Oxidation by Reduction: Efficient and Selective Oxidation of Alcohols by the Electrocatalytic Reduction of Peroxydisulfate

Abstract

Alcohol oxidation is an important class of reaction that is traditionally performed under harsh conditions and most often requires the use of organometallic compounds or transition metal complexes as catalysts. Here, we introduce a new electrochemical synthetic method, referred to as reductive oxidation, in which alcohol oxidation is initiated by the redox-mediated electrocatalytic reduction of peroxydisulfate to generate the highly oxidizing sulfate radical anion. Thus, and counter-intuitively, alcohol oxidation occurs as a result of an electrochemical reduction reaction. This approach provides a selective synthetic route for the oxidation of alcohols carried out under mild conditions to aldehydes, ketones, and carboxylic acids with up to 99% conversion yields. First-principles density functional theory calculations, ab initio molecular dynamics simulations, cyclic voltammetry, and finite difference simulations are presented that support and provide additional insights into the S₂O₈^2–-mediated oxidation of benzyl alcohol to benzaldehyde.

Introduction

The hydroxyl group is the most abundant functional group found in natural products and small molecules with medicinal properties.^1,2 Consequently, alcohol derivatization and functionalization have great technological importance. Functionalizing a hydroxyl moiety usually involves the oxidation of an alpha C–H bond to form a primary carbon radical center, often requiring the use of organometallic catalysts containing Pd,³ Ru,⁴ or Ir⁵ metal centers or transition metal complexes such as pyridinium chlorochromate.⁶ Furthermore, this process is frequently carried out in the presence of auxiliary reagents and the input of light or heat.⁷⁻⁹ These issues make C–H activation one of the most challenging reactions in the field of synthetic chemistry.

Electrochemical methods have been widely used over the past two decades to carry out selective organic transformations under mild reaction conditions.¹⁰ For example, Birch¹¹ and C–C cross coupling reactions^12,13 can be performed at room temperature without the use of toxic or expensive catalyst/auxiliary reagents. Still, electrochemical C–H activation remains one of the main challenges in the field of synthetic electrochemistry.^14,15 The standard redox potential, E⁰, for oxidation of benzyl alcohol to benzaldehyde or benzoic acid are −0.34 and −0.23 V versus Ag/AgCl, respectively.¹⁶ However, the electrooxidation of benzyl alcohol requires the application of potentials greater than 1.40 V versus Ag/AgCl, vide infra. The large overpotential for benzyl alcohol oxidation η (i.e., the difference between the observed peak potential and E⁰) is likely due to the slow kinetics of coupled electron and proton transfer from C–H and O–H moieties.

Aminoxyl redox mediators, such as TEMPO, have been successfully employed to reduce η for electrochemical C–H activation in alcohol oxidation^17,18 and α-cyanation¹⁹ reactions. Unfortunately, these mediators are relatively expensive, unsuitable for large-scale reactions, and show limited modularity. In an attempt to resolve the modularity issue, N-ammonium ylides were developed by Baran and co-workers.²⁰ Nonetheless, alcohol oxidation with these mediators still requires the use of large positive potentials (i.e., 0.5–1.5 V).

Peroxydisulfate (i.e., S₂O₈^2–) has been employed as an oxidant in several synthetic reactions, including Newman-Kwart rearrangement,²¹ cyclization,²² C–H oxygenation,²³ C-3 functionalization,²⁴ alcohol oxidation,²⁵ and arylation reactions.²⁶⁻²⁹ This anion is very appealing for industrial uses because of its low cost and extreme oxidizing potency.³⁰ The homolytic cleavage of the peroxide bond of S₂O₈^2– via heat or light input, also known commonly as peroxydisulfate activation, results in the formation of the sulfate radical anion SO₄^•–, which can efficiently abstract a hydrogen atom from inactive moieties to form corresponding radicals. Indeed, owing to the extreme oxidation capacity of SO₄^•– (i.e., E⁰ = 2.24 V vs Ag/AgCl (3.5 M KCl)),^31,32 peroxydisulfate is widely employed for advanced oxidation of environmental pollutants.³³ In synthetic applications, the activation of peroxydisulfate is usually mediated by an organometallic complex made of precious transition metals such as Pd and requires the input of heat or UV radiation.^34,35

Electrochemical methods offer an attractive alternative method for generating SO₄^•– for organic synthesis. Specifically, S₂O₈^2– undergoes a one-electron reduction at platinum (Pt) or carbon (C) electrodes to form the radical anion, S₂O₈^3•–, which rapidly decomposes to SO₄^2– and SO₄^•–. In the absence of any homogenous chemical reaction, SO₄ immediately undergoes a second reduction at the electrode. The overall sequence is presented in eqs 1–3.

1

2

3

In the presence of suitable substrates, for example, alcohols, the highly oxidizing SO₄^•– generated via eq 2 can be intercepted and used to carry out useful oxidative transformations, eq 4, prior to it undergoing electroreduction at the electrode, eq 3.

4

Using this strategy, C–H activation with electrogenerated SO₄^•– was first reported by Bard and co-workers in the dimerization of diphenylbenzo[k]fluoranthene and 7,14-diphenyloacetonaphthol[1,2-k]-fluoranthene.³⁶ More recently, Maiyalagan and co-workers reported on the peroxydisulfate-mediated electrooxidation of benzyl alcohol and derivatives in biphasic media.³⁷ While similar to the strategy reported herein, this latter study employed constant current electrolysis in an undivided cell using Pt electrodes, conditions where alcohol oxidation occurs directly at the anode without any requisite participation of SO₄^•– generated at the cathode. Therefore, the claims of S₂O₈^2– mediated oxidation of alcohol in that report appear unsupported by experimental results.

In the present report, we showcase the in situ electrogeneration of SO₄^•– for C–H activation in various types of alcohols. The alcohol and S₂O₈^2– are placed in the cathode compartment of a divided cell, and electrolysis is carried out at −0.5 V versus Ag/AgCl. Our results indicate a conversion yield of up to 99% with electrolysis duration less than 2 h. Our approach features efficient C–H oxidation resulting from a redox-mediated electrochemical reduction. Consequently, we refer to this method of alcohol oxidation as reductive oxidation. An earlier example of reductive oxidation of alcohols via electrochemical reduction of aromatic halide was reported by Amatore et al.³⁸ In our experiments, electrochemical reductive oxidation offers selective alcohol oxidation via adjustments in pH of the catholyte (i.e., pH = 6.5 and pH = 12.5 for electrosynthesis of aldehydes and carboxylic acids, respectively). Additionally, the mechanism of electrochemical reductive oxidation of alcohols was investigated using cyclic voltammetry (CV), density functional theory calculations, ab initio molecular dynamics simulations, and finite difference simulations.

Results and Discussion

The oxidation of benzyl alcohol using SO₄^•– (eq 4) generated by the direct reduction of S₂O₈^2–at the working electrode suffers from two major drawbacks: (1) a high overpotential for reduction of S₂O₈^2– and (2) the direct reduction of SO₄^•– at the working electrode. The E⁰ value for the overall reduction of S₂O₈^2– to SO₄^2– is 1.74 V versus Ag/AgCl (i.e., the sum of eqs 1–3). However, CV studies (Figure 1A, black line) reveal that reduction of peroxydisulfate occurs at very negative potentials, displaying a peak current at ∼−1.5 V at glassy carbon (GC) electrodes. Thus, an overpotential of ∼3 V is needed to drive the direct reduction of S₂O₈^2– at transport-controlled rates. Additionally, as noted above, electrogenerated SO₄^•– may be rapidly reduced at the electrode (eq 3) before it can react with an intended substrate, eq 4. To circumvent both of these issues, S₂O₈^2– was homogeneously reduced using the electrogenerated one-electron outer-sphere redox reductant Ru(NH₃)₆^2+ 39 (eqs 5 and 6), resulting in the generation of SO₄^•– (eq 2) close to E⁰ for the Ru(NH₃)₆^3+/2+ system (∼−0.2 V vs Ag/AgCl). The use of Ru(NH₃)₆^3+/2+ mediator results in a nearly 1.3 V reduction in overpotential, Figure 1A.

5

6

Figure 1.

(A) Voltametric responses and E⁰ values for the oxidation of benzyl alcohol, reduction of S₂O₈^2–, and reduction of the Ru(NH₃)₆³⁺ redox mediator. All voltammograms were recorded in O₂-free aqueous solutions containing 0.10 M Na₂SO₄ (pH = 6.50). (B) Reaction mechanism for formation of benzaldehyde and benzoic acid following the reduction of Ru(NH₃)₆³⁺, homogenous reduction of S₂O₈^2–, oxidation of benzyl alcohol with SO₄^•– to form benzyl alcohol radical, and subsequent oxidation of benzyl alcohol radical in acidic (top) and basic (bottom) solutions. For the oxidation of benzyl alcohol radical in acidic solutions, two alternative paths that include HAT and PCET are proposed in the main text.

More importantly, because eq 6 is a homogeneous reaction, the use of a redox mediator results in the generation of SO₄^•– at a significant distance (10–100 μm) from the electrode surface, vide infra. This strategy reduces the loss of SO₄^•– via direct reduction at the electrode (eq 3). The addition of eqs eqs 2–6 yields the overall 2e^– reduction of S₂O₈^2– (eq 7) consistent with Ru(NH₃)₆^3+/2+ acting as an electrocatalyst.

7

Figure 1A shows the voltammetric response of a 1.49 mm radius GC disk in aqueous solution containing 0.5 mM Ru(NH₃)₆³⁺ in the presence and absence of 4.5 mM S₂O₈^2–. In the absence of the S₂O₈^2– (green curve, Figure 1A), the reduction of Ru(NH₃)₆³⁺ is completely reversible, consistent with its known behavior as a fast outer-sphere redox system.⁴⁰ The addition of S₂O₈^2– results in an increase in the cathodic peak height associated with Ru(NH₃)₆³⁺ reduction (blue curve) and the disappearance of the reverse anodic wave. These observations are consistent with the Ru(NH₃)₆²⁺-mediated reduction of S₂O₈^2– to generate S₂O₈^3•–, eq 6, followed by bond cleavage, eq 2, and a second electron transfer, eq 3.

The voltammetric response shown in Figure 1A (blue curve) is similar to that observed for a classic 1e^– electrocatalytic mechanism (E_rC_i′) reaction. However, it is more complex as Ru(NH₃)₆³⁺ can also be regenerated by the reaction of Ru(NH₃)₆²⁺ with the SO₄^•– produced by eq 2, as shown in eq 8.

8

Ru(NH₃)₆³⁺ generated by eq 8 can then be reduced at the electrode, eq 5, resulting in the transfer of the second electron to the electrode. Summation of eqs 5 (2×), 6, 2, and 8 again yields the overall 2e^– reduction of S₂O₈^2–, eq 7. As discussed in more detail later, effective alcohol oxidation involves a careful consideration of the relative rates of SO₄^•– generation (via eq 6 followed by eq 2), relative to the rates of consumption by Ru(NH₃)₆²⁺ (eq 8) or direct reduction at the electrode (eq 3).

As shown in Figure 2, upon addition of benzyl alcohol to an O₂-free H₂O/MeCN (90% H₂O v/v) solution containing Ru(NH₃)₆³⁺ and S₂O₈^2–, the electrocatalytic peak current associated with the mediated reduction of S₂O₈^2– decreases. (Note: the use of the mixed H₂O/MeCN solvent is to improve the solubility of alcohols.) This decrease in current can be readily understood as resulting from the homogenous oxidation of benzyl alcohol by SO₄^•–, thereby preventing a second electron transfer from the electrode to SO₄^•– via either the direct (eq 3) or chemical pathways (eq 8).

Figure 2.

CV study of (black) 0.50 mM Ru(NH₃)₆³⁺, (red) 0.50 mM Ru(NH₃)₆³⁺ and 5 mM S₂O₈^2–, and (blue) 0.50 mM Ru(NH₃)₆³⁺, 5 mM S₂O₈^2–, and 2.50 mM benzyl alcohol. CV studies were carried out with 1.49 mm GC electrode in an O₂-free H₂O/MeCN solution (90% H₂O v/v) containing 0.10 M Na₂SO₄ (pH = 6.50).

In the following section, exhaustive electrolysis and product analysis are used to identify the products generated by oxidation of benzyl alcohol during the mediated electroreduction of S₂O₈^2– via Ru(NH₃)₆³⁺. In the latter sections, computational analyses and more detailed voltammetric studies are presented to support the proposed mechanism of product formation. Table 1 summarizes all reactions involved in the electrocatalytic reduction of S₂O₈^2– and the subsequent oxidation of alcohols.

Table 1. Reactions Involved in the Reductive Oxidation of Benzyl Alcohol.

eq no.	reaction
5
6
2
8
3
13
14
15
BA = benzyl alcohol; BAR = benzyl alcohol radical; BAL = benzaldehyde

Electrosynthesis of Ketones/Aldehydes and Carboxylic Acids via Reductive Oxidation

The voltammetric studies presented above suggest that benzyl alcohol is oxidized upon electrocatalytic reduction of S₂O₈^2–. Controlled-potential electrolysis (CPE) was performed to determine the final product of this reaction. A detailed description of the electrolysis cell and procedure, product isolation, and characterization is presented in the Supporting Information. CPE was performed in a divided cell where the cathode and anode compartments were separated with an ion-exchange membrane supported on a fritted glass membrane. The 20 mL cathode compartment includes a reticulated vitreous carbon (RVC) cathode and an Ag/AgCl (3.5 M KCl) reference electrode. A graphite rod was employed in the anode compartment. A typical example of the current–time (i–t) trace for the mediated reductive oxidation of benzyl alcohol is shown in Figure 3. Before electrolysis begins, the catholyte (16 mL of 0.1 M Na₂SO₄) was deaerated with Ar. Additionally, the catholyte was bubbled with Ar for the duration of the electrolysis. Next, a constant potential of −0.5 V versus Ag/AgCl (3.5 M KCl) was applied to the RVC electrode. Once the current stabilized at the baseline, benzyl alcohol (4.10 μL, 0.04 mmol) dissolved in 2 mL of MeCN was injected to the cathodic compartment of the cell. After 5 min, Na₂S₂O₈ (0.1 g, 0.4 mmol) dissolved in 2 mL of H₂O was injected into the cathode compartment. Next, 1 mL of the catholyte was removed from the cell and used to dissolve Ru(NH₃)₆³⁺ (8 mg, 0.02 mmol). Upon injection of Ru(NH₃)₆³⁺ solution into the cathode compartment, an immediate increase in cathodic current was observed. Electrolysis concludes when all starting Na₂S₂O₈ in the cathode compartment is reduced to SO₄^2– (i.e., ∼1 h). The catholyte was worked up as discussed in the Supporting Information, and the product was purified using column chromatography and characterized by NMR spectroscopy.

Figure 3.

Example of an i-t trace for electrolysis of 2 mM benzyl alcohol in an O₂-free solution of H₂O–MeCN (90% H₂O v/v) containing 20 mM Na₂S₂O₈, 1 mM Ru(NH₃)₆Cl₃, and 0.10 M Na₂SO₄ (pH = 6.50). (a) Application of −0.5 V versus Ag/AgCl to the RVC cathode; (b) addition of benzyl alcohol; (c) addition of Na₂S₂O₈; (d) addition of Ru(NH₃)₆³⁺; (e) conclusion of the electrolysis.

Under optimized conditions, the reductive oxidation of 0.3 mmol benzyl alcohol, 3 mmol Na₂S₂O₈^2–, and 0.05 mmol Ru(NH₃)₆³⁺ at pH 6.5 yielded 0.23 mmole benzaldehyde, corresponding to a 79% yield. The same reaction performed in a pH = 12.5 solution with 3 mmol Na₂S₂O₈ yielded 86% benzoic acid, indicating that the mechanism is strongly pH-dependent. To demonstrate the scope of the reductive oxidation strategy using S₂O₈^2–, as well as the generality of using pH to control the mechanism, a variety of alcohols were electrolyzed by the method described above at pH 6.50 and 12.5. As shown in Figure 4, we were able to synthesize 14 different aldehydes and ketones and 7 different carboxylic acids. Furthermore, conversion yields of up to 99% were achieved in reaction times of less than 2 h. Our results indicate that the mediated reductive oxidation via SO₄^•– is most efficient in forming ketones via oxidizing secondary benzylic alcohols, and it is least effective in oxidizing aliphatic compounds. We hypothesize that this reactivity trend is partly due to the stability of the secondary benzylic alcohol radical relative to the aliphatic alcohols.

Figure 4.

Scope of reductive electrooxidation for selective conversion of alcohols to (A) aldehydes/ketones at pH = 6.5 and (B) carboxylic acids in basic solution with pH = 12. Values shown in the figure refer to the conversion yield calculated using mole of product/mole of the starting compound.

Based on the product analysis, the overall redox reactions for the reductive oxidation of benzyl alcohol via electrogenerated SO₄^•– as a function of pH are given in eqs 9 and10.

9

10

As depicted in Figure 1B, we propose that the mechanism of converting alcohols to aldehydes at pH = 6.5 proceeds by H atom abstraction by SO₄^•– to form the benzyl alcohol radical, followed by a second H atom abstraction by SO₄^•– to yield the corresponding aldehyde. Reductive oxidation at pH = 12.5 to yield carboxylic acids is more complex and most likely involves OH^•. Flash photolysis of S₂O₈^2– in deaerated alkaline solutions (pH > 9) has been shown to yield OH^• via eq 11.^41,42

11

In the following section, we present computational analysis to gain insight into the energetics and mechanism of oxidation of benzyl alcohol to benzaldehyde (eq 9). Density functional theory (DFT) calculations were carried out to analyze the stepwise mechanism of S₂O₈^2– reduction via Ru(NH₃)₆²⁺ and the oxidation of benzyl alcohol by SO₄^•–. Guided by the mechanistic insights from the DFT analysis, finite difference simulations are used to model the voltammetric behavior and to extract rate constants for several intermediate steps.

Computational Analysis

First-principles DFT calculations and ab initio molecular dynamics (AIMD) simulations were used to analyze the Ru(NH₃)₆^3+/2+/S₂O₈^2–-mediated reductive oxidation of benzyl alcohol to benzaldehyde, as proposed in Figure 1B and by the elementary reaction steps in Table 1. Periodic AIMD simulations were carried out using the Vienna Ab initio Simulation Package⁴³ to determine the appropriate solvation shells of the different solution-phase reactants, intermediates, and product complexes. The solvated structures were subsequently optimized using static DFT calculations with Gaussian 16 software package⁴⁴ to determine the lowest energy structures of reactants, intermediates, transition states, and products. Additional computational details are provided in the Supporting Information. These solvated species were subsequently used to obtain the reaction energies (thermodynamics) and activation barriers (kinetics) involved in the proposed elementary steps of the reaction mechanism.

The S₂O₈^2– redox reactions, eqs 1–3, were first examined. The calculated free energies for electron transfer reactions were used to determine the corresponding redox potentials using the computational hydrogen electrode model⁴⁵ (Supporting Information). The reduction potentials for the 1e^– reductions of S₂O₈^2– (eq 1) and SO₄^•– (eq 3) were first calculated using the implicit SMD (solvation model based on density) solvation model.⁴⁶ The calculated reduction potentials (E⁰_implicit, Table 2) for S₂O₈^2– and SO₄^•– show significant deviations from the values (E⁰_expt, Table 2) obtained from experimentally determined free energies of formation of the different reaction species (Supporting Information).³² The lack of explicit water molecules that can directly interact with and solvate-charged intermediates significantly underpredicts the stabilization of these highly anionic species in the solvent phase. A detailed investigation justifying the choice of the DFT functional, the basis set, the reference electrode, and the explicit solvation model employed in this work has been further carried out and presented (Supporting Information). These results help to further affirm that the large errors in the DFT-predicted potentials of these anionic species are significantly influenced by the nature of solvation over the choice of methods.

Table 2. Experimental and Computed E⁰ Values (vs Ag/AgCl (3.5 M KCl)).

AIMD simulations were subsequently carried out for these ions in an explicit water solvent to determine the solvent structures and hydrogen bonding networks for these ions. The lowest-energy molecular structures, which include the S₂O₈^2– and SO₄^•– ions and two hydration shells from the solvent network around these ions, were extracted from the AIMD simulations (shown in Figure 5) and used to calculate the explicit reduction potentials (E⁰_explicit, Table 2). The E⁰_explicit values calculated using explicit solvent molecules are significantly closer to E⁰_expt, Table 2. This improvement is the result of stabilizing these anionic species with the hydrogen bonding network, which is enabled by the introduction of explicit water molecules in the calculations. This stabilization is further evident from the explicit solvent solute structures, which show the proton ends of the solvent coordinated to the anions and the oxygen ends of the solvent coordinated to cations. The structures of the solvation shells obtained with dynamics in this work agree with the previously reported computational and experimental results on these solvation shells.⁴⁷ A detailed comparison is provided in the Supporting Information. Further details on the sampling of the different solvation shells of the solute species is also provided in the Supporting Information, which includes simulations run at longer time scale and simulations with temperature cycling followed by longer time scale run.

Figure 5.

DFT-optimized explicit water networks and structures for (A) S₂O₈^2–/SO₄^•– + SO₄^2– (eq 1 and 2), (B) SO₄^•–/SO₄^2– (eq 3), and (C) Ru(NH₃)₆³⁺/Ru(NH₃)₆²⁺ (eq 5) redox pairs. The water molecules highlighted in green are part of the first hydration shell. The surrounding nonhighlighted water molecules comprise the second hydration shell. The E⁰ values calculated from DFT are shown in red, while values based on the experimental results are in purple. All potentials are reported versus the Ag/AgCl (3.5 M KCl) reference electrode.

The decomposition of S₂O₈^3•– to form SO₄^•– and SO₄^2– (eq 2) was subsequently examined using AIMD simulations. The static DFT calculations reported above showed a significant increase in the peroxo bond length from 1.40 to 2.40 Å upon electron transfer to S₂O₈^2– to generate S₂O₈^3•–. AIMD simulations further revealed that the S₂O₈^3•– species fragments within 5 × 10^–12 s to form SO₄^•– and SO₄^2– (see Supporting Information for the energies), suggesting that this disproportionation reaction is very fast with a free energy barrier of ΔG^‡∼ 0 kJ/mol. These findings indicate that the reduction of S₂O₈^2– likely occurs in a single concerted step to form SO₄^•– and SO₄^2–, rather than in a stepwise mechanism. Hence, the DFT-optimized S₂O₈^3•– structures essentially represent the complexes of SO₄^•– and SO₄^2– species, which result upon electron transfer to S₂O₈^2– and simultaneous cleavage of the peroxo bond. Thus, eqs 2 and 3 can be combined and written as a single step

12

The calculated reduction potentials for S₂O₈^2– at E⁰ = 1.06 V and for SO₄^•– at E⁰ = 1.96 V versus Ag/AgCl indicate that both reactions are thermodynamically downhill at the potential used to oxidize benzyl alcohol (−0.5 V vs Ag/AgCl). Electrostatic work, however, is required to bring these anionic species to the electrode to enable heterogeneous electron transfer, which likely leads to high overpotentials and the kinetically limited electron transfer (Figure 1A). To address the kinetics of electron transfer for the reduction of S₂O₈^2– and SO₄^•–, we used Marcus Theory^48,49 together with DFT calculations to determine the inner- and outer-sphere reorganization energies (Supporting Information). The calculated inner-sphere reorganization energy (λ_i) for the reduction of S₂O₈^2- is very high at λ_i = 416 kJ/mol as there are significant structural changes that result from the elongation and dissociation of the peroxo bond. This results in a free energy barrier of ΔG^‡ > 53 kJ/mol and occurs in the normal regime of Marcus theory. The inner-sphere reorganization energy for the reduction of SO₄^•– is significantly lower at λ_i = 23 kJ/mol as there is little change in the SO₄^•– structure upon electron transfer. The reduction of SO₄^•–, on the other hand, falls within the inverted regime of Marcus theory, thus resulting in a free energy barrier of ΔG^‡ ∼ 0 kJ/mol. The significantly higher inner-sphere reorganization energy, more negative reduction potential, and higher anionic charge should make the direct heterogeneous reduction of S₂O₈^2– (eq 1) much more kinetically limited than the reduction of SO₄^•– (eq 3). This is consistent with the results from voltametric experiments (Figures 1A and 2).

Employing the Ru(NH₃)₆^3+/2+ mediator results in the electrocatalytic reduction of S₂O₈^2– via a homogeneous electron transfer and significantly reduces the overpotential for S₂O₈^2– reduction. The reduction potential for the 1e^– reduction of Ru(NH₃)₆³⁺ (eq 5) was calculated using both the implicit SMD solvation model and explicit water solvation shown in Figure 5c. As shown in Table 2, DFT-predicted reduction potentials from both approaches agree with the experimental values, with an overprediction of 50 mV and an underprediction of 130 mV, respectively. For comparisons with the previous studies,⁵⁰ see Supporting Information.

Heterogeneous and homogeneous electron transfer barriers for the Ru(NH₃)₆^3+/2+ mediated reactions were calculated using the optimized water-solvated S₂O₈^2–/3•–, SO₄^•–/2– and Ru(NH₃)₆^3+/2+ intermediates to simulate eqs 5–8 of the reaction using Marcus Theory^48,49 (see Supporting Information). The E⁰ value for 1e^– reduction of Ru(NH₃)₆³⁺ was calculated to be −0.33 V versus Ag/AgCl suggesting that this reduction (eq 5) should be facile at −0.50 V versus Ag/AgCl at the GC electrode. A more detailed analysis of the reaction kinetics carried out using Marcus Theory gives a free energy barrier of ΔG^‡ = 10 kJ/mol (detailed in Supporting Information) for the reduction of Ru(NH₃)₆³⁺ to Ru(NH₃)₆²⁺. The subsequent reduction of S₂O₈^2– to S₂O₈^3•– by Ru(NH₃)₆²⁺ (eq 6) was calculated to have a reaction free energy of ΔG_reac= −134 kJ/mol and a free energy barrier of ΔG^‡ = 62 kJ/mol. The reaction energy and total reorganization energy of λ = 477 kJ/mol indicate that this homogeneous electron transfer reaction proceeds in the normal regime of Marcus theory (). Similar to the direct reduction of S₂O₈^2– at the electrode, the high inner-sphere reorganization energy and activation barrier for this homogeneous electron transfer is due to the significant increase of the peroxo bond from 1.40 to 2.40 Å. The subsequent reduction of SO₄^•– by Ru(NH₃)₆²⁺ (eq 8) was calculated to have a reaction free energy of ΔG_reac = −221 kJ/mol and a free energy barrier of ΔG^‡ = 47 kJ/mol. The reaction energy and total reorganization energy of λ = 91 kJ/mol indicate that this homogeneous electron transfer reaction falls within the inverted regime of the Marcus theory (). Further details of a parameter sensitivity analysis from the sampled different solvation shells of the solute species is provided in the Supporting Information. This analysis gauges the effect of the electron transfer reaction energy on the heterogeneous electron transfer barriers.

The calculated reaction free energies and activation free energies for the elementary steps involved in the reduction of S₂O₈^2– to form the reactive SO₄^•– intermediate were subsequently used to construct the reaction free-energy profile in Figure 6. The higher activation barrier and the less exergonic reaction free energy for electron transfer associated with the reduction of S₂O₈^2– by Ru(NH₃)₆²⁺ (eq 6), relative to the reduction of Ru(NH₃)₆³⁺ (eq 5) and the reduction of SO₄^•– by Ru(NH₃)₆²⁺ (eq 8) indicate that eq 6 is likely the rate-limiting electron transfer reaction. This is consistent with finite difference simulations of the voltametric results, presented in the next section.

Figure 6.

Reaction coordinate diagram for the Ru(NH₃)₆^3+/2+ mediated reduction of S₂O₈^2– showing eqs 5, 6, and 8 of the mechanism, with eq 6 being the rate-limiting electron transfer step.

The SO₄^•– radical anion that is formed homogeneously can subsequently oxidize the benzyl alcohol to the corresponding benzaldehyde. The two-step oxidation of the alcohol to aldehyde can proceed via (a) two sequential hydrogen atom abstractions steps, (b) two sequential proton-coupled electron transfer (PCET) reactions, or (c) a hydrogen atom abstraction followed by a PCET reaction. While SO₄^•– can mediate hydrogen abstraction, both SO₄^•– and Ru(NH₃)₆³⁺ mediators can act as a sink of electrons for the PCET step. Both implicit and explicit solvation models were employed to capture the different aspects of these reactions. As shown below, while implicit solvation captures the intrinsic reactivity in the hydrogen abstraction mechanism, explicit solvation is imperative to capture the proton transfer steps in the PCET mechanism.

The two-step hydrogen atom abstraction mechanism can proceed via two distinct steps: initial C–H activation followed by O–H activation or initial O–H activation followed by C–H activation. The reaction energies and activation barriers for the SO₄^•–-mediated hydrogen abstraction steps were initially modeled using implicit water solvation, as shown in Figure 7. The calculated activation barriers indicate that the first hydrogen abstraction proceeds via the activation of the benzylic C–H bond (TS_CH (green)), resulting in a low free energy barrier of ΔG^‡ = 0.2 kJ/mol (eq 13). The barrier to initially activate the O–H bond of the benzyl alcohol (TS_OH (red)) was calculated to be significantly higher with ΔG^‡ = 12.9 kJ/mol. The lower barrier for the initial C–H activation is attributed to the formation of a stable benzylic carbon radical [P–CH (green)], which results in an exergonic reaction free energy of ΔG_reac = −72.9 kJ/mol. The phenolic radical that forms from the initial O–H activation (P_OH (red)) is less stable and results in an endergonic reaction free energy of ΔG_reac = 1.8 kJ/mol. Activation barriers calculated with explicit solvation also show the same trend as those with implicit solvation for initial C–H versus initial O–H activation (Supporting Information).

Figure 7.

Free energy-reaction coordinate diagram for the first hydrogen abstraction step from benzyl alcohol depicting initial C–H activation (green) versus initial O–H activation (red).

The subsequent activation of the O–H bond of the benzyl alcohol radical intermediate by a second SO₄^•– has no barrier for activation and readily results in the formation of the benzaldehyde product (eq 14). We could not isolate a transition state for this facile second hydrogen abstraction step at the alcoholic oxygen position, even with explicit solvent molecules (Supporting Information). The second hydrogen abstraction from the benzyl alcohol radical appears to be a barrierless process that is solely limited by the rate of mass transfer.

13

The oxidation of benzyl alcohol to benzaldehyde as two sequential PCET steps was subsequently examined. The oxidation potentials of benzyl alcohol (eq 16) and benzyl alcohol radical (eq 17) were first calculated using implicit and explicit solvation (Supporting Information). The first oxidation step of benzyl alcohol is computed to occur at a potential between 1.65 and 1.83 V versus Ag/AgCl (Supporting Information), in agreement with experimental cyclic voltametric data that show that the peak current for direct benzyl alcohol oxidation is at ∼1.5 V (Figure 1A, red trace). This positive oxidation potential, with no evidence of a proton transfer to the solvation shell in the DFT structures upon oxidation, suggests that the first step of benzyl alcohol oxidation is indeed a hydrogen abstraction step rather than a PCET step (Supporting Information).

16

17

For the second step, the oxidation of benzyl alcohol radical is computed to occur between –0.84 and –0.33 V versus Ag/AgCl (Supporting Information). This more negative range of potentials (i.e., more thermodynamically favorable) coupled with a simultaneous or sequential proton transfer to the adjacent water solvent during oxidation suggests that the second step of benzyl alcohol oxidation could indeed proceed via a PCET step or proceed via the hydrogen abstraction mechanism, as shown in Figure 1B and in the Supporting Information. Using Marcus theory, we further calculated the barrier for this PCET step involving Ru(NH₃)₆³⁺ mediation of benzyl alcohol radical to afford benzaldehyde and Ru(NH₃)₆²⁺ (eq 15). Our analysis indicates that eq 15 can occur with a low free energy barrier of ΔG^‡ = 7 kJ/mol and an overall reaction free energy of ΔG = – 48 kJ/mol. A PCET step with SO₄^•– mediation of benzyl alcohol radical to yield benzaldehyde and SO₄^2–, however, carries a higher free energy barrier of ΔG^‡ = 49 kJ/mol. Hence, the mediation by SO₄^•– more likely proceeds via a hydrogen abstraction step (eq 14) rather than a PCET step (eq 15) (Supporting Information).

Digital Simulation of the Voltammetric Behavior Observed During Reductive Oxidation of Benzyl Alcohol

The above DFT analysis reveals four key findings: (1) the 1e^– reduction of S₂O₈^2– most likely occurs by a concerted mechanism to yield SO₄^•– and SO₄^2–; (2) the rate-determining step during reductive oxidation of benzyl alcohol is the homogenous reduction of S₂O₈^2– by Ru(NH₃)₆²⁺ (eq 6); (3) the first step of the benzyl alcohol oxidation to the benzyl alcohol radical proceeds through eq 13; (4) the oxidation of benzyl alcohol radical to benzaldehyde can proceed either via eq 14 or eq 15. In addition to the computational results, we presented preliminary voltametric data in Figure 2 for the catalytic reduction of S₂O₈^2– by electrogenerated Ru(NH₃)₆²⁺ (eq 6), resulting in the formation of SO₄^•– (eq 2 or eq 12). The SO₄^•– is then reduced to SO₄^2– at the electrode (eq 3) or by reaction with a second electrogenerated Ru(NH₃)₆²⁺ (eq 8). We also showed in Figure 2 that upon the addition of benzyl alcohol to the solution, the peak current associated with the catalytic reduction of S₂O₈^2– decreases, a consequence of the electrogenerated SO₄^•– reacting with benzyl alcohol (eqs 13 and 14 ), preventing it from being reduced at the electrode.

We now present a more detailed voltametric study of the benzyl alcohol oxidation to examine the key findings suggested with DFT calculations. The voltametric data are analyzed using finite difference computational methods to simulate the voltametric behavior and extract kinetic rates for several reaction steps. Finite difference simulations are based on numerical solutions to the time- and potential-dependent differential equations that describe the coupled electrochemical reactions, diffusive transport [Fick’s laws, J_i = −D_i(dC_i/dx) and dC_i/dt = D_i(d²C_i/dx²)], and homogenous kinetics (e.g., dC_i/dt = −kC_iC_j) of the various species i, j,...proposed in the reaction mechanism. Comparison of the experimental and simulated voltammograms is a common and powerful method to validate proposed mechanisms and to extract kinetic rate information. An introduction to finite difference-based simulations in electrochemistry is discussed elsewhere.⁴⁹ Details specific to the simulations and analyses described here are provided in the Supporting Information.

As noted above, the concerted reduction of S₂O₈^2– directly to SO₄^2– and SO₄^•–, eq 12, is equivalent to a two-step mechanism in which S₂O₈^3•– is first produced and then rapidly dissociates to form SO₄^2– and SO₄^•– (eqs 1 and 2). We chose to simulate and discuss the voltametric response in terms of the two-step sequence, although identical results are obtained using the one-step concerted mechanism. As discussed below, our simulation results are consistent with the lifetime of S₂O₈^3•– being very short (<1 μs), in agreement with the DFT predictions (5 × 10^–12 s).

We first examined the electrocatalytic reduction of S₂O₈^2– by Ru(NH₃)₆²⁺ (eq 6). Figure 8 shows the experimental (solid lines) and simulated (dashed lines) voltammograms for solutions containing 0.5 mM Ru(NH₃)₆³⁺ in the presence of 0 (red), 2 (blue), 4 (green), and 8 mM (black) S₂O₈^2–. The results shown in Figure 8 were obtained at the scan rate (ν) = 100 mV/s; voltametric results for the same system, but obtained at ν = 50 and 200 mV/s, are presented in the Supporting Information.

Figure 8.

Experimental (solid lines) and simulated (dashed lines) voltammograms for solutions containing 0.50 mM Ru(NH₃)₆³⁺ (red lines) and S₂O₈^2– at 0 mM (red trace), 2 mM (blue trace), 4 mM (green trace), and 8 mM (black trace). CV studies were conducted in H₂O–MeCN (90% H₂O v/v) solution containing 0.10 M Na₂SO₄ (pH = 6.50). All voltammograms were recorded at ν = 100 mV/s at a 1.49 mm radius GC working electrode.

The simulated response of the voltammogram for the 1e^– reduction of Ru(NH₃)₆³⁺ (eq 5) was performed using the Butler-Volmer model to describe the electron transfer step, with the reported standard rate constant (k⁰) of 17 cm/s and the transfer coefficient (α) of 0.50.⁴⁰ A diffusivity (D) of 3.90 × 10^–6 cm²/s for both Ru(NH₃)₆³⁺ and Ru(NH₃)₆²⁺ was used based on the literature value for Ru(NH₃)₆³⁺ (5.71 × 10^–6 cm²/s)³⁷ and after adjustment for the viscosity of the H₂O–MeCN (90% H₂O v/v) solution (see the Supporting Information). For the moderate ν employed in the experiment (50–200 mV/s), these kinetic parameters yield a reversible voltammetric response for the reduction of Ru(NH₃)₆³⁺ in the absence of S₂O₈^2– with a cathodic/anodic peak splitting of ∼59 mV, equal cathodic and anodic peak heights, and a cathodic peak current increasing in proportion to ν^1/2.

Upon the addition of S₂O₈^2–, the voltametric peak current for the reduction of Ru(NH₃)₆³⁺ increases, while the reverse anodic wave disappears, consistent with Ru(NH₃)₆²⁺-mediated reduction of S₂O₈^2– to generate S₂O₈^3•– (eq 6), followed by rapid S₂O₈^3•– dissociation to produce SO₄^•– (eq 2) and the reduction of SO₄^•– at the electrode surface (eq 3) or via reduction with Ru(NH₃)₆²⁺ (eq 8). Thus, we simulated the voltammetric response in the presence of S₂O₈^2– by building on the simulation of Ru(NH₃)₆³⁺ alone (described above) and then including eqs 2, 3, 6, and 8. The rate constants for each reaction step, k_i, used in the simulations were guided by the DFT results and adjusted to optimize the fit of the simulations to the experimental voltammograms. The diffusivities of S₂O₈^2–, S₂O₈^3•–, and SO₄^•– in the H₂O–MeCN (90% H₂O v/v) solutions were all assumed to be 0.70 × 10^–5 cm²/s based on the range of D values reported for S₂O₈^2– in aqueous solutions (∼0.80–1.20 × 10^–5 cm²/s).⁵¹

The DFT results presented above indicate that the lifetime of the S₂O₈^3•– is on the order of ∼10^–12 s; thus, we used k₂ = 10¹² s^–1, although the simulated voltammograms are insensitive to k₂ for values greater than 10⁶ s^–1 (note: subscripts on rate constants refer to the reaction number). We also assumed that the direct reduction of SO₄^•– at the electrode surface (eq 3) is diffusion controlled, with a heterogenous rate constant, k⁰, of ∼10 cm/s. The remaining two homogenous chemical steps involve the reactions of Ru(NH₃)₆²⁺ with S₂O₈^2– (eq 6) and with SO₄^•– (eq 8). By varying k₆ and k₈, we found that the simulated voltammograms were consistent with k₈ > k₆, in agreement with DFT analyses that indicate that eq 6 is the rate-limiting step. Best fits to the voltammogram were obtained using k₆ = 2.0 × 10⁵ M^–1s^–1 and setting k₈ to any value greater than 10⁷ M^–1s^–1 (see the Supporting Information for fitting opimizations).

Using the above kinetic rates, excellent agreement between simulated and experimental voltammograms was obtained for all S₂O₈^2– concentrations (2–8 mM) and ν = 50–200 mV/s examined, indicating the robustness of the kinetic model (Figure 8 and Supporting Information). We conclude from this analysis that the initial electron transfer step between Ru(NH₃)₆²⁺ and S₂O₈^2– (eq 6) is the rate-determining step in the electrocatalytic reduction of S₂O₈^2–,which is consistent with the DFT results. The value of k₆ obtained from the simulations is insensitive to the assumed values of D = 0.70 × 10^–5 cm²/s for the intermediates S₂O₈^3•–and SO₄^•– but is sensitive to the value of D for the reactant S₂O₈^2–. We estimate that the uncertainty introduces up to a ∼20% error in the value of 2.0 × 10⁵ M^–1s^–1 for k₆.

While the above analyses are based on measurements in the H₂O–MeCN (90% H₂O v/v) mixed solvent, we also repeated the measurements using pure H₂O as the solvent (see the Supporting Information). We obtained qualitatively identical voltammetric responses, with simulations again yielding k₆ ∼ 2.0 × 10⁵ M^–1s^–1. Thus, our conclusions regarding the mechanism for the electrocatalytic reduction of S₂O₈^2– are equally applicable to pure aqueous solutions.

Upon the addition of the benzyl alcohol to a solution containing 0.50 mM Ru(NH₃)₆³⁺ and 4.50 mM S₂O₈^2–, the voltammogram remains irreversible (no anodic peak on the reverse scan), but the height of the cathodic peak decreases, Figure 9A. Shown in Figure 9B is a plot of the normalized peak current, , as a function of the benzyl alcohol concentration, where i_p0 and i_p are the heights of the catalytic peak current in the absence and presence of benzyl alcohol, respectively. With the introduction of benzyl alcohol,decreases rapidly, approaching ∼0.20 when the concentration of benzyl alcohol, C_BA, is greater than 15 mM.

Figure 9.

(A) Experimental (solid lines) and simulated (dashed lines) voltammograms for solutions containing 0.45 mM Ru(NH₃)₆³⁺, 4.5 mM S₂O₈^2–, and 0 (black) 1 (red), 2 (green), 4 (blue), and 30 mM (magenta) benzyl alcohol. Simulations of the voltammetric waves are based on mechanism B using rate constants eqs 2, 3, 5, 6, 8, 12, and eq 14 discussed in the text. (B) Dependence of the normalized voltammetric peak currents, i_p/i_p0, as a function of benzyl alcohol concentration. The black squares show experimental results, while the blue and green lines show simulated values of i_p/i_p0 based on mechanism A (eqs 2, 3, 5, 6, 8, 13, and eq 14) or mechanism B (eqs 2, 3, 5, 6, 8, 13, and eq 15). Error bars (red) shown in B are obtained by averaging i_p values of three successive CVs at each benzyl alcohol concentration (Supporting Information). See text for values of rate constants. Experimental voltammograms are obtained in O₂-free solution of H₂O–MeCN (90% H₂O v/v) containing 0.10 M Na₂SO₄ (pH = 6.5). All voltammograms are recorded at ν = 100 mV/s using a 1.49 mm radius GC working electrode.

The decrease in the height of the cathodic peak with increasing C_BA is due to the oxidation of benzyl alcohol by the electrogenerated SO₄^•–, preventing the reduction of SO₄^•– at the electrode. As considered in the DFT computational analyses, two possible mechanisms are proposed for the oxidation of benzyl alcohol: (A) two successive hydrogen atom transfers (HATs) (eqs 13 and 14 ) or (B) one HAT followed by a PCET (eqs 13 and 15 ). Digital simulation was employed to determine which of these mechanisms provided a better description of the experimental results, details of which are reported in the Supporting Information. Briefly, mechanism A includes eqs 2, 3, 5, 6, 8, 13, and eq 14, while mechanism B comprises eqs 2, 3, 5, 6, 8, 13, and eq 15. For both mechanisms, identical kinetic parameters were employed for eqs 2, 3, 5, 6, and 8. Additionally, DFT calculations indicated that eq 14 is essentially a barrierless event. Hence, k₁₄ was assumed to be diffusion controlled in the finite difference simulations, 1 × 10¹⁰ M^–1 s^–1.

For mechanism A, the rate of reaction for the first HAT (k₁₃) was varied between 1.0 × 10⁶ and 1.0 × 10⁹ M^–1 s^–1, and the best fit to the experimental voltammograms was found for k₁₃ = 8.1 × 10⁸ M^–1 s^–1 while maintaining the rate of the second HAT reaction (k₁₄) at 1.0 ×10¹⁰ M^–1 s^–1 (i.e., diffusion limited). The value of k₁₃ determined by this method is in good agreement with the reported experimental values for the oxidation of the alcohols by SO₄^•– generated by pulse radiolysis (i.e., 2.60 × 10⁶–3.10 × 10⁸ s^–1).⁵² As shown in Figure 9B, the simulations based on mechanism A (with eqs 13 and eq 14) yields good agreement with the experimental values of i_p/i_p0 at low C_BA. However, the simulated values level off at ∼0.60 with increasing C_BA (green line in Figure 9B), well below the experimental limiting value of ∼0.20. No combined adjustment of k₁₃ and k₁₄ yielded agreement of the simulated and experimental values over the entire range of C_BA.

Simulation of mechanism B was carried out using the value of k₁₃ obtained above (8.1 × 10⁸ M^–1 s^–1) while varying the rate of the eq 15 (k₁₅). The best fit of the simulated CVs to the experimental results was achieved for k₁₅ values of 1.0 × 10⁵ M^–1 s^–1, as shown in Figure 9A (black dashed lines). The plot of i_p/i_p0 versus alcohol concentration is also in reasonable agreement with the experimental values, with both values asymptotically approaching a limiting value of ∼0.20. Based on the better agreement between simulation results for mechanism B and the experimental observations, we conclude that the oxidation of benzyl alcohol to benzaldehyde most likely proceeds through reactions 13 and eq 15. However, we are unable to completely reject reaction 14 as a possible simultaneous pathway.

Digital simulations based on mechanism B were also employed to gain insights into the concentration distributions of the benzyl alcohol radical and SO₄^•– during CV, Figure 10. These simulations correspond to the same conditions as in Figure 9 (0.45 mM Ru(NH₃)₆³⁺, 4.5 mM S₂O₈^2–, and E =–0.5 V vs Ag/AgCl) in the presence of 0.5 mM or 30 mM BA. As expected, with increasing C_BA, the concentration of the benzyl alcohol radical in the diffusion layer increases (Figure 10A), while the concentration of SO₄^•– decreases. However, the concentration of SO₄^•– is always very low relative to that of the benzyl alcohol radical (Note the 1000x concentration scale difference in 9A and 9B). At low C_BA (0.5 mM)_, the profiles in Figure 10B show that SO₄^•– reaches a concentration of ∼6 nM at ∼ 25 μm of the working electrode. In this case, SO₄^•– is reduced to SO₄^2– by both reaction with BA and by direction reduction at the electrode. When C_BA is increased, the amount of free SO₄^•– near the electrode decreases by 100-fold to 60 pM, confirming that electrocatalytically generated SO₄^•– is efficiently intercepted by benzyl alcohol within the diffusion layer.

Figure 10.

Simulations of the concentration distributions of the (A) benzyl alcohol radical and (B) SO₄^•– during CV (at E =–0.5 V vs Ag/AgCl and ν = 100 mV/s) in solutions containing 0.45 mM Ru(NH₃)₆³⁺ and 4.5 mM S₂O₈^2–. Distributions are shown for initial benzyl alcohol concentrations of 0.50 and 30 mM. Simulations were performed assuming mechanism B, using the same simulation parameters employed in Figure 9A (see text).

Conclusions

In this work, we have shown that SO₄^•– is electrocatalytically generated through the mediated reduction of S₂O₈^2– by Ru(NH₃)₆²⁺ and that it can be used to carry out the reductive oxidation of benzyl alcohols to yield either benzoic acids or benzaldehydes. The high reaction efficiency observed is partially due the strategy of employing a mediator to homogeneously generate SO₄^•– away from the working electrode, mitigating the competitive direct reduction of SO₄^•– at the electrode. Detailed DFT calculations, CV, and finite difference element simulations have been used to establish the mechanism for reductive oxidation of benzyl alcohol. The experimental and computational results are in excellent agreement. The rate-determining step of benzyl alcohol oxidation was determined to be the 1e^– reduction of S₂O₈^2– by Ru(NH₃)₆²⁺, with a rate constant of 2.0 × 10⁵ M^–1s^–1. We have shown that the oxidation of benzyl alcohol by electrogenerated SO₄^•– proceeds via the initial activation of the C–H bond (eq 13) followed most likely by a PCET step between the benzyl alcohol radical and Ru(NH₃)₆³⁺ (eq 15).

Reductive oxidation chemistry has been employed in electrolysis experiments to synthesize both carboxylic acids and aldehydes (or ketones) from various benzyl alcohol derivatives. We believe the use of reductive oxidation for C–H activation at negative potentials provides further opportunities to carry out more complex reactions at very mild conditions such as α-carbon functionalization and cross-coupling reactions.

Finally, a byproduct of this study was to establish that the initial 1e^– reduction of S₂O₈^2– to form SO₄^•– and SO₄^2– (eq 12) occurs by a concerted electron transfer disproportionation mechanism rather than in a stepwise manner (eqs 1 and 2). The lifetime of S₂O₈^3•– is estimated to be ∼5 × 10^–12 s.

The data generated in this study has been made publicly available on the ioChem server (https://doi.org/10.19061/iochem-bd-6-144) and includes density functional theory computational input and output data. Additionally, all data for product characterization and parameters for model and computational efforts can be found in the Supporting Information.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.2c07305.

• Details of CV studies and parameters used for digital simulation of cyclic voltammograms, procedure used for CPE, product isolation procedure, ¹H NMR spectra, and details of DFT calculations (PDF)

The authors declare no competing financial interest.