‐suppression method [45]. Sample preparation consisted of 490 µL of sample, 90 µL of deuterated water, and 20 µL of 0.2%_vol dimethyl sulfoxide (DMSO) solution, used as an internal standard. Each product's Faradaic efficiency was calculated using the following equation:

$${\mathrm{FE}}_i\hspace{0.17em}\left(\mathrm{\%}\right)=\hspace{0.17em}\frac{z\hspace{0.17em}F\hspace{0.17em}{n}_i}{I}\hspace{0.17em}100$$ (1)

where z is the number of electrons involved in the oxidation reaction according to the chemical equations described in the supporting information (Equations S1–S5), F is the Faraday constant (96,485 C mol⁻¹), n _i is the molar flow rate of the obtained product (mol s⁻¹), and I is the total current supplied by the potentiostat (0.9 A). Global Faradaic efficiency was calculated as the sum of the Faradaic efficiencies of all the individual products as shown in Equation (2).

$$\mathrm{FE}\left(\mathrm{\%}\right)=\sum {\mathrm{FE}}_i$$ (2)

Production, selectivity, and energy consumption were calculated using the equations detailed in the Supporting Information (Equations S6–S9).

3. Results and Discussion

3.1. Electrode Selection

This study examines three representative materials from different broad families of electrocatalysts with distinct electrocatalytic properties and reaction mechanisms. Nickel cobalt oxide (NiCo) was chosen for the first‐row transition‐metal catalyst group due to its high selectivity toward formic acid, although it exhibits activity only at high overpotentials [20]. For the noble metal catalyst group, a gold–indium alloy (AuIn) was selected for its low GOR overpotential, despite not having the highest selectivity [37]. Finally, a palladium–nickel alloy (PdNi) was chosen from the hybrid catalyst group for its low GOR onset potential and high selectivity toward three‐carbon products, such as dihydroxyacetone and glycerate [29].

The electrocatalysts were deposited over two different supports: a flat iron sheet (Fe sheet) for the electrochemical characterization experiments and nickel foam (Ni‐foam) for the flow cell electrolysis experiments. This choice reflects the requirements of the respective techniques, as both reflectance and Raman spectroscopy require flat surfaces for an adequate signal, whereas the flow cell electrolysis requires a porous support to facilitate electrolyte flow. FESEM imaging (Figure 2) shows that all three materials present a similar morphology regardless of the support. NiCo has a plain‐threaded morphology, while AuIn and PdNi display a rough cauliflower‐like structure, with AuIn showing a larger grain size than PdNi. EDS measurements (Table S1) confirmed the correct deposition of the three electrocatalysts on both supports, as the catalyst metals were detected in all cases. However, due to the contribution of the support, the atomic composition could not be accurately determined. Capacitance measurements corroborated the higher active area of the Ni‐foam support compared to the Fe sheet (Figure S1 and Table S2).

FIGURE 2.

FESEM imaging of the three electrocatalysts deposited over (1) Nickel foam and (2) iron sheet supports. (a) NiCo with a plain threaded morphology, (b) AuIn with a cauliflower morphology, and (c) PdNi with a cauliflower morphology.

The crystalline structure of the electrocatalysts was analyzed by XRD using the samples deposited on the flat iron sheet support. As shown in Figure S2a, the support exhibits three clear peaks that correspond to cubic iron (JCPDS 00‐001‐1262). Figure S2b shows that, after the thermal treatment, the NiCo catalyst possesses a cubic CoNiO₂ structure (JCPDS 00‐010‐0188) with additional peaks corresponding to the oxidation of the support into iron oxide (JCPDS 00‐001‐1111). Figure S2c shows that the AuIn catalyst presents a hexagonal Au₇In₃ alloy composition (JCPDS 00‐029‐0647) with the presence of various iron oxides. Finally, Figure S2d shows that the PdNi catalyst has three peaks that can be assigned to metallic cubic Pd (JCPDS 00‐002‐1439). However, these peaks are slightly shifted to higher diffraction angles, and no nickel peaks are detected, suggesting the formation of a PdNi alloy with a palladium‐rich composition.

3.2. Electrochemical Characterization

Figure 3 shows the cyclic voltammograms of the three selected materials in a 1 M KOH electrolyte, both with and without 0.1 M glycerol. As expected, the NiCo electrode (Figure 3a) exhibits the typical behavior of the oxyhydroxide mechanism. In the absence of glycerol, we can see an oxidation peak preceding the OER corresponding to the formation of the oxyhydroxide species near 1.4 V_RHE (Equation I), and in the reverse scan, reduction of the oxyhydroxide results in a peak just above 1.3 V_RHE. In the presence of glycerol (dotted voltammogram), the GOR onset potential closely coincides with the onset of the oxyhydroxide formation since this species is the intermediate for glycerol oxidation (Equations II and III). In contrast, the AuIn electrode (Figure 3b) exhibits the typical behavior of the adsorbed hydroxide mechanism. In the absence of glycerol, we can only see the oxidation of the surface at around 1.2 V_RHE. Meanwhile, in the presence of glycerol, the electrode starts being active around 0.6 V_RHE, but loses its activity once its surface becomes oxidized. On the reverse scan, we can see that the activity of the electrode is recovered once the surface gets reduced back to its metallic state.

FIGURE 3.

Cyclic voltammograms normalized by ECSA recorded at a scan rate of 50 mV s⁻¹ for (a) NiCo, (b) AuIn, and (c) PdNi electrodes in 1 M KOH electrolyte (full line) and 1 M KOH + 0.1 M glycerol electrolyte (dotted line).

Finally, the PdNi electrode (Figure 3c) shows characteristics of both the oxyhydroxide and the adsorbed hydroxide mechanisms of the GOR, indicating that this hybrid material is able to perform both oxidation mechanisms. Below 1 V_RHE, glycerol oxidation begins at a less negative potential (0.5 V_RHE) than that on the AuIn electrode, but its electrocatalytic activity is also lost much earlier (0.8 V_RHE), coinciding with the onset of surface oxidation. Such inhibition of the GOR is consistent with an adsorbed hydroxide mechanism. Above 1 V_RHE, oxidation and reduction of oxyhydroxide species are observed in the glycerol‐free electrolyte at around 1.3 V_RHE. These features disappear in the presence of glycerol, indicating an oxyhydroxide‐mediated pathway. Thus, depending on the applied potential, the PdNi electrode behaves either as a noble metal catalyst, oxidizing glycerol via an adsorbed hydroxide mechanism, or as a first‐row transition‐metal oxide, operating via the oxyhydroxide mechanism.

To determine the individual effect of the nickel and palladium sites, we deposited Ni and Pd electrodes using the PdNi bath with only nickel or palladium in it and performed cyclic voltammetries with and without glycerol. In Figure S3a, we can see that nickel sites are inactive until the formation of the oxyhydroxide peak and subsequent OER evolution starts. Meanwhile, palladium sites experience a continuous oxidation that seems to increase around the oxyhydroxide formation potentials. This seems to imply that palladium is also being oxidized into active species for glycerol oxidation, which is corroborated in Figure S3b, where we can see that all three materials are active at the oxyhydroxide formation potentials. This shows that in the PdNi electrode, at low potentials, glycerol oxidation is performed exclusively by the palladium active sites, whereas at high potentials, glycerol oxidation is performed by both nickel and palladium active sites.

3.3. UV–vis and Raman In Situ Measurements

To better understand the relationship of the adsorbed hydroxide mechanism and surface oxidation, we coupled the AuIn and PdNi voltammograms with in situ reflectance and Raman measurements over a wide potential range. In glycerol‐free 1 M KOH, the reflectogram of AuIn (Figure 4a) shows a linear decrease in reflectance in the positive scan with an increasing potential between 0.1 and 1 V_RHE. This is mainly caused by the electroreflectance of gold, which is maximal around 500 nm and is essentially due to a decrease in the free electron density on the metal surface as the potential increases [46, 47], indicating the absence of gold oxide on the electrode surface over this potential range. However, whereas pure gold exhibits a reflectance decrease of <1% over 1 V_RHE [48], AuIn shows a larger decrease over a similar potential interval. This suggests that In contributes to the surface electroreflectance, probably through affecting the electronic structure of Au. The onset of gold oxidation around 1.2 V_RHE coincides with an accelerated decrease in reflectance, which is consistent with the formation of a less reflective gold oxide layer on the surface. This is corroborated by in situ Raman spectroscopy (Figure S4a), where a broad peak at 600 cm⁻¹, corresponding to gold oxide, appears at 1.4 V_RHE and disappears when the surface is reduced back to metallic gold at 1.3 V_RHE [49]. When the scan is reversed, the reflectance remains initially constant, as no more gold oxide is being formed, until reduction of the gold oxide provokes an increase in reflectance. However, after oxidizing the surface beyond 1.1 V_RHE, the reflectance of AuIn remains lower in the negative‐going scan than in the positive‐going one, unlike pure Au, which shows a reflectance overlap in the negative‐ and positive‐going sweeps after the reduction peak. This suggests a considerable degree of irreversibility of surface oxidation, most likely due to the formation of indium oxide. Increasing the anodic limit from 1.4 to 1.65 V_RHE (Figure S5a), where the OER starts, leads to a further decrease in the surface reflectance, which implies that the extent of surface oxidation is determined by the upper potential limit.

FIGURE 4.

Reflectograms at λ = 500 nm and cyclic voltammetry recorded at a scan rate of 50 mV s⁻¹ of (a,b) AuIn and (c,d) PdNi in 1 M KOH without (a,c) and with (b,d) 0.1 M glycerol.

Figure 4b shows that for AuIn, glycerol oxidation starts at around 0.6 V_RHE, followed by a rapid surface deactivation for GOR just below 1.3 V_RHE. Such deactivation coincides with a sharp decrease in reflectance due to sudden oxide formation. Moreover, the superposition of the glycerol and glycerol‐free reflectograms (Figure S6a) reveals that the presence of glycerol delays surface oxidation by approximately 0.1 V, although once initiated, oxidation is completed quickly, reaching a similar extent at 1.4 V_RHE in both electrolytes. This potential delay is also seen in the in situ Raman measurements (Figure S4b), where the gold oxide band appears at 1.5 V_RHE and disappears at 1.4 V_RHE in the glycerol‐containing electrolyte, 0.1 V higher than in the absence of glycerol. This suggests that adsorbed OH is needed for both glycerol and surface oxidation. However, since the latter process requires a critical OH coverage to begin (Equation VII), the constant consumption of the adsorbed OH by glycerol oxidation inhibits surface oxidation [48].

$$\mathrm{M}-{\mathrm{OH}}_{\mathrm{ad}}\hspace{0.17em}+\hspace{0.17em}{\mathrm{OH}}^{-}\hspace{0.17em}\to \hspace{0.17em}\mathrm{MO}\hspace{0.17em}+\hspace{0.17em}{\mathrm{H}}_2\mathrm{O}\hspace{0.17em}+\hspace{0.17em}{\mathrm{e}}^{-}\hspace{0.17em}$$ (VII)

Looking at the behavior of AuIn voltammetry, just before the inhibition happens, it seems that the decline in the current density could be linked to insufficient glycerol mass transport. Because of that, we performed additional voltammetries under stirring (Figure S7a), which showed that this is indeed the case. This shows that the inhibition of the AuIn electrode happens due to an accumulation of adsorbed hydroxides. However, even in the absence of mass transport limitations, it is also possible to observe an inhibition of the surface if the potential is high enough. This indicates that since the oxidation involves an electron transfer, as the potential increases, the required OH_ad coverage will decrease until the process is thermodynamically favorable. This leads to the buildup of the oxide layer and quenching of the electrocatalytic activity.

During the reverse scan, the sudden anodic current spike in the cyclic voltammogram coincides with a sudden increase in the surface reflectance, evidencing that the surface reactivation results from reduction of the surface oxide. Additionally, the reactivation is delayed at higher voltammetric limits, due to an increased buildup of the oxide layer (Figure S5b). The reflectograms and Raman spectra further show that oxide reduction occurs at a less negative potential in the presence of glycerol, providing additional evidence of the competition between glycerol and surface oxidation for adsorbed OH. Interestingly, in the presence of glycerol, the forward and reverse scans of the reflectogram overlap in the potential region below 1 V_RHE, which may indicate that glycerol inhibits indium oxidation, suggesting that indium oxide could remain active toward glycerol oxidation.

Figure 4c shows that, in the absence of glycerol, the PdNi electrode exhibits two oxidation processes: oxidation of palladium and formation of oxyhydroxide species. When the positive limit of the voltammogram is set at 1.2 V_RHE (Figure S8a), where only palladium oxidation occurs, the reflectogram shows the expected continuous decrease in reflectance due to the formation of a surface palladium oxide layer. When the scan is reversed and the oxide is reduced, the reflectance returns to its initial value. Increasing the positive limit of the voltammogram to 1.4 V_RHE leads to an increase in reflectance, associated with the second oxidation process. This can be attributed to the oxidation of insulating oxides to metallically conducting oxyhydroxides, which is confirmed by the corresponding increase in reflectance. These two processes are also observed in the in situ Raman spectra (Figure S9a): palladium oxidation is indicated by the appearance of a small peak at 530 cm⁻¹ starting at 0.9 V_RHE [50], and the oxyhydroxide formation is evidenced by the appearance of two extra peaks at 470 and 550 cm⁻¹ which mask the palladium oxide peak starting at 1.5 V_RHE [51].

Figure 4d shows that, similarly to AuIn, the loss of activity in the case of PdNi is due to the formation of palladium oxide. At potentials where palladium remains metallic, the reflectance‐potential dependence is similar to that of AuIn, albeit with a smaller electroreflectance response. The onset of Pd oxide formation shifts to more positive potentials in the presence of glycerol, which, like in the case of AuIn, implies a competition for adsorbed OH between glycerol and surface oxidation (Figure S6b). However, this difference is smaller than that observed for AuIn and is not clearly resolved in the in situ Raman spectra (Figure S9b). Moreover, under stirring (Figure S7b), the PdNi electrode does not experience a delay in activity loss, indicating that it is not as affected by mass transport limitations as the AuIn electrode. When the potential is positive enough to form oxyhydroxide species, the reflectance increases again, but to a lesser extent than in the absence of glycerol. This strongly supports that oxidation of glycerol in this potential region occurs via the reaction with oxyhydroxide species, which are reduced back to oxide in the process (Equation V).

Figure S8b shows that, for PdNi electrodes, the onset potential of glycerol oxidation decreases with the upper potential limit of voltammetry. This suggests that oxyhydroxide formation may induce structural changes that slightly change the catalyst's performance. To study this effect, the PdNi electrode was cycled in a 1 M KOH electrolyte until stabilization, changing the limit from 1.2 V_RHE to 1.4 V_RHE and back to 1.2 V_RHE. Figure S10 shows that increasing the upper limit shifts the palladium reduction onset potential, demonstrating that oxyhydroxide formation leads to structural changes. However, this change is reversible, as the initial reduction onset potential is restored when the upper limit is returned to 1.2 V_RHE. This could be explained by a surface rearrangement via an interfacial place‐exchange mechanism during oxidation, with different anodic limits producing a distinct local environment of the palladium atoms, resulting in a slight shift in their reduction potential.

3.4. Effect of Temperature

Figure 5 shows the effect of temperature on glycerol oxidation for the three studied materials. As expected, increasing temperature leads to higher currents and lower onset overpotentials; however, distinct behaviors are observed between the oxide and metallic electrocatalysts. The current increase of the NiCo oxide electrode can be linked to the advancement of the oxyhydroxide formation peak from 1.4 V_RHE at 25 °C to 1.2 V_RHE at 60 °C, as can be seen in Figure S11a. This proves that glycerol oxidation is mediated by the oxyhydroxide phase formation, which is reduced back to the Ni(II)/Co(II) oxide and subsequently reoxidized electrochemically to regenerate the active oxyhydroxide, completing the cycle.

FIGURE 5.

Cyclic voltammetry recorded at a scan rate of 50 mV s⁻¹ at different temperatures (25, 40, and 60 °C) in 1 M KOH + 1 M glycerol electrolyte. (a) NiCo, (b) AuIn, and (c) PdNi.

In contrast, the increased currents observed for metallic electrodes (AuIn and PdNi) cannot be related to changes in the surface oxidation state, since as can be seen in Figure S11b,c, no new electrochemical processes are observed in the absence of glycerol. This means that this negative shift in the GOR onset potential is likely due to an earlier formation of the adsorbed OH species. Another interesting effect is that the loss of activity linked to surface oxidation is delayed to higher potentials as the temperature increases. This suggests that as the temperature increases, surface oxidation is impeded at higher temperatures. The voltammetry measurements in Figure S11b,c show that the surface oxidation onset potential is not heavily affected by temperature, implying that this process is not kinetically restricted and occurs once a critical OH coverage is reached. Therefore, the delayed activity loss can be attributed to an increased glycerol oxidation reaction rate at higher temperatures, which consumes the adsorbed OH more rapidly, lowers its surface coverage, and shifts oxide formation to higher potentials. During the reverse scan, surface reactivation occurs at more positive potentials, as the temperature also accelerates the reduction of the material.

Scheme 2 summarizes the overall oxidation mechanism. As the potential increases, metallic surfaces start to adsorb hydroxide ions, leading to an increased surface coverage. Once a critical coverage is reached, OH_ad reacts with the surface, oxidizing it. However, if glycerol is present in the solution, it reacts with OH_ad and reduces the overall coverage, shifting the onset of surface oxidation to higher potentials. At higher potentials, the oxide is further oxidized into oxyhydroxide species, which can also react with glycerol and be reduced back to the oxide state. Increasing temperature enhances both oxyhydroxide formation and glycerol oxidation, while surface oxidation remains unaffected. In consequence, the OH_ad coverage decreases, delaying the onset of oxide formation.

SCHEME 2.

Summary of the surface and glycerol oxidation reactions and the effect of glycerol on them.

3.5. Flow Cell Electrolysis

Having determined the effect of temperature on the glycerol oxidation mechanisms, we decided to evaluate its impact on production and selectivity by performing flow cell electrolysis experiments at a current density of 90 mA cm⁻² and a flow rate of 5.7 mL min⁻¹ at three temperatures (25, 40, and 60 °C). Moreover, we evaluated the effect of residence time on the catalyst surface. To do this, additional electrolysis experiments were performed at a higher flow rate of 17.1 mL min⁻¹ to promote the removal of reaction intermediates by reducing their residence time on the electrode surface.

Figure 6 shows the anodic potential during the different electrolysis conditions, where two distinct regions can be identified, corresponding to the two mechanisms. The first region colored in red, ranging from 1.3 to 1.6 V_RHE, corresponds to oxidized electrodes (MO _x ) operating via the oxyhydroxide mechanism. The second region colored in blue, from 0.7 to 1.1 V_RHE, corresponds to metallic electrodes (M⁰) operating via the adsorbed hydroxide mechanism. As expected, NiCo consistently acts as an oxide electrocatalyst, while AuIn acts as a metallic electrocatalyst, indicating that it is able to avoid inhibition under these conditions. PdNi is also expected to operate in the metallic region and initially does so (Figure S12). However, while at 90 mA cm⁻² AuIn can operate at a potential well below that at which deactivation occurs, PdNi operates very close to the tipping point, leading to a gradual buildup of oxide through oxidation of adsorbed OH, which eventually shifts its operation potential to the oxide region. Increasing the temperature from 25 to 40 °C delays this shift, and at 60 °C, the electrode is able to maintain activity in the metallic region for the entire duration of the electrolysis (Figure S13). It is important to note that electrode deactivation can also be prevented by reducing the applied current density, as it would lower the anodic potential, ensuring that all adsorbed hydroxide reacts with glycerol and does not oxidize the surface. The beneficial effect of higher temperatures resides in enabling the material to perform at higher current densities.

FIGURE 6.

Anode electrode potential during the electrolysis at 90 mA cm⁻² for different flow rates (5.7 and 17.1 mL min⁻¹) and temperatures (25, 40, and 60 °C). The red region indicates potentials where the oxyhydroxide mechanism predominates, whereas the blue region indicates potentials where the adsorbed mechanism predominates.

Figure 7 shows the effects of temperature and flow rate on the product distribution for the NiCo electrocatalyst. The catalyst shows high selectivity toward C1 products, primarily formic acid, which accounts for approximately 80% of the Faradaic efficiency. This indicates that the oxyhydroxide mechanism favors the formation of highly oxidized species, due to the strong oxidative power of the oxyhydroxide intermediates. Increasing the flow rate enhances product formation by reducing the residence time of reactants at the electrode surface, interrupting the reaction prematurely. As a result, less oxidized products are favored, but overoxidation toward CO₂ is also impeded, leading to an increase in global Faradaic efficiency. The effect of temperature is more complex; some products increase while others decrease. Formic acid production declines with increasing temperature, which is due to a higher conversion to CO₂, reflected by the reduction in global Faradaic efficiency, indicating that higher temperature promotes complete oxidation. In contrast, glycolic acid production increases. Although its oxidation rate is expected to rise, its formation also increases due to the higher degradation rates of glyceraldehyde and glycolaldehyde, whose production decreases.

FIGURE 7.

Product distribution of NiCo flow cell electrolysis, categorized into C3_DHA, C3_GLAD, C2, and C1.

To evaluate the performance of the electrodes, we focused on four key figures of merit, which are reported in Tables S3–S5: Faradaic efficiency (FE), selectivity, production rate, and energy consumption. A global FE below 100% indicates losses either due to CO₂ overoxidation or competition with the OER. Selectivity and production rate are two parameters that have to be evaluated in conjunction, because high selectivity is only meaningful if it is accompanied by good production values and vice versa. Finally, energy consumption (EC) is arguably the most important parameter as it integrates the production rate with the energy cost of the process to give the product production cost. Based on the data in Table S3, we can conclude that a high flow rate is more desirable because it yields more formic acid and has a lower energy consumption, with only a slight reduction in selectivity. Regarding temperature, all three conditions yield similar selectivity, with a reduction of formic acid production and energy consumption as temperature increases. This means that even though higher temperatures yield less product, the associated decrease in the cell potential compensates for this, enabling the production of formic acid at lower energy cost. Overall, the optimal conditions correspond to a flow rate of 17.1 mL min⁻¹ and a temperature of 60 °C, yielding a formic acid production of 2530 µmol m⁻² s⁻¹ with a selectivity of 83% and an energy consumption of 258 kWh kmol⁻¹.

Figure 8 shows the effect of temperature and flow rate on the product distribution for the AuIn electrocatalyst. This material suffers from low selectivity, with all product categories contributing at least 15%. Interestingly, C1 and C2 products are formed in similar proportions, and global Faradaic efficiency is close to 100%, indicating that the adsorbed hydroxide mechanism is able to cleave the C3 products’ C─C bonds but limits further oxidation. This behavior reflects the lower oxidative strength of the adsorbed hydroxide species compared to oxyhydroxides, as they are generated at lower potentials. Increasing the flow rate leads to higher C3 selectivity and reduced C2 and C1 formation, suggesting that lower surface residence times impede the cleavage process and favor less oxidised products. Similarly, increasing temperature promotes C3 products, with glyceric acid becoming the main product at elevated temperatures, while C2 and C1 products decrease. This trend is partially due to suppression of C─C cleavage, as higher temperatures accelerate the initial glycerol oxidation step, consuming adsorbed hydroxide at a faster rate and limiting its availability for further oxidation steps. However, global Faradaic efficiency decreases with increasing temperature. This is because when increasing the temperature, C2 and C1 oxidation processes stop being impeded, and thus, overoxidation toward carbon dioxide starts being possible. At first glance, this may seem undesirable. However, with the overoxidation to carbon dioxide, there is an increase in selectivity toward glyceric acid, which is beneficial for reducing downstream separation costs. Temperature also affects C3_DHA products: lactic acid production increases while dihydroxyacetone production decreases, consistent with the accelerated degradation of dihydroxyacetone via the temperature‐dependent Cannizzaro rearrangement.

FIGURE 8.

Product distribution of AuIn flow cell electrolysis, categorized into C3_DHA, C3_GLAD, C2 and C1.

The figures of merit in Table S4 indicate that higher flow rates slightly outperform lower ones across all key parameters. Increasing temperature significantly enhances both glyceric acid production and selectivity. This makes a temperature of 60 °C and 17.1 mL min⁻¹ the optimal conditions, reaching a production of 1020 µmol m⁻² s⁻¹ with a selectivity of 40% and an energy consumption of 532 kWh kmol⁻¹.

Figure 9 shows the effect of temperature and flow rate on the product distribution for the PdNi electrocatalyst. Previously, we observed that this material exhibits two distinct anodic potentials, each corresponding to a different oxidation mechanism, which is reflected in the product distribution.

FIGURE 9.

Product distribution of PdNi flow cell electrolysis, categorized into C3_DHA, C3_GLAD, C2, and C1.

At lower temperatures (25 and 40 °C), PdNi exhibits a high selectivity toward C1 products, similar to that of the NiCo electrocatalyst, consistent with the prevalence of the oxyhydroxide mechanism. However, there is also a notable production of C3_GLAD products, indicating that even though the adsorbed hydroxide mechanism is inhibited, the presence of palladium in the material reduces the C─C cleavage of C3 products. At higher temperature (60 °C), the selectivity shifts toward C3_DHA and C3_GLAD products, resembling the behavior of AuIn, indicating a transition to the adsorbed hydroxide mechanism. However, unlike AuIn, PdNi is able to effectively suppress C─C bond cleavage, resulting in a much higher selectivity for C3 products.

As PdNi yields two main products, it is necessary to determine which is more relevant from an industrial viewpoint. Table S5 shows that both products have similar selectivity, but formic acid has higher production. However, glyceric acid has a significantly higher market value, making it economically more attractive even at lower production rates. To maximize glyceric acid production, operation at 60 °C and higher flow rates is required, as these conditions substantially reduce energy consumption without heavily compromising production and selectivity. Under these conditions, we achieve a production of 1840 µmol m⁻² s⁻¹ of glyceric acid, with a selectivity of 71% and an energy consumption of 272 kWh kmol⁻¹.

In summary, operating at higher temperatures and flow rates, we are able to reduce the energy consumption of the process by lowering the cell potential. Table 1 compares the performance of the three electrocatalysts under optimal conditions (60 °C and 17.1 mL min⁻¹). At first glance, NiCo excels in three of the four merit figures closely followed by PdNi, while AuIn shows the worst performance of the three electrocatalysts. However, this does not account for the fact that NiCo produces formic acid, while AuIn and PdNi produce more valuable glyceric acid. Additionally, AuIn is the only material that reaches low cell potentials regardless of the operating conditions, highlighting that every material has its own strengths and weaknesses.

TABLE 1.

Comparison of main product, production, Faradaic efficiency (FE), selectivity, energy consumption (EC), anodic potential (E _ano), and cell potential (E _cell) of the three electrocatalysts at 60 °C and 17.1 mL min⁻¹ operational conditions.

Catalyst	Main product	Production, µmol m−2 s−1	FE, %	Selectivity, %	EC, kWh kmol−1	E ano, V	E cell, V
NiCo	FA	2530	72	83	258	1.32	2.62
AuIn	GLA	1020	44	40	532	1.00	2.19
PdNi	GLA	1840	79	71	272	0.75	2.02

4. Conclusion

This work explored the influence of catalyst composition, flow rate, and operating temperature on glycerol electrooxidation using three representative materials: NiCo oxide, AuIn alloy, and PdNi alloy. Each catalyst exhibits a distinct reaction mechanism: oxyhydroxide‐mediated for NiCo, adsorbed hydroxide‐mediated for AuIn, and a dual mechanism for PdNi. These mechanisms result in different product distributions and electrode potentials.

Electrochemical in situ UV–vis reflectance and Raman measurements revealed the key role of glycerol in modulating surface oxidation. In noble and hybrid catalysts, glycerol consumption delays the formation of surface oxides by depleting the adsorbed hydroxide species. This process preserves the metallic state and maintains the catalytic activity. This effect has been observed for the AuIn and PdNi electrodes, where reflectance measurements demonstrate that glycerol prevents the inhibition of the surface, enabling sustained operation at lower potentials.

Temperature further enhances this protective effect by accelerating the consumption of adsorbed hydroxides and facilitating surface reactivation. In flow cell electrolysis, higher temperatures and flow rates positively influence product selectivity and energy consumption. Elevated temperatures promote reaction kinetics and delay catalyst passivation. Increased flow rates, on the other hand, reduce reactant residence time on the electrode surface, favoring the production of intermediate products.

Overall, the interplay between catalyst composition, temperature, and flow conditions governs both the reaction mechanism and product distribution. These insights provide a foundation for optimizing glycerol electrooxidation in continuous‐flow systems and have implications for energy‐efficient chemical valorization.

Funding

This work was supported by the Ministerio de Ciencia e Innovación (PID2022‐138491OB‐C33, PID2022‐138127NA‐I00, CEX2021‐001202‐M and CEX2024‐001490‐S), European Commission (101077243 and 101115204), and Agència de Gestió d'Ajuts Universitaris i de Recerca (2024 FI‐1 00421).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supplementary Material

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.