Polysubstituted ferrocenes as tunable redox mediators

A series of four ferrocenyl ester compounds, 1-methoxycarbonyl- (1), 1,1’-bis(methoxycarbonyl)- (2), 1,1’,3-tris(methoxycarbonyl)- (3) and 1,1’,3,3’-tetrakis(methoxycarbonyl)ferrocene (4), has been studied with respect to their potential use as redox mediators. The impact of the number and position of ester groups present in 1–4 on the electrochemical potential E1/2 is correlated with the sum of Hammett constants. The 1/1+–4/4+ redox couples are chemically stable under the conditions of electrolysis as demonstrated by IR and UV–vis spectroelectrochemical methods. The energies of the C=O stretching vibrations of the ester moieties and the energies of the UV–vis absorptions of 1–4 and 1+–4+ correlate with the number of ester groups. Paramagnetic 1H NMR redox titration experiments give access to the chemical shifts of 1+–4+ and underline the fast electron self-exchange of the ferrocene/ferrocenium redox couples, required for rapid redox mediation in organic electrosynthesis.


Introduction
Since its discovery, ferrocene (FcH) has been established as versatile redox-active building block [1][2][3]. Ferrocene can be reversibly oxidized to the 17 valence electron ferrocenium cation (FcH + ) at a useful electrochemical potential (FcH/FcH + +630 mV vs NHE; +380 mV vs SCE in CH 3 CN) [4]. The 0/+ redox couple of ferrocene and its derivatives possesses high electron self-exchange rates k ex = 10 6 -10 7 M −1 s −1 , remarkably independent on the electrolyte and solvent [5,6]. Both, the ferrocene/ferrocenium and the decamethylferrocene/ decamethylferrocenium redox couples are well established as internal reference redox systems for electrochemical analyses in non-aqueous media [7][8][9][10]. Important requirements for redox couples with respect to useful applications are: (i) Both components of the redox couple should be soluble. (ii) Homogeneous and heterogeneous electron-transfer (ET) reactions should be fast. (iii) Both components should be stable under the electrolysis conditions and should not react irreversibly with any component of the supporting electrolyte [8]. In general, the redox mediators used as redox catalysts in indirect organic electrosyntheses should comprise the same characteristics [11][12][13][14]. A mediator is a reversible redox couple with a fast ET between itself and the electrode (heterogeneous) and between itself and the substrate (homogeneous). The benefit of the presence of a mediator is the switch of the sluggish heterogeneous electron Scheme 1: Selected transformations with ferrocene/ferrocenium as SET reagents (a) [27], catalyzed (b,c) [29][30][31] and mediated transformations (d-f) [34][35][36] by the ferrocene/ferrocenium redox couple.
transfer between electrode and substrate to a rapid homogeneous redox reaction between mediator and substrate. Further, the mediator's redox potential must be below or above of that of the substrate for oxidation or reduction processes, respectively. This avoids the often kinetically hindered direct ET between electrode and substrate and diminishes overoxidation or overreduction of the substrate.
Ferrocenyl esters 1-4 are synthetically accessible via the acids of 1 [45,46], 2 [57], 3 and 4 [54] in a direct selective metalation of ferrocene [54,[57][58][59][60], quenching with carbon dioxide, followed by esterification [45][46][47][48]54]. The 1,1'-disubstituted ferrocene 2 can also be obtained by direct coordination of the respective substituted cyclopentadienyl ligand (CpR) to iron(II) [49]. An alternative route to the mono-, 1,1'-di-and 1,1',3-tricarboxylic acids of ferrocene is the oxidation of the respective acetylferrocenes [47,48,53]. Ferrocene carboxylic acid is also available via basic hydrolysis of ferrocenyl aryl ketones [61]. Together with the redox potentials of ferrocene, 1 and 2, the hitherto unknown electrochemical potentials of 3 and 4 should cover a wide potential range. This will meet the requirements of different substrates for the potential application of 1-4 and their ferrocenium ions as selective redox mediators or SET reagents. Apart from the redox potentials of the redox mediators FcH and 1-4, the stability of the 18 and 17 valence electron species as well as their solubility and the availability of spectroscopic probes to monitor reaction progress and stability are important issues. These fundamental aspects will be addressed in this study.

Results and Discussion
Electrochemistry of esters 1-4 The esters 1-4 were studied by cyclic and square wave voltammetry in 0.1 M CH 2 Cl 2 solutions of [n-Bu 4 N][B(C 6 F 5 ) 4 ], using platinum working and counter electrodes. All esters 1-4 show an essentially reversible behaviour for the ferrocene/ferrocenium oxidation process ( Figure 1, Figure S1, Supporting Information File 1). The electrochemical potentials cover a wide range, E 1/2 = 260-900 mV vs FcH/FcH + ( Figure 1, Table 1).   [38]. The latter three are only irreversibly oxidized at room temperature precluding any application as mediators. The data are in full accordance with the increasing electron-withdrawing character of the cyclopentadienyl ligands from 1 to 4. The position of the ester groups has a slight influence on the electrochemical potential. 1-or 1'-substitution with a methoxycarbonyl group raises the potential by ca. 250 mV (FcH → 1, 1 → 2), while substitution in 3-and 3'-position has only a smaller impact with an increase of the potential by ca. 200 mV (2 → 3, 3 → 4). According to Lever et al. [39], the calculated electrochemical pa-rameters E L (L) for 1-(methoxycarbonyl)cyclopentadienyl and 1,3-bis(methoxycarbonyl)cyclopentadienyl ligands amount to E L (L 1 ) = 250 mV and E L (L 2 ) = 450 mV vs FcH/FcH + , respectively. Indeed, the electrochemical potential E 1/2 = 700 mV of 3 perfectly corresponds to the sum E L (L 1 ) + E L (L 2 ) = 700 mV. Consequently, the ligand contributions to the electrochemical potential of substituted cyclopentadienyl complexes are essentially additive for 1-4.
This characteristic relationship is supported by correlating the electrochemical data with the Hammett substituent constants [37,39,74,75]. Typically, the E 1/2 data of substituted ferrocenes correlate linearly with the sum ∑σ p of the Hammett values σ p of para-substituents [37,39,74].
For esters 1-4, the electrochemical potentials E 1/2 (vs FcH/ FcH + ) versus sum of Hammett values ∑σ p did not give a satisfactory linear relation. Within this approach, the relative positions of ester groups and hence their different electronic influence to the electrochemical potential is not considered. The influence of a methoxycarbonyl substituent in 1-or 1'-position is indeed best described with σ p = 0.45 [75]. On the other hand, substituents in the 3-or 3'-position require using σ m = 0.37 [75] for meta-substituents, to give an excellent linear correlation of E 1/2 with ∑σ p/m ( Figure 2, Table 1). The generalizable use of σ p and especially σ m to include the effect on the relative positions of substituents for E 1/2 of polysubstituted ferrocenes has to be further validated with other series of polysubstituted ferrocenes.
In contrast to the solid-state IR spectra, only a single broad C=O band is observed for 1-4 in solution ( Figure 3, Figure 4, Figures  S7-S14, Supporting Information File 1). In the series 1-4, the C=O bands shift to higher wavenumbers in solution = 1712-1724 cm −1 with increasing number of electronwithdrawing COOMe groups (Figure 3b). The DFT calculated IR spectra with unscaled energies of the C=O vibrations = 1710-1724 cm −1 fully support these findings (Table  S1, Figures S15-S22, Supporting Information File 1).    Table  S1, Supporting Information File 1).
Triester 3 and tetraester 4 cannot be quantitatively oxidized to 3 + and 4 + in the SEC cell up to a potential of 1.1 V and 1.4 V, respectively, probably due to a fast diffusion of 3 and 4 to the anode in the beam path (Figure 4, Figures S11-S14, Supporting Information File 1). In addition, precipitation of some poorly soluble [4][X] also occurs. During oxidation to the respective ferrocenium cations, the C=O stretching vibration bands of 1-4 decrease in intensity, while the C=O bands of 1 + -4 + appear, crossing in clean isosbestic points. Expectedly, the C=O stretching vibrations of 1 + -4 + are shifted to higher wavenumbers by 26-19 cm −1 ( = 1738-1743 cm -1 ) with an increasing electron-withdrawing character of the Cp ligands. The substituent effect is attenuated by the positive charge at the iron atom in 1 + -4 + ( = 5 cm −1 ), compared to 1-4 ( = 12 cm −1 ), respectively ( Figure 3b, Table S1, Figures S7-S14, Supporting Information File 1) [78]. The unscaled energies of the DFT calculated C=O bands of 1 + -3 + fit very well to the experimental observations of 1 + -3 + (Figures S7-S12, S16, S18, S20, Table S1, Supporting Information File 1). Unexpectedly, the calculated data of 4 + are significantly lower than the experimental ones, which remain unexplained at the moment.
For all redox couples of the ferrocenyl esters, the C=O stretching vibration delivers a useful in operando probe substantiating the stability of the redox mediator and enabling quantification of both redox partners and hence estimation of the actual concentration-dependent redox potential in solution.
The energy of the absorption bands decreases almost linearly with the number n of the electron-withdrawing COOMe substituents for 1-4 ( Figure 6a).  Isosbestic points indicate clean conversions of 1 → 1 + , 2 → 2 + and 3 → 3 + , respectively. For example, this set of bands and shoulders (sh) IV-I is observed at λ max = 485 nm (IV), λ sh = 559 nm (III), 601 nm (II) and λ max = 659 nm (I) for 3 + . During oxidation of 4 to 4 + , isosbestic points between the absorption bands of 4 and 4 + cannot be observed ( Figures S29  and S30, Supporting Information File 1). Probably, precipitation of the poorly soluble tetraester 4 + could be responsible for this effect, as already suggested for the IR-SEC experiments of 4/4 + . On the other hand, isosbestic points are observed in the UV-vis spectra upon re-reduction of 4 + to 4 ( Figure S31, Supporting Information File 1). The energy of the absorptions of the ferrocenium cations 1 + -4 + decreases with the electron-withdrawing nature of the Cp ligands in the series 1 + -4 + , similar to the vis absorption maxima of the neutral ferrocenes 1-4. For the prominent band I of the cations 1 + -4 + , a linear and stronger dependency of the energy on the number n of methoxycarbonyl substituents can be found than for the ligand field band of the ferrocenes 1-4 ( Figure 6b). The lowest energy band (band I) in the UV-vis spectra of 1 + -4 + is assigned to ligand-to-metal charge transfer (LMCT) transitions [79,[81][82][83]. The bands II-IV are assigned to mainly d-d transitions [79]. TD-DFT calculations on the B3LYP, def2-TZVP, RIJCOSX, ZORA, CPCM (CH 2 Cl 2 ) level do not give satisfactory results concerning energy, number of bands and oscillator strength of electronic transitions ( Figures S32-S35, Supporting Information File 1). The poor agreement of TD-DFT calculated electronic spectra of metallocenes and derivatives with experimental data has been noted before. Improvements have been achieved by testing different functionals [84,85] and by including vibrational distortions of the ferrocene geometry into the calculations [86]. Nevertheless, the LMCT character of the prominent band I is confirmed by the calculations. The intensity of band I scales with the amount of the corresponding ferrocenium ion present and consequently the actual potential in solution can be estimated by UV-vis spectroscopy.

NMR spectroscopy of esters 1-4 and 1 + -4 +
In contrast to typical organic paramagnetic redox mediators, the relaxation properties of proton nuclei of paramagnetic ferrocenium derivatives allow the observation of reasonable sharp resonances [87]. The fast electron self-exchange of the ferrocene/ ferrocenium redox couple and derivatives on the NMR timescale leads to the observation of resonances with averaged chemical shifts δ in the 1 H NMR spectra of ferrocene/ferrocenium mixtures [5,6,[70][71][72]. The molar fraction of FcH/FcH + can be calculated from the averaged 1 H NMR resonance frequencies of a mixture and the known resonance frequencies of FcH and FcH + , respectively [6]. This relation gives χ P = (δ − δ D )/( δ P − δ D ) for the molar fraction of the paramagnetic species, expressed in the chemical shift scale with δ D being the chemical shift of the diamagnetic species, δ P being the resonance of the paramagnetic species and δ being the averaged chemical shift of the mixture.
In some cases, e.g., 3 + , the different Cp protons can still be distinguished in spite of the broadened resonances ( Figure 8). The broadening is much more severe for the Cp proton resonances, while the methyl proton resonances are still rather sharp allowing the discrimination and assignment of the different methyl protons of 3 + (Figure 8).
This substituent effect is larger for the paramagnetically shifted resonances of 1 + -4 + than for the diamagnetic complexes 1-4.
In CD 3 CN, the treatment of 3 with Magic Green led to the disappearance of the resonances of 3. However, paramagnetically shifted resonances of 3 + are absent suggesting that the initially formed 3 + undergoes further reactions with the coordinating solvent CD 3 CN ( Figure S39, Supporting Information File 1). This finding underscores that the solvent has to be carefully chosen with respect to the mediated reaction and stability of the mediator.
From the observed 1 H NMR chemical shifts -either of the cyclopentadienyl or methyl resonances -the relative concentrations of the ferrocene and ferrocenium ion can be extracted, again allowing the estimation of the actual potential in solution by spectroscopic techniques.

Conclusion
Ferrocenyl esters 1-4 with one to four ester substituents are reversibly oxidized to the respective ferrocenium cations 1 + -4 + , spanning a broad electrochemical potential range from 260 mV for 1 to 900 mV for 4 vs the ferrocene/ferrocenium redox couple. The electrochemical potentials E 1/2 of 1-4 correlate linearly with the sum of Hammett substituent parameters ∑σ p/m . However, the position of ester substituents has to be taken into account by employing σ p for 1-and 1'-substituents and σ m for  With all the data in hand, the molar fraction of the ester-substituted redox couples 1/1 + -4/4 + can be accessed a) from the C=O stretching vibrations of the ester groups, b) the ferrocenium CT bands or c) from the averaged 1 H NMR chemical shifts of the Cp or ester methyl protons. Ongoing investigations focus on the spectroscopic monitoring of 1-4 as redox mediators in selected electrosynthetic transformations.

Experimental
Dichloromethane, CD 2 Cl 2 and CD 3 CN were distilled from calcium hydride. Electrochemical experiments were carried out on a BioLogic SP-50 voltammetric analyzer using a platinum working electrode, a platinum wire as counter electrode, and a 0.01 M Ag/AgNO 3 CH 3 CN electrode as reference electrode. The measurements were carried out at a scan rate of Spectroelectrochemical experiments were performed using a Specac omni-cell liquid transmission cell with CaF 2 windows equipped with a Pt-gauze working electrode, a Pt-gauze counter electrode and an Ag wire as pseudo-reference electrode, meltsealed in a polyethylene spacer (approximate path length 0.5 mm) in dichloromethane (68, 35, 13, 2 mM solutions of 1-4 in CH 2 Cl 2 , containing 0.1 M [n-Bu 4 N][B(C 6 F 5 ) 4 ]) [88]. UV-vis/near-IR spectra were recorded on a Varian Cary 5000 spectrometer using 1.0 cm cells (Hellma, Suprasil). IR spectra were recorded on a Bruker Alpha FTIR spectrometer with ATR unit, containing a diamond crystal.
DFT calculations were carried out using the ORCA program package (version 4.0.1) [90]. All calculations were performed using the B3LYP functional [91][92][93] and employ the RIJCOSX approximation [94,95]. Relativistic effects were calculated at the zeroth order regular approximation (ZORA) level [96]. The ZORA keyword automatically invokes relativistically adjusted basis sets. To account for solvent effects, a conductor-like screening model (CPCM) modeling dichloromethane was used in all calculations [97]. Geometry optimizations and TD-DFT calculations (50 vertical transitions) were performed using Ahlrichs' split-valence triple-ξ basis set def2-TZVP which comprises polarization functions for all non-hydrogen atoms [98,99]. The presence of energy minima was checked by numerical frequency calculations. Explicit counterions and/or solvent molecules were not taken into account.

Supporting Information
The Supporting Information file contains square wave voltammograms, IR and UV-vis spectra of the spectroelectrochemical experiments, (TD)-DFT calculated IR and UV-vis spectra, a table with IR data, 1 H NMR spectra of the oxidation titration experiments and Cartesian coordinates of DFT calculated structures of 1-4.

Supporting Information File 1
Mediators measured and calculated spectra, IR data and Cartesian coordinates.