The electrocarboxylation of a series of benzophenones under galvanostatic conditions has been carried out in aprotic solvents using an undivided bulk electrolysis cell equipped with a Mg sacrificial anode. Systematic studies have been carried out in order to establish the qualitative relationships between the yield of carboxylation reaction and the operational and intrinsic parameters for the electrocarboxylation of benzophenones. For the diaryl ketones chosen for these studies, the yields of the target benzilic acids have been found to be strongly dependent on different parameters such as solvents, supporting electrolytes, cathode materials, current density, temperature and the nature of the substrates.

Introduction

Carbon dioxide is recognized to be a naturally abundant, cheap, recyclable, nontoxic building block in organic synthesis. However, the conversion and utilization of carbon dioxide have been challenging for chemists from the ecological and economic points of view due to its high chemical stability. In recent years electrosynthesis has been recognized as a convenient, low cost and eco-friendly method for CO2 fixation [1]. Particularly relevant to our studies, many efforts have been devoted to the studies of the electrocarboxylation of several substrates, including alkenes [2-4], alkynes [5-7], alcohols [8,9], halides [10-12], imines [13,14], aldehydes [15,16] and ketones [16-24] under mild conditions.

Benzilic acids are an important class of compounds, which are widely used as the important intermediates in the production of some pharmaceutical materials, such as anodyne, cosmetic drugs [25-27], spasmolytics and anti-inflammatory drugs [28]. Traditionally, aromatic acids were prepared via the corresponding cyanohydrins using ketones as precursors [29]. However, the need of using highly toxic cyanides makes this method less useful for pharmaceutical applications. Alternatively, benzilic acids can be prepared in a much more environmental friendly way via the electrocarboxylation of the corresponding diaryl ketones with sacrificial anodes [16]. In recent years, different studies have been carried out to investigate the reaction mechanism [17-19,22] and the effects of both the operative parameters and the nature of the substrate [19,30,31] on the electrocarboxylation of aromatic ketones, especially alkyl aryl ketones. Scialdone et al. have shown that the yields of the electrocarboxylation of ketones strongly depend on the nature of substrate [19,21-23]. Low yields of target hydroxyl acids together with the corresponding alcohols and pinacols were obtained in the electrocarboxylation of alkyl aryl ketones due to the short lifetime of their one-electron reduced forms. Apart from these main products, detectable amounts of substituted benzoic acids and cycloexene carboxylic acids were also obtained in the electrocarboxylation of some alkyl aryl ketones. The product distribution of the electrocarboxylation of ketones are also influenced by various other operative parameters, such as the working potential, temperature, the water content and the ratio between the carbon dioxide and the ketone concentrations. Since the electrocarboxylation of ketones involves several competitive reactions which lead to the formation of the acidic products and the corresponding alcohols side products, usually, the products of these electrocarboxylation reactions are mixtures of the target product and the side products. Therefore, not all ketones are applicable in the practical production of aromatic acids by electrocarboxylation. In the case of benzophenones, high yields of benzilic acids were usually obtained since CO2 can readily react with the highly stable one-electron reduced form of benzophenones [18-21].

From an applicative point of view, a highly pure benzilic acid is desirable in view of its destination as a pharmaceutical drug. In order to make the production of benzilic acids by electrocarboxylation practical and commercial useful, it is desirable to develop a cost effective and environmental friendly system with high yield. Since the yield of electrocarboxylation is influenced dramatically by the operative parameters, we carried out a series of systematic studies on the effects of these factors on the yield of benzilic acids in an undivided cell equipped with a Mg sacrificial anode in aprotic solvents. The other major focus of our studies is the investigation of the effect of an electron-donating group to the keto carbonyl group on the yield of electrocarboxylation reaction. The presence of an electron-donating group to the keto carbonyl group favors the formation of the side products, such as alcohol and pinacols, and thus is not ideal for electrocarboxylation applications [23]. However, to what extent the presence of this group may affect the yield of the target acid has not been reported previously.

Results and discussion

Galvanostatic electrolysis in the electrocarboxylation of benzophenone

Benzophenone (1a) was first chosen as a model molecule for the investigation of the electrocarboxylation of benzophenones (Scheme 1 ). Experiments have been conducted to investigate the effects of various experimental parameters, such as electrode materials, solvents, supporting electrolytes, current density, and temperature on the yield of target product.

Scheme 1

Reaction scheme for electrocarboxylation of benzophenones.

Influence of the nature of the cathode

Cyclic voltammetric studies of benzophenone were first carried out in DMF containing 0.1 M Et4NBF4 using a commercial 3 mm diameter glassy carbon (GC) electrode as the working electrode. The results are shown in Fig. 1 . Two reduction processes were observed. The reversible potential (E, taken as the average of oxidation and reduction potentials) of the first process is −1.465 V vs. Ag/AgI/I−. This process corresponds to the one-electron reduction of benzophenone. The fact that the sizes of oxidation and reduction peaks are comparable suggests that this process is chemically reversible in the voltammetric timescale. The separation between the oxidation and reduction peaks is 84 mV when the scan rate is 0.1 V s−1. This peak separation value is larger than the theoretical value of 57 mV for a reversible one electron transfer process at 25 °C [32] due to the effect of uncompensated resistance (the measured uncompensated resistance value is 272 Ω under our experimental conditions. Under this situation, the result obtained from numerical simulation suggests that a peak separation of 85 mV would be expected for a one-electron reversible process). Therefore, the first reduction process is reversible in the timescale of the voltammetric measurements. The second reduction process is chemically irreversible with a reduction peak potential of −1.955 V vs. Ag/AgI/I−. These findings are in good agreement with literature results [33]. The voltammetric processes are summarized in the following equations. where HA stands for solvent or water.

Fig. 1

Cyclic voltammograms at a GC electrode (3 mm diameter, scan rate = 0.1 V s−1 and temperature = 25 °C) in DMF (0.1 M Et4NBF4) for saturated CO2 (a) and for 2 mM benzophenone (1a) in the absence (b and c) and presence (d) of saturated CO2.

When the solution was saturated with CO2 at 1 atm, the first reduction process becomes chemically irreversible and the size of peak current increases significantly, while the peak potential shifted to a more positive value and the second reduction process disappeared completely. This voltammetric response is due to a chemical reaction between the electrogenerated intermediate and CO2 followed by a rapid one electron reduction process to form benzilic acid. The controlled experiment (curve a in Fig. 1) suggests that direct electroreduction of CO2 at a GC electrode does not occur within the same potential region. In fact, our experimental results suggest that direct reduction of CO2 occurs at potential much more negative potential than benzophenone reduction process even at metallic Cu, one of the most effective electrode materials for CO2 reduction [34]. Therefore, the total reaction in all cases can be outlined in Scheme 1.

Among all electrochemical reaction conditions, the nature of electrode materials has the greatest influence on the yield or reactivity in the electrocarboxylation process. Therefore, it is important to find a suitable cathode for the electrocarboxylation of 1a. In order to examine the influence of cathodes on the electrocarboxylation of 1a, the voltammetric responses of 1a in the CO2 saturated solution at 1 atm were also investigated using other commonly used cathodes, e.g., Ti, Pt, Ni and Cu. The results are summarized in Fig. 2 . In all cases, the voltammetric characteristics of 1a were similar to those obtained at the GC electrode (curve e), except that the reduction peak potentials are different and are in the following order: Ti (−1.680 V, curve a) < GC (−1.500 V, curve e) < Ni (−1.400 V, curve b) < Pt (−1.393 V, curve c) < Cu (−1.392 V, curve d).

Fig. 2

Cyclic voltammograms of 10 mM benzophenone (1a) (scan rate = 0.1 V s−1 and temperature = 25 °C) in DMF (0.1 M Et4NBF4) saturated with CO2 at different electrodes: (a) Ti (b) Ni (c) Pt (d) Cu (e) GC.

In the previous studies [18,21,30], the electrolysis of aromatic ketones was investigated on Hg, lead and carbon cathode due to the large hydrogen evolution overpotential at these electrodes. From the applicative point of view, a cost effective and environmental friendly system is desirable in order to obtain a sustainable electrochemical process. Therefore, bulk electrolysis experiments in our studies were conducted at a wider range of working electrodes, including Ti, Ni, Pt, Ag, Cu, stainless steel and graphite (C). The yields, defined as the ratio between the amount of reaction product and that predicted based on Faraday’s law, are summarized in Table 1 . The results suggest that stainless steel is the most efficient cathode material, which resulted in a good electrocarboxylation yield of 92% (entry 6). The other cathode materials gave relatively moderate yields of 69–87%. The results also suggest that the carboxylic yields depend on the cathode materials in the following order: C (69%, entry 4) < Ti (77%, entry 1) < Ni (83%, entry 2) < Pt (85%, entry 3) < Cu (87%, entry 5) < stainless steel (92%, entry 6), which is apparently correlated to the reduction potentials of carboxylation reaction. Higher yield was obtained at the cathode when the reduction potential for carboxylation reaction is more positive. This may be due to the fact that the unwanted background Faradaic processes is less significant when the applied potential for the electrolysis is further away from the limit of the negative potential window. In all cases, GC–MS results show benzilic acid is the major product and the side products are negligible.

Table 1

Influence of cathode material on the electrocarboxylation of benzophenone.a

Entry	Cathode	Yield (%)b
1	Ti	77
2	Ni	83
3	Pt	85
4	C	69
5	Cu	87
6	Stainless steel	92

Experimental conditions: 0 °C, benzophenone 0.1 M, n-Bu4NI 0.1 M, DMF 10 mL, current density 7.65 mA cm−2, electricity 2 F mol−1and CO2 1 atm.

GC yield. The yield based on starting benzophenone.

Influence of the solvent

Electrochemical reactions are often influenced by the solvents [35]. To investigate the effect of the solvents, the electrocarboxylation of 1a was carried out in both DMF and MeCN. The results are summarized in Table 2 . The yield of 2a in DMF was considerably higher than that in MeCN, even though the solubility of CO2 in MeCN is higher [36]. This is in agreement with previous results [19,20].

Table 2

Influence of solvents on the electrocarboxylation of benzophenone.a

Entry	Solvent	Yield (%)b
1	MeCN	72
2	DMF	92

Experimental conditions: 0 °C, stainless steel cathode, benzophenone 0.1 M, n-Bu4NI 0.1 M, current density 7.65 mA cm−2, electricity 2 F mol−1 and CO2 1 atm.

GC yield. The yield based on starting benzophenone.

Voltammograms of 1a were also obtained in MeCN in the presence and absence of CO2 saturated (1 atm). The results are shown in Fig. 3 . Voltammogram of 1a obtained in DMF was shown for comparison. The controlled experiment (curve a in Fig. 3) suggests that direct electroreduction of CO2 does not occur within the same potential region. In the absence of CO2, the results show that the reversible potential of the first reduction potential only shifts slightly more positive when MeCN was used instead of DMF, while the peak potential of the second reduction process shifts significantly to a much more positive value since MeCN is a stronger proton donor compared to DMF [37] which favors the formation of the corresponding alcohol according to Eq. (3). The results in Fig. 3 also show that in the presence of CO2, an irreversible process is observed and the negative potential window of MeCN becomes significantly narrower compared to that of DMF (Fig. 1) and the limit of the negative potential window is very close to the peak potential of the electrocarboxylation process. Therefore, the unwanted background Faradaic current may be higher and the yield of electrocarboxylation may be lower in the case of MeCN. Apart from this, DMF is a solvent with strong basicity which may also favor the formation of acidic product, benzilic acid. We also noticed that the salt of Mg2+ formed from anodic oxidation has much lower solubility in MeCN than in DMF. Therefore, the lower reaction yield obtained in MeCN could also be partially due to the fact that the anodic reaction is inefficient in MeCN because of the precipitation of nonconductive Mg2+ salt onto the surface of Mg anode.

Fig. 3

Cyclic voltammograms at a GC electrode (3 mm diameter, scan rate = 0.1 V s−1 and temperature = 25 °C) in MeCN (0.1 M Et4NBF4) for saturated CO2 (a), for 2 mM benzophenone (1a) in the absence (b and c) and presence (d) of saturated CO2 and for 2 mM benzophenone (1a) in the absence of CO2 in DMF (0.1 M Et4NBF4).

Influence of the supporting electrolytes

The influence of the supporting electrolyte has also been investigated. As shown from the results summarized in Table 3 , the nature of the supporting electrolytes had significant influence on the yield of 2a. The results suggest that both the anion and the cation of electrolyte have pronounced effect on the reaction yields. For the supporting electrolytes with common cation R4N+, the yield of 2a decreased depending on the anions in the following order: I−  > Br−  > Cl−. The origin of this effect is unclear at this stage. For supporting electrolytes with the same anion, the performance of n-Bu4N+ was superior to that of Et4N+, which is possibly due to the fact that n-Bu4N+ is more stable against electroreduction [38].

Table 3

Influence of supporting electrolytes on the electrocarboxylation of benzophenone.a

Entry	Supporting electrolyte	Yield (%)b
1	Et4NCl	45
2	Et4NBr	57
3	Et4NI	59
4	n-Bu4NBr	73
5	n-Bu4NI	92

Experimental conditions: 0 °C, stainless steel cathode, benzophenone 0.1 M, supporting electrolyte 0.1 M, DMF 10 mL, current density 7.65 mA cm−2, electricity 2 F mol−1 and CO2 1 atm.

GC yield. The yield based on starting benzophenone.

Influence of the current density

Current density also has effect on the reaction yields. When the current density is too high or too low, significant contribution from undesired Faradaic process results in the low yield of product. Therefore, the electrocarboxylation of 1a was examined under various current densities in order to find the optimal current density. The results summarized in Table 4 suggest that the optimal current density is 7.65 mA cm−2.

Table 4

Influence of current density on the electrocarboxylation of benzophenone.a

Entry	Current density (mA cm−2)	Yield (%)b
1	3.75	80
2	5.05	87
3	6.35	90
4	7.65	92
5	8.95	81
6	10.25	69

Experimental conditions: 0 °C, stainless steel cathode, benzophenone 0.1 M, n-Bu4NI 0.1 M, DMF 10 mL, electricity 2 F mol−1 and CO2 1 atm.

GC yield. The yield based on starting benzophenone.

Influence of the temperature

The influence of the temperature on the electrocarboxylation reaction is complicated. On one hand, the increase of temperature will lead to the decrease of the solubility of reactant CO2 in DMF; but on the other hand, the increase of temperature will affect both the thermodynamics and the kinetics of the electrocarboxylation. To investigate the influence of the temperature on the electrocarboxylation of 1a, electrolysis was conducted at different temperatures. The reaction yields obtained are summarized in Table 5 . When the temperature was increased from −20 °C to 20 °C, the electrocarboxylation yield increased from 75% to 95% (entries 1–4), but the yield decreased to 81% when the temperature was further increased to 30 °C (entry 5). Therefore, 20 °C is the optimal temperature.

Table 5

Influence of temperature on the electrocarboxylation of benzophenone.a

Entry	Temperature (°C)	Yield (%)b
1	−20	75
2	−5	85
3	0	92
4	20	95
5	30	81

Experimental conditions: the electrochemical cell was water-jacketed, stainless steel cathode, benzophenone 0.1 M, n-Bu4NI 0.1 M, DMF 10 mL, current density 7.65 mA cm−2, electricity 2 F mol−1and CO2 1 atm.

GC yield. The yield based on starting benzophenone.

Influence of the nature of substrates

To study the substituent effect on the yield of electrocarboxylation, we have also investigated the electrocarboxylation of various p-substituted benzophenones (1b–1e) which are the derivatives of 1a with electron donating groups.

The voltammetric characteristics of 1b–1e at a GC electrode in the absence of CO2 are similar to those of 1a, except that the reversible potentials (as shown in Table 6 ) are different as expected. The value of E shifted to more negative potential value in the following order: −1.665 V (1e, entry 5) < −1.591 V (1d, entry 4) < −1.560 V (1c, entry 3) < −1.508 V (1b, entry 2) < −1.459 V (1a, entry 1).

Table 6

Influence of the nature of substrates on the electrocarboxylation of benzophenones.a

Entry	1	Eo (V)b	Ip1 (mA)b	Ep2 (V)c	Ip2 (mA)c	Ip2/Ip1	Yield (%)d
1	1a	−1.459	0.1783	−1.449	0.3186	1.79	95
2	1b	−1.508	0.1876	−1.481	0.3220	1.72	89
3	1c	−1.560	0.1873	−1.518	0.2991	1.60	75
4	1d	−1.591	0.1934	−1.572	0.2879	1.49	67
5	1e	−1.665	0.1704	−1.661	0.2333	1.37	54

Electroanalytical experimental conditions: 20 °C, working electrode GC, benzophenones 0.01 M, Et4NBF4 0.1 M, 10 mL DMF, CO2 1 atm. Reaction conditions as Table 5 entry 4.

Reversible potentials (E V vs. Ag/AgI/I−) and peak currents of the first reduction peak of benzophenones are measured in the absence of CO2.

Peak potentials ( V vs. Ag/AgI/I−) and peak currents of the reduction of benzophenones are measured in the presence of CO2.

GC yield. The yield based on starting benzophenones.

The Hammett constant (σ) is a well-established parameter that takes into account the effect of the electron donating/withdrawing properties of the substituents on the reversible potentials of the complexes. In the case of benzophenone reduction, the dependence of reversible potential on σ can be described by a linear Eq. (4) of the form [39].where X stands for the substituent on the phenyl ring. p is para position of the phenyl ring of benzophenone. The reversible potentials of the first reduction processes of benzophenones are plotted vs. σ. The results are shown in Fig. 4 . Indeed, a good linear correlation of E with σ is obtained.

Fig. 4

Dependence of E of substituted benzophenones 1a–1e on Hammett substituent constants.

When the solutions were saturated with CO2 at 1 atm, the voltammetric characteristics of 1a–1e were similar. In all cases, a chemically irreversible process was observed with a considerable increase in the reduction peak currents compared to the cases where CO2 was absent. However, the peak current ratio decreases in the sequence of 1a  >  1b  >  1c  >  1d  >  1e which suggest that the introduction of electron withdrawing groups to the phenyl ring of 1a lowers the activity of the ketyl radical anion towards CO2 activation. This negative impact of electron withdrawing groups towards CO2 activation also showed up in the longer timescale of electrolysis. The reaction yields obtained decreased in the same sequence. This trend of decrease in reaction yield could also be partially attributed to the fact the contribution from non-desired Faradaic process increases when the value is closer to the limit of negative potential window of the media. GC–MS results suggest that the introduction of electron withdrawing groups to the phenyl ring of 1a does not affect the high selectivity towards benzilic acid. In all cases, the formation of side products is negligible.

Conclusion

We have systematically investigated the effect of operative parameters and intrinsic properties on the electrocarboxylation reactions involving benzophenones. The results obtained allow optimal performance of the electrocarboxylation of benzophenone to be achieved in terms of its yield towards the target carboxylate.

Experimental section

Chemicals and instrumentation

1H NMR spectra were recorded on an AVANCE 300 (300 MHz) spectrometer in CDCl3 with Me4Si as an internal standard. Mass spectra were obtained on a 5973 N spectrometer connected with a HP 6890 gas chromatograph. The N,N-dimethylformamide (DMF) and acetonitrile (MeCN) of an analytical pure were dried over 4 A molecular sieves to remove residual water prior to use. Other reagents (purities >98%) were purchased from Sigma–Aldrich and were used as received.

Electroanalytical procedure

The electroanalytical experiments were performed in DMF solution containing 0.1 M Et4BF4 as the supporting electrolyte in a one-compartment electrochemical cell, equipped with a gas inlet and outlet, a glassy carbon (GC, d  = 3 mm) as the working electrode, a platinum spiral (Pt) as the counter electrode and an Ag/AgI/0.1 M n-Bu4NI in DMF as the reference electrode. The diameters of other working electrodes (Ti, Ni, Cu and Pt) are 2 mm. Voltammetric measurements were performed on a CHI 650C electrochemical station (Shanghai Chenhua Instruments Company).

Typical electrolysis procedure

In a typical experiment, the galvanostatic electrolysis was carried out in DMF solution containing 0.1 M n-Bu4NI as the supporting electrolyte in an undivided tank glass cell with cylindrical geometry, equipped with a gas inlet and outlet, a sacrificial magnesium rod anode and a ring reticulate stainless steel cathode at 0 °C, until 2 F mol−1 of charge was passed. The volume of the electrolyte solution was 10 mL. The electrolytic solution was continuously stirred during electrolysis process.

The reaction mixture was esterified in DMF by adding anhydrous K2CO3 (1 mmol) and MeI (3 mmol) and stirring the mixture at 50 °C for 5 h. The solution was hydrolyzed and extracted with Et2O, dried over anhydrous MgSO4 and evaporated. The methyl esters corresponding to acids were isolated by column chromatography for qualitative analysis. The quantification of the products was done on a GC instrument (GC-2014, Shimadzu).

α-Hydroxy-α-phenylbenzeneacetic acid methyl ester (2a) GC–MS (m/e, %): 242 (M+, 3), 183 (100), 105 (63), 77 (30), 165 (5), 51 (5). 1H NMR (300 MHz, CDCl3) δ: 7.42–7.33 (10H, m), 3.86 (3H, s).

α-Hydroxy-4-methyl-α-phenylbenzeneacetic acid methyl ester (2b) GC–MS (m/e, %): 256 (M+, 1), 197 (100), 105 (66), 119 (35), 77 (27), 91 (21). 1H NMR (300 MHz, CDCl3) δ: 7.42–7.27 (7H, m), 7.15 (2H, d, J  = 9.1 Hz), 3.85 (3H, s), 2.35 (3H, s).

α-Hydroxy-4-methoxy-α-phenylbenzeneacetic acid methyl ester (2c) GC–MS (m/e, %): 272 (M+, 2), 213 (100), 105 (40), 135 (29), 77 (18), 197 (8). 1H NMR (300 MHz, CDCl3) δ: 7.41–7.31 (7H, m), 6.87 (2H, d, J  = 8.4 Hz), 3.85 (3H, s), 3.81 (3H, s).

α-Hydroxy-4-methyl-α-(4-methylphenyl)benzeneacetic acid methyl ester (2d) GC–MS (m/e, %): 270 (M+, 1), 211 (100), 119 (76), 91 (26), 65 (7), 195 (6). 1H NMR (300 MHz, CDCl3) δ: 7.29 (4H, d, J  = 8.1 Hz), 7.14 (4H, d, J  = 8.1 Hz), 3.84 (3H, s), 2.35 (6H, s).

α-Hydroxy-4-methoxy-α-(4-methoxyphenyl)benzeneacetic acid methyl ester (2e) GC–MS (m/e, %): 302 (M+, 1), 243 (100) 135 (75), 227 (49), 32 (41), 77 (19). 1H NMR (300 MHz, CDCl3) δ: 7.32 (4H, d, J  = 8.7 Hz), 6.86 (4H, d, J  = 8.7 Hz), 3.84 (3H, s), 3.81 (6H, s).