Redox-Rich Metallocene Tetrazene Complexes: Synthesis, Structure, Electrochemistry, and Catalysis

Abstract

Thermal or photochemical metal-centered cycloaddition reactions of azidocobaltocenium hexafluoridophosphate or azidoferrocene with (cyclooctadiene)(cyclopentadienyl)cobalt(I) afforded the first metallocenyl-substituted tetrazene cyclopentadienyl cobalt complexes together with azocobaltocenium or azoferrocene as side products. The trimetallic CpCo compounds are highly conjugated, colored, and redox-active metallo-aromatic compounds, as shown by their spectroscopic, structural, and electrochemical properties. The CpCo-tetrazenido complex with two terminally appended cobaltocene units catalyzes electrochemical proton reduction from acetic acid at a mild overpotential (0.35 V). Replacing cobaltocene with ferrocene moieties rendered the complex inactive toward catalysis.

Introduction

Tetraaza-1,3-butadiene (”tetrazene”) ligands (R–N=N–N=N–R) are unstable as free ligands but can be prepared in situ from organic alkyl or aryl azides in the coordination sphere of suitable middle- to late-transition-metal fragments,¹ thereby affording stable metallo-aromatic compounds.² The main interest in metal-tetrazene complexes stems from their intriguing electronic structures^1,2 due to the often redox-noninnocent character of the tetrazene ligand (Scheme 1). The latter may exist as a diamagnetic neutral (A), paramagnetic monoanionic (B), or diamagnetic dianionic (C) species, depending on the redox potential of the coordinated metal/coligand entity.²

Scheme 1. Electronic Structures of Redox-Noninnocent Tetrazene Ligands and Tetrazenido Complexes.

Abbreviations: R = alkyl or aryl; M = middle to late transition metal; L = coligand.

In this contribution, we combine tetrazene coordination chemistry with metallocene chemistry, aiming at electronically and catalytically interesting, redox-responsive compounds. Using recently accessible azidocobaltocenium hexafluoridophosphate (CcN₃⁺PF₆^–; Cc = cobaltoceniumyl)³ and azidoferrocene (FcN₃, Fc = ferrocenyl)⁴ as starting materials, we report on the first metallocenyl-substituted tetrazene cyclopentadienyl-cobalt (CpCo) complexes. The choice of cobaltoceniumyl and ferrocenyl substituents as well as CpCo as the central metal fragment is governed by their very well behaved reversible redox chemistry. In combination with the highly conjugated, quasi-aromatic CpCo(tetrazene) “cobaltole” metallacycles, we expect rich electrochemistry, switchable redox states, and potential applications in photo or redox catalysis: e.g., in artificial photosynthesis for solar-driven water splitting.⁵ Furthermore, the highly polar, cationic cobaltoceniumyl substituents allow solubility in polar solvents such as water, an advantageous feature for green chemistry applications. Ferrocenyl and cobaltoceniumyl moieties have inversely addressable redox couples (Fc, Fe(II)/Fe(III); Cc, Co(III)/Co(II)/Co(I)), which will allow a study of the influence of this difference in redox-catalytic applications.

Results and Discussion

Synthesis

CpCo tetrazene complexes have been known for 50 years and are prepared by thermal reaction of an excess of an alkyl or more commonly an aryl azide with cobaltocene or dicarbonyl(cyclopentadienyl)cobalt(I), respectively.^1,2a,2d,2e These reactions are far from being chemoselective and usually afford rather low yields (<50%) with unidentified diamagnetic and paramagnetic byproducts. In contrast to these simple organic azide substrates,^2a,2d,2e the organometallic starting materials CcN₃⁺PF₆^– (1) and FcN₃ (2) required distinct reaction conditions, due to their different polarities, charges, and donor/acceptor character (Scheme 2).

Scheme 2. Synthesis of Tetrazene Complexes 4 and 5 and Azometallocenes 6 and 7.

We found that, as a starting material for the CpCo fragment of the tetrazene complexes, cyclooctadiene(cyclopentadienyl)cobalt(I)⁶ (3) is superior to the traditionally used dicarbonyl(cyclopentadienyl)cobalt(I),^1,2a,2d,2e which is easily rationalized by the more labile diolefin ligand in comparison to strong-field CO ligands. The generally assumed mechanism of these reactions involves initial formation of a metal-nitrene/imido complex with elimination of N₂ followed by a [2 + 3] cycloaddition reaction with a second equivalent of the azide.¹ Due to these singlet/triplet-nitrene ((S/T)-nitrene) intermediates, their dimerization via their T-nitrenes to 1,2-diazenes (azo compounds) is always encountered in these reactions, depending on the reaction conditions and on the population of the T state. To maximize the yield of the desired tetrazene complexes 4 and 5, the formation of their T-nitrenes has therefore to be suppressed as much as possible. It is well-known that acceptor-substituted nitrenes, e.g. pentafluorophenylnitrene,⁷ predominantly exist in the S state. Cationic cobaltoceniumyl is a strongly electron withdrawing group. Therefore, it proved possible to synthesize dicobaltoceniumyl-tetrazene complex 4 by photochemical activation of azidocobaltocenium hexafluoridophosphate³ (1) at room temperature in an excellent yield of 90% in the crude product mixture according to NMR analysis. Byproducts include only 1% of azocobaltocenium bis(hexafluoridophosphate) (6) and 6% of aminocobaltocenium hexafluoridophosphate (8),³ formed via hydrogen abstraction by cobaltoceniumylnitrene from the solvent. However, as is always the case in cobaltocenium chemistry, purification of mixtures of (di)cationic species is challenging, as chromatographic separation is usually not possible. Purification by fractional crystallization resulted in a severe loss of material, providing analytically pure 6 in 57% yield (see the Experimental Section). It is remarkable that such a photochemical route is possible at all, given the fact that known aryl tetrazene-CpCo complexes (aryl = phenyl, pentafluorophenyl)^2e are photochemically unstable. Dicationic tetrazene complex 4 therefore represents a photochemically as well as thermally stable compound (no degradation during 6 h of irradiation with a 700 W high-pressure mercury lamp; mp > 300 °C), properties that are highly beneficial for potential applications in photo/redox catalysis in polar media (vide infra).

In contrast to azidocobaltocenium hexafluoridophosphate (1), azidoferrocene⁴ (2) is a donor-substituted, slightly air and light sensitive azide. In this work, an improved and reliable synthetic protocol for 2 was developed (see the Experimental Section), making it now available in 89% yield, a significant improvement over the reported yield of 69%.^4b Since diferrocenyl-tetrazene complex 5 proved to be unstable toward air and light (vide infra), no photochemical synthesis was possible. Reaction of 2 with 3 at room temperature gave no satisfactory conversion; therefore, thermal activation proved necessary (Scheme 2). Unfortunately, under these conditions the desired 5 is obtained after chromatographic workup in only 22% yield, together with 15% of azoferrocene⁸7, the dimerization product of T-ferrocenylnitrene. Whereas azoferrocene 7 is a known compound that has been frequently studied,⁸ its cobaltocenium analogue 6 represents a new dimetallocenyldiazene of interest in electrochemistry. To make 6 available on a preparative scale, a chemoselective synthesis via oxidation of aminocobaltocenium hexafluoridophosphate (8) was developed (Scheme 3). After some screening of potential oxidizing agents, the HgO/I₂ reagent⁹ proved best suited for our purposes, affording 6 in 87% isolated yield. In contrast to photolabile azoferrocene 7, azocobaltocenium bis(hexafluoridophosphate) (6) is a highly stable, polar compound, soluble even in water (see the Experimental Section).

Scheme 3. Synthesis of Azocobaltocenium 6.

Solid-State Structures

Single-crystal structure analyses are available for all three of the new products 4–6 (Figures 1–3). Overall, these cobaltoceniumyl-substituted compounds contain regular, undistorted cobaltocenium moieties with structural metrics comparable to those of other monofunctionalized cobaltocenium derivatives.^3,10 The two tetrazene complexes 4 and 5 have similar tetraazacobalta metallacycles, comparable to that of bis(pentafluorophenyl)tetrazene(cyclopentadienyl)cobalt.^2d Most importantly, the central N–N bonds (4, N(2)–N(3) = 1.272(9) Å; 5 N(2)–N(3) = 1.292(5) Å) are shorter than the terminal N–N bonds (4, N(1)–N(2) = 1.365(9) Å, N(3)–N(4) = 1.367(8) Å; 5, N(1)–N(2) = 1.351(5) Å, N(3)–N(4) = 1.348(5) Å), supporting the proposition that the electronic structure C of Scheme 1 is the accurate description in both cases. These dianionic ligands are therefore named in the following as “tetrazenido” ligands and the oxidation state 3+ is assigned to the central cobalt of these compounds. Although dicationic dicobaltoceniumyl-tetrazenido complex 4 and neutral diferrocenyl-tetrazenido complex 5 differ in their overall charge, their peripheral metallocen(ium)yl substituents are arranged in a similar syn conformation, most likely due to steric constraints imposed by the central CpCo moieties.

Figure 1.

Molecular structure of the dication of 4 with thermal ellipsoids at the 30% probability level. Selected bond lengths (Å): N(1)–N(2) = 1.365(9), N(2)–N(3) = 1.272(9), N(3)–N(4) = 1.367(8), N(1)–Co(3) = 1.834(7), N(4)–Co(3) = 1.832(6).

Figure 3.

Molecular structure of the dication of 6 with thermal ellipsoids at the 30% probability level. Selected bond lengths (Å): N(1)–N(1A) = 1.254(3).

Figure 2.

Molecular structure of 5 with thermal ellipsoids at the 30% probability level. Selected bond lengths (Å): N(1)–N(2) = 1.351(5), N(2)–N(3) = 1.293(5), N(3)–N(4) = 1.348(5), N(1)–Co(3) = 1.819(4), N(4)–Co(3) = 1.826(4).

The molecular structure of azocobaltocenium 6 (Figure 3) shows the expected trans conformation of the azo moiety with cobaltoceniumyl substituents arranged in the electrostatically preferred anti conformation. Bond lengths and angles are in line with expectations. Overall, the structure of 6 is very similar to the published structure of its neutral azoferrocene^2d analogue.

Spectroscopic Properties

¹H NMR spectra of tetrazenido complexes 4 and 5 and of azocobaltocenium 6 showed the common pattern of monosubstituted metallocene groups (s, 5H; pseudo-t, 2H; pseudo-t, 2H) with signals shifted to lower field for cobaltoceniumyl substituents (>5.5 ppm) in comparison to those of ferrocenyl substituents (<5.5 ppm) (see the Experimental Section and spectra in the Supporting Information). The additional Cp ligand in tetrazenido complexes 4 and 5 gave rise to an additional singlet of 50% intensity in comparison to the corresponding signals of the unsubstituted Cp rings of the two peripheral metallocene(ium)yl groups. ¹³C NMR spectra displayed the expected number of signals for symmetrically substituted compounds, and cobaltoceniumyl signals were again observed generally at lower field (>75 ppm) in comparison to ferrocenyl signals (<75 ppm), except for signals of the quaternary carbons (4, 129.0 ppm; 5, 115.4 ppm; 6, 119.3 ppm), indicating CpCo-tetrazenido to be a weaker donor in comparison to the azo group. The signals of the CpCo moiety were observed close to those of the metallocene(ium)yl groups (4, 81.1 ppm; 5, 78.2 ppm), indicating the pseudoaromatic character of the tetrazenido ligand framework.

Further spectroscopic characterization included HRMS, IR, and UV/vis data. High-resolution mass spectrometry of 4–6 gave signals of molecular ions in excellent agreement with calculated values, thereby proving further the identity of these compounds. In the IR spectra of dicationic tetrazenido complex 4 and azocobaltocenium 6 the presence of the hexafluoridophosphate counterions was clearly evident from the strong diagnostic ν_P–F absorptions¹¹ observed at approximately 815 and 550 cm^–1 (see the Experimental Section and spectra in the Supporting Information).

Compounds 4 (red), 5 (green), and 6 (orange) are highly colored organometallic chromophores. Figure 4 gives a comparison of the two tetrazenido complexes 4 and 5, whereas Figure 5 compares azocobaltocenium 6 with its known azoferrocene^8c analogue 7. As anticipated, donor-substituted ferrocenyl complexes 5 and 7 absorb at significantly higher wavelengths in comparison to acceptor-substituted cobaltoceniumyl compounds 4 and 6 (for an assignment of the underlying transitions see the following section). Interestingly, tetrazenido complexes 4 and 5 have comparable dye properties as chlorophylls with their distinctive Soret band at ca. 400 nm and their Q bands ranging from 600 to 800 nm.

Figure 4.

UV–vis spectra of tetrazenido complexes. 4 (CH₃CN): λ_max 388 nm (ε = 8865 L mol^–1 cm^–1), λ_max 528 nm (ε = 8805 L mol^–1 cm^–1), λ_max = 706 nm (ε = 1551 L mol^–1 cm^–1). 5 (CH₂Cl₂): λ_max 419 nm (ε = 6807 L mol^–1 cm^–1), λ_max 462 nm (ε = 7124 L mol^–1 cm^–1), λ_max 759 nm (ε = 3997 L mol^–1 cm^–1).

Figure 5.

UV–vis spectra of azometallocenes. 6 (CH₃OH): λ_max 292 nm (ε = 13200 L mol^–1 cm^–1), λ_max 356 nm (ε = 5220 L mol^–1 cm^–1), λ_max 427 nm (ε = 1400 L mol^–1 cm^–1); 7 (CH₂Cl₂): λ_max 321 nm (ε = 11200 L mol^–1 cm^–1), λ_max 378 nm (ε = 2887 L mol^–1 cm^–1), λ_max 529 nm (ε = 3084 L mol^–1 cm^–1).

Electrochemistry, UV/Vis/NIR Spectroelectrochemistry, and Quantum Chemistry

In the following, we discuss the electrochemical properties of the two azometallocene (or metallocenium) and the trinuclear tetrazenido complexes along with their electronic spectra in all experimentally accessible oxidation states. Assignments of the respective bands as based on experimental observations and (TD)DFT calculations are also provided. Electrochemistry data are collected in Table 1, whereas Table 2 compiles the data pertinent to the electronic spectra.

Table 1. Cyclic Voltammetry Data of the Complexes^a.

complex	E1/2Mc/Mc+ (mV), McIII-Co-McII/III	E1/2Mc/Mc+ (mV), McII/III-Co-McII	E1/2CpCo(bridge) (mV), Mcn+-CoIII/II-Mcn+	ΔE1/2b (mV)
4c	–1585	–1775	–600	985/190
4d	–1600	–1840	–675	925/240
5d	23	–66	–1695	89/1629
5e	86	–28	–1660	114/1632
6c	–600	–890	n.a.	290
7e	332	118	n.a.	214

^aAll data in mV vs Cp₂Fe^0/+ at v = 0.1 V/s.

^bHalf-wave potential difference between two consecutive redox processes.

^cIn CH₃CN/NBu₄⁺PF₆^– (0.1 M).

^dIn THF/NBu₄⁺PF₆^– (0.1 M).

^eIn CH₂Cl₂/NBu₄⁺PF₆^– (0.1 M).

Table 2. UV/Vis/NIR Data for the Complexes in Charge State n^a.

complex	charge state n	λmax (nm) (εmax (10–3 M–1 cm–1))
4a	2+	388 (8.9), 526 (8.3), 711 (1.5)
	+	476 (8.4), 941 (2.5)
	0	440 (6.6), 546 (7.2), 1031 (1.6)
	–	452 (8.69), 520 (sh), 780 (3.1)
5b	0	414 (8.8), 463 (8.8), 750 (4.4)
	–	345 (13.5), 505 (5.4)
6a	2+	354 (5.5), 429 (sh, 1.4)
	+	465 (2.2), 557 (2.9), 1135 (1.6)
	0	423 (sh, 2.2), 569 (5.8)
7c	2+	361 (sh), 965 (0.2)
	+	358 (9.6), 641 (2.4), 1634 (0.4)
	0	321 (11.2), 379 (3.1), 530 (3.1)

^aIn CH₃CN/NBu₄⁺PF₆^– (0.1 M).

^bIn THF/NBu₄⁺PF₆^– (0.1 M).

^cIn CH₂Cl₂/NBu₄⁺PF₆^– (0.1 M).

The electrochemistry of azoferrocene 7 as well as the vis/NIR spectroscopic characteristics of its mixed-valent radical cation have already been reported on earlier occasions.^8d,12 Our own observations during the present study are in perfect agreement with these previous results, providing a pattern of two consecutive, reversible one-electron voltammetric waves with a significant half-wave potential splitting ΔE_1/2 of 214 mV and identical half-wave potentials despite the NBu₄⁺PF₆^– supporting electrolyte being used instead of NBu₄⁺ClO₄^– (see Table 1 and Figure S19).^8d We also confirmed the partial bleaching and red shift of the prominent metal to ligand charge-transfer (MLCT) band from 530 to 641 nm and the appearance of a weak intervalence charge-transfer band at 1634 nm on one-electron oxidation in a spectroelectrochemical setup (see Table 2 and Figure S20).¹³ The slightly different values of 672 and 1560 nm in ref (8d) might be due to an inflection of their spectra to negative values, which distorts the MLCT band, and the differences in the supporting electrolyte. Our setup also allowed us to monitor the spectroscopic changes concomitant with further oxidation of 7⁺ to dication 7²⁺, which is accompanied by the bleaching of the IVCT and the vis bands (Figure S20). The vis band of 7⁺ is therefore reassigned as a shifted MLCT instead of a LMCT (ligand to metal charge transfer) band, in contrast to the previous proposal (note that, in the case of an LMCT, this band would be expected to further red shift and increase in intensity on further oxidation as the number of ferrocenium electron acceptors doubles).

The electrochemical behavior of azocobaltocenium 6 is just a mirror image of that of 7, showing two consecutive and likewise reversible one-electron reductions at −600 and −890 mV and an enhanced half-wave potential splitting of 290 mV in comparison to its ferrocene congener (see Table 1 and Figure S21). The large anodic shifts of both waves in comparison to the cobaltocenium/cobaltocene (Cc⁺/Cc) redox couple (E_1/2 = −1330 mV)¹⁴ is a clear testimony to the strongly electron withdrawing character of the (acceptor-substituted) azo ligand. During the first reduction of dicationic 6 to its mixed-valent radical cation, we observe the growth of a considerably more intense IVCT band (Cc⁺ → Cc) at a much lower wavelength/higher energy (1135 nm/8820 cm^–1) in comparison to that for the azoferrocene radical cation (1634 nm/6120 cm^–1) during the first reduction of dicationic 6 to its mixed-valent radical cation (Table 2 and Figure S22) in UV/vis/NIR spectroelectrochemical experiments. This is indicative of a stronger electronic coupling in the mixed-valent state. We also note the growth of a prominent vis absorption feature with a main peak at 557 nm and a smaller peak at 465 nm (Table 2). During the second reduction to neutral azocobaltocene (Figure S22 and Table 2), the absorptivity of this band further increases along with a slight red shift, while the IVCT band bleaches. This again argues for an MLCT character with charge transfer from the electron-rich Cc site(s) to the electron-accepting azo bridge, just as in azoferrocene.

As already mentioned, the tetrazenido ligand in the trinuclear complexes 4 and 5 is present in its dianionic state, thus providing a considerably more electron-rich connector as opposed to the neutral azo bridge in complexes 6 and 7. As a consequence, the three reversible one-electron reduction waves of 4 at −600, −1585, and −1775 mV (see Figure 6) are assigned to the stepwise reductions of first the central CpCo^III(tetrazenido) fragment and then the individual Cc⁺ entities. Thus, replacing the azo bridge of 6 (d_Co···Co = 6.611 Å) by the more extended CpCo^II-1,4-tetrazenido bridge (d_Co···Co = 7.682 Å), which is present after the first reduction of 4, decreases ΔE_1/2 from 290 to 190 mV.

Figure 6.

Cyclic (left) and square-wave voltammogram (right) of complex 4 in CH₃CN/0.1 M NBu₄⁺PF₆^– (v = 0.1 V/s or square-wave frequency 50 Hz, step height 2 mV, at room temperature).

That assignment of redox events is further supported by our quantum chemical calculations, which place the LUMO almost exclusively on the CpCo(tetrazenido) fragment of 4 (for a compilation of relevant frontier MOs see Figure S23). Consequently, all three vis bands in the electronic spectrum of 4 are assigned as LMCT from occupied, ligand-based MOs, spread over mainly the tetrazenido bridge with lesser contributions from the attached Cp rings, to the central CpCo^III fragment. Figure 7 (top) provides graphical accounts of the charge density differences concomitant with the individual absorptions of 4, while Figure S24 demonstrates the good match between the computed and the experimental spectra.

Figure 7.

Changes in the UV/vis/NIR spectra on stepwise reduction of complex 4 to the radical cation (top), the neutral state (middle), and the anion (bottom) (CH₃CN/0.1 M NBu₄⁺PF₆^–, room temperature) and charge density different plots for the individual transitions with the corresponding experimental wavelengths. The blue color indicates a loss and the red color a gain of electron density.

As is shown in Figure 7 (top), the first one-electron reduction causes the bleaching of the least intense vis absorption at 711 nm and the growth of a much stronger band at an even lower energy (λ_max 941 nm, 10640 cm^–1) and merges the two more intense vis bands into an equally intense, composite band at an intermediate position (Table 2). With reference to our quantum chemical calculations, the transition underlying the low-energy band can be assigned as metal to metal charge transfer (MMCT) from the CpCo^II entity to the appended Cc⁺ sites, while the absorptions at higher energies have more of an MLCT origin with charge transfer from the same donor moiety to the central azo unit of the tetrazenido bridge and mainly the ipso atoms of the Cc⁺ Cp ligands. Charge density difference plots for these transitions are also provided in Figure 7, while Figure S26 compares the experimental and computed spectra.

During the second reduction of 4 (Figure 7, middle), the intensity of the NIR band diminishes somewhat along with a red shift by 930 cm^–1 to 1031 nm (Table 2). Moreover, the prominent vis absorption splits into two separate bands at higher and lower energies. According to our quantum chemical calculations, doubly reduced 4^2– has a triplet ground state. The principal character of the NIR band as MMCT from the central CpCo^II(tetrazenido) fragment to the remaining Cc⁺ site is retained with hardly any admixture of intervalence charge transfer (IVCT) from reduced Cc to Cc⁺. The vis bands are more mixed in character with, again, MMCT between CpCo^II(tetrazenido) and Cc⁺, augmented by MLCT at the Cc⁺ entity or have dominant MLCT within the Cc unit (see Figure 7, middle; for a comparison of computed and experimental spectra see Figure S27).

Triply reduced 4^3– has three paramagnetic Co^II centers and, according to our calculations, a quartet ground state where all three spins are aligned parallel. Our calculations assign the remaining vis absorptions at 780, 520, and 452 nm (see Figure 7, bottom) as charge transfer from the central CpCo^II(tetrazenido) to the Cc units.

Cyclic and square wave voltammograms of complex 5 as represented by Figure 8 show two closely spaced one-electron waves for the stepwise oxidation of the peripherally appended ferrocenyl substituents and a reduction of the central CpCo^III(tetrazenido) unit. As was found for the comparison of azaferrocene 7 and azacobaltocenium 6, the half-wave potential splitting for stepwise oxidation of the ferrocene sites is substantially smaller than that observed for its cobaltocenium analogue. In 5, these two oxidations are merged into one composite wave or peak, and half-wave potential splitting as estimated from deconvolution of the square-wave peak is down to 89 mV in THF/NBu₄⁺PF₆^– or 114 mV in CH₂Cl₂/NBu₄⁺PF₆^–, in comparison to 190 mV in 4 (see Figure 8 and Figure S28). Replacing the cationic Cc⁺ appendants by neutral ferrocenes shifts the reduction potential of the CpCo^III(tetrazenido) entity by almost 1100 mV, to −1660 (CH₂Cl₂) or −1700 mV (THF). This provides clear evidence of how strongly the appended metallocene/metallocenium entities influence the electronic properties of the tetrazenido bridge and, thus, the entire trinuclear arrays. In CH₂Cl₂, the reduction of 4 is followed by a chemical step which consumes the reduced form at the expense of a new species, which itself is being reduced at a peak potential of −1.90 V (see Figure S28 for a representative CV). That process is outrun at sweep rates of 400 mV/s or faster. No such complications were encountered in THF, where reduced 4 proved stable on the time scale of the voltammetric and of the spectroelectrochemical experiments.

Figure 8.

Cyclic (left) and square-wave voltammogram (right) of complex 5 in CH₃CN/0.1 M NBu₄⁺PF₆^– (v = 0.1 V/s or square-wave frequency 50 Hz, step height 2 mV, at room temperature).

Figure 9 collects the UV/vis/NIR spectra of complex 5 in its neutral and monanionic states along with the charge density difference plots for the transitions at the lowest energies. Phenomenologically, the electronic spectrum of 5 resembles that of singly reduced 4^– (compare Figure 7 and Figure 9 and see Table 2), but with inverted donor/acceptor roles of the peripheral and central sites. Thus, while charge transfer in singly reduced 4^– occurs from the central CpCo^II unit to the appended Cc⁺ acceptors, it is directed from the outer Fc donors to the inner CpCo^III acceptor in 5 (for further charge density difference plots see Figure S29). We again note an excellent match between computed and experimental spectra (as documented in Figure S30). In the same vein, the spectrum and the nature of the electronic transitions of reduced 5^– resemble those of triply reduced 4^3–, which is also present as a monoanion with three metal ions in oxidation state 2+. Unfortunately, oxidized 5 proved to be unstable on the time scale of the spectroelectrochemical experiment in any solvent we tried, and no data can therefore be presented.

Figure 9.

Changes in the UV/vis/NIR spectra on stepwise reduction of complex 5 to the radical anion (THF/0.1 M NBu₄⁺PF₆^–, room temperature) and charge density different plots for the relevant transitions with the corresponding experimental wavelengths. The blue color indicates loss and the red color gain of electron density.

Catalytic Studies

The electrocatalytic proton reduction activity of 4 was probed in acetonitrile using acetic acid as the proton source (pK_a^MeCN = 22.3). Since the reduction potential for generating doubly reduced 4^2– or fully reduced 4^3– is more negative than the standard reduction potential of acetic acid in acetonitrile (−1.39 V), such electrocatalysis is thermodynamically possible.¹⁵ Upon sequential additions of acetic acid (from 5 to 200 mM), the two reversible reduction processes for the Cc⁺ units coalesce to a single irreversible wave at −1.65 V with an ∼12-fold increase in peak current, i_p (Figure 10A). To investigate this proton-responsive behavior, we recorded cyclic voltammograms of 4 with smaller increments of acetic acid (Figure S31). In the presence of low concentrations of acid (0–5 mM), the reduction wave at −1.78 V was rendered irreversible and the peak current was enhanced by ∼2.7-fold before plateauing at ∼5 mM acetic acid (Figure 10B). At higher acid concentration (>5 mM), the peak at −1.58 V started to grow, and above 50 mM, the two peaks eventually merged into a single reductive wave which continued to grow with increasing acid concentration. Emergence of one reductive wave at higher acid concentration suggests that protonation of the tetrazenido ligand disrupts the electronic communication between the two appended Cc⁺ units, leading to smaller half-wave potential splitting. However, it should be noted that the tetrazenido ligand in 4 does not undergo protonation by acetic acid at the resting state, as the first reduction wave at −0.6 V remained unaffected in acidic condition. Foot-of-the-wave analysis (FOWA) of the cyclic voltammograms was carried out to acquire kinetic information on the catalytic process. The overall rate constant for H₂ evolution (k_obs), which equals the maximum turnover frequency (TOF_max), was estimated to be 46.1 ± 0.3 and 59.5 ± 0.5 s^–1 in the presence of 0.05 and 0.10 M acetic acid, respectively (Figure S32A).¹⁶ Notably, 4 shows a relatively early onset potential for electrocatalytic reduction of acetic acid, which is illustrated by the catalytic Tafel plot (Figure S32B). Approximately 300 mV overpotential was required to achieve maximum turnover frequency.

Figure 10.

(A) Cyclic voltammograms of 4 (0.5 mM) with various concentrations of acetic acid recorded in 0.1 M NBu₄⁺PF₆^–/acetonitrile at a potential scan rate of 0.1 V s^–1. (B) Dependence of the reductive peak currents at −1.62 V (black) and −1.82 V (red) on the concentration of acetic acid.

Controlled-potential electrolysis was performed at −1.75 V in the presence of 0.2 M acetic acid to verify that the current enhancement at −1.65 V is caused by electrocatalytic proton reduction (Figure S33). Evolution of hydrogen gas was confirmed by gas chromatographic (GC) analysis, and the Faradaic efficiency of H₂ formation was estimated to be 80(±5)%. One hour of electrolysis at 350 mV overpotential (η) using 1.4 μmol of 4 generated 4.1 μmol of H₂, which corresponds to a turnover number (TON) of 2.9 per catalyst molecule in the electrolyte solution. Electrolysis over a longer time period (2 h) led to lower Faradaic efficiency for H₂ evolution, suggesting degradation of the catalyst.

The catalytic peak current increases linearly with the concentration of 4, consistent with a mechanism for H₂ evolution that is first order in catalyst (Figure S34). Emergence of the catalytic wave at the Cc⁺/Cc reduction potential hints that the central tetrazenido ligand may act as a proton shuttle to the central cobalt center, the likely site for electrocatalysis. Reduction of Cc⁺ units enhances the electron density around the tetrazenido ligand and facilitates its protonation. A mechanism can be tentatively proposed that subsequently involves proton transfer and reduction to generate a cobalt-hydride intermediate species, which undergoes protonation to form molecular hydrogen (Scheme S1).

On the other hand, voltammograms of 5 were unaffected by addition of acetic acid, which shows that the singly reduced CpCo^II-tetrazenido fragement is not sufficiently basic to bind a proton and mediate electrochemical H₂ evolution. This further emphasizes the influential role played by the appended metallocene units in modulating the electrochemical property of the central cobalt-tetrazenido moiety.

The cyclic voltammogram of 6 in the presence of acetic acid consists of two reversible peaks at −0.60 and −1.56 V (Figure 11). The peak currents for these two processes were double that observed for the one-electron reductions of 6 at −0.6 and −0.89 V in the absence of acid, which indicates that the new peaks correspond to two-electron processes. However, no further increase in peak current was observed upon addition of excess acetic acid, suggesting that 6 is not an electrocatayst under these conditions. Interestingly, the half-wave potential (E_1/2) of the first reduction process at −0.6 V shows an incremental anodic shift with increasing acid concentration, which suggests a proton-coupled electron transfer (PCET) process (Figure S36). The redox processes could be tentatively assigned to the reduction of the two Cc⁺ centers and 2e^– reduction of the azo linkage. In contrast, CVs of 7 display no such proton-responsive behavior (Figure S37).

Figure 11.

Cyclic voltammograms of 6 (0.5 mM) recorded in 0.1 M [NBu₄][PF₆]/acetonitrile at a potential scan rate of 0.1 V s^–1 in the absence (black trace) and in the presence of 1 mM acetic acid (red trace).

Conclusion

The first trimetallic tetrazenido complexes containing peripheral metallocenyl substituents of opposite donor/acceptor character have been prepared from cobaltoceniumyl- and ferrocenylazide, respectively, and (cyclooctadiene)(cyclopentadienyl)cobalt(I). These highly colored, conjugated, and redox-responsive complexes were fully characterized by NMR, IR, MS, UV/vis, and X-ray diffraction. The redox-noninnocent tetrazenido ligand is present in its dianionic form, thereby assigning redox state 3+ to the central cobalt atom. Detailed spectroelectrochemical studies in combination with DFT calculations allowed a precise assignment of electronic transitions depending on the charge/redox state of the complexes. The dicationic, highly stable, and polar dicobaltoceniumyl-tetrazenido-CpCo complex catalyzes electrochemical proton reduction in acidic solution at a modest overpotential. Furthermore, these tetrazenido complexes were compared in their properties to their related diazenes “azoferrocene” and “azocobaltocenium bis(hexafluoridophosphate)”, respectively.

Experimental Section

General Considerations

Standard methods and procedures of organometallic synthesis were applied. Chemicals were obtained commercially and used as received. Solvents were dried, degassed, and distilled before usage. Starting materials azidocobaltocenium hexafluoridophosphate (1),³ cyclooctadiene(cyclopentadienyl)cobalt(I)⁶ (3) and aminocobaltocenium hexafluoridophosphate³ (8) were synthesized according to literature procedures. Spectroscopic characterization methods and single-crystal structure analysis were performed as detailed recently.^10a

Cyclic Voltammetry

Cyclic voltammetry was performed in a one-compartment cell with 5–7 mL of the respective solvent and NBu₄PF₆ (0.1 M) as the supporting electrolyte. A platinum electrode (Ø = 1.1 mm, BASI) was used as the working electrode. It was polished with diamond pastes (1.5 and 1 μm particle size) from Buehler&Wirtz. A computer-controlled BASi EPSILON potentiostat was used for recording of the voltammograms. An Ag/AgCl wire pseudo reference electrode and a Pt wire as an auxiliary electrode were used in the measurements. The cell was connected to an argon gas bottle. Potential calibration was performed by adding appropriate quantities of decamethylferrocene (Cp*₂Fe) or Cc⁺PF₆^– after all scans of interest had been acquired. Potentials are reported against the ferrocene/ferrocenium (Cp₂Fe^0/+) couple, which is 550 mV positive of the Cp*₂Fe^0/+ couple and 1330 mV positive of the Cc⁺/Cc couple under our conditions.

IR and UV/Vis/NIR Spectroelectrochemistry

IR and UV/vis/NIR spectroelectrochemistry was performed in a self-built optically transparent thin-layer electrolysis (OTTLE) cell according to the design of Hartl.¹³ A platinum minigrid as the working and counter electrodes and a thin silver foil as reference electrode were welded in a polyethylene spacer, incorporated into a Teflon housing with electrical connectors, and sandwiched between the CaF₂ plates of a conventional liquid IR cell. 1,2-C₂H₄Cl₂/0.1 M NBu₄PF₆ was used as the supporting electrolyte. IR/NIR spectra were recorded on a FT-IR Bruker Tensor II instrument. The UV/vis/NIR measurements were performed on a TIDAS diode array unit by j&m ANALYTIK AG with a spectroscopic window of 250–2100 nm. A WENKING POS3 computer-controlled potentiostat was used for controlled-potential electrolysis.

Density Functional Theory (DFT) Calculations

These calculations were performed on the full model complexes using the GAUSSIAN 09 program package.¹⁷ Geometry optimizations were performed without any symmetry constraints. Electronic transitions were calculated by the time-dependent DFT (TD-DFT) method. Within G09 calculations the quasi-relativistic effective core pseudopotentials and the corresponding optimized set of basis functions for Ru were used.¹⁸ Polarized double-ζ basis sets (6-31G(d), geometry optimization) were employed together with the pbe1pbe functional.¹⁹ Solvation effects were modeled by the polarizable continuum model (PCM)²⁰ in TD-DFT calculations.

Catalytic Studies

Cyclic voltammetry was performed on a PalmSens EmStat potentiostat using a conventional three-electrode setup with a glassy-carbon working electrode (3 mm diameter), Pt-wire counter electrode, and a Ag/AgCl/KCl (saturated) reference electrode. A 0.1 M solution of NBu₄⁺PF₆^– in MeCN was used as supporting electrolyte. Bulk electrolyses were carried out in a two-compartment H cell connected by a glass frit using a Bio-Logic science multichannel potentiostat. A glassy-carbon rod with a surface area of ∼0.2 cm² was used as a working electrode, and platinum mesh as a counter electrode. Prior to electrolysis, the electrolyte solution was deaerated by sparging N₂, and then the electrochemical cell was kept closed and gastight during the electrolysis. Typically, the volume of electrolyte in the working compartment was 6 mL, and that in the counter compartment was 4 mL. H₂ produced during electrolyses was quantified with a Shimadzu Tracera GC-2010 Plus gas chromatograph kept at 130 °C equipped with a barrier ionization discharge (BID) detector and a molecular sieve column with He as the carrier gas. Methane (2% CH₄ in N₂) was used as internal standard.

Azidoferrocene (2):⁴ Improved Synthesis

A 50 mL Schlenk vessel was charged with 22 mL of dry, deoxygenated ethanol and 1.000 g of bromoferrocene (3.775 mmol, 1.0 equiv). The mixture was cooled with stirring in an ice bath, and 0.482 g of anhydrous CuCl (4.869 mmol, 1.3 equiv) was added under protection of light. In a separate vessel, 0.550 g of NaN₃ (8.460 mmol, 2.2 equiv) was dissolved in 1.6 mL of Ar-saturated water. An aqueous solution of NaN₃ was added to the solution of bromoferrocene and CuCl, the mixture was sonicated at 0 °C for 3 h under protection of light, and stirring was continued in the dark overnight at 0 °C. Workup was performed under ambient conditions. The orange-brown reaction mixture, containing mostly yellow-white CuBr with some green copper residues, was filtered through a short column of Celite to remove the copper salts, and the column was washed with five portions of ethanol, giving an orange-brown solution of the crude product. The solvent was reduced in volume on a rotary evaporator to ∼5 mL, and 60 mL of water was added to dissolve excess NaN₃ and to achieve better phase separation in the following liquid–liquid extraction with three to five 150 mL portions of diethyl ether. The orange-yellow organic phases were combined and washed with two 100 mL portions of cold water followed by one 50 mL portion of saturated aqueous NaCl solution. The organic phase was dried over Na₂SO₄, and the solvent was stripped off on a rotary evaporator, giving an orange-brown oil of the pure product. This material was kept in the refrigerator overnight under an atmosphere of Ar. On opening of the flask in the next morning, the oil solidified spontaneously, affording 0.759 g of yellow-brown 2 (3.343 mmol, 89% yield; reported yield^4b in the literature 69%) as a highly crystalline material. Spectral data concur with published values.^4b Azidoferrocene (2) is slightly light sensitive in solution and should be stored under protection of light at low temperature.

Dicobaltoceniumyl-Tetrazenido Complex (4)

A quartz Schlenk tube was charged with 20 mL of 1,2-dichloroethane under protection from air by an argon atmosphere and cooled to −35 °C. A 0.080 g portion of azidocobaltocenium hexafluoridophosphate (1; 0.213 mmol, 2.0 equiv) and 0.025 g of freshly prepared (1,5-cyclooctadiene)(cyclopentadienyl)cobalt(I)⁶ (3; 0.107 mmol, 1.0 equiv) were added, and the reaction mixture was irradiated with a 700 W high-pressure water-cooled mercury lamp for 6 h. A dark red-brown product mixture was obtained, containing yellow-brown aminocobaltocenium hexafluoridophosphate³ and orange azocobaltocenium bis(hexafluoridophosphate) (6) as byproducts, as well as intense red product 4. The reaction mixture was placed in a freezer and cooled to −20 °C for 2 h. The black-red solid was filtered off on a Büchner funnel and thoroughly washed three times with diethyl ether, the receiving flask was replaced, and the fine powdery product was diluted with acetonitrile. Note: theoretical yield at this stage by NMR analysis is approximately 90%, with 6% of aminocobaltocenium hexafluoridophosphate (8) and 1% of azocobaltocenium bis(hexafluoridophosphate) (6) as byproducts. Purification was achieved via diffusion crystallization in acetonitrile out of diethyl ether overnight at room temperature concomitant with significant yield losses. Product 4 crystallized as black-red needles and was filtered off on a Büchner funnel. Only the needles were collected, and the powder in the pores was discarded. Note: the powder in the pores can only be gained by dissolution with acetonitrile and still contains ∼1% of the azocobaltocenium bis(hexafluoridophosphate) (6) byproduct. Drying in vacuo afforded 0.051 g (0.060 mmol) of 4, representing a yield of 57%. Compound 4 is intensely red and highly air, heat, and water stable. It dissolves well in acetonitrile, acetone, and dimethyl sulfoxide, moderately well in methanol and water, and slightly in 1,2-dichloroethane and dichloromethane. ¹H NMR (300 MHz, CD₃CN): δ 5.49 (s, 5H, Co-Cp), 5.61 (s, 10H, Cc-Cp), 5.84 (pseudo-t, 4H, J = 2.1 Hz, C3/C4 of substituted Cc-Cp), 6.60 (pseudo-t, 4H, J = 2.3 Hz, C2/C5 of substituted Cc-Cp) ppm. ¹³C NMR (75 MHz, CD₃CN): δ 79.1 (C3/C4 of substituted Cc-Cp), 81.1 (Co-Cp), 82.9 (C2/C5 of substituted Cc-Cp), 129.0 (quart. carbon of substituted Cc-Cp) ppm. MS (MALDI pos, [m/z]): 700.96 [M – PF₆^–]⁺. IR (ATR): 3125 (ν_C–H), 2927 (ν_C–H), 1726, 1462 (ν_C=C), 1420 (ν_C=C), 1260, 1031, 814 (ν_P–F), 554 (ν_P–F), 515, 470 (ν_as(Co-ring)) cm^–1. UV/vis (CH₃CN): λ_max 388 nm (ε = 8865 L mol^–1 cm^–1), λ_max 528 nm (ε = 8805 L mol^–1 cm^–1), λ_max 706 nm (ε = 1551 L mol^–1 cm^–1). Complex 4 crystallizes readily via diffusion crystallization in acetonitrile out of diethyl ether overnight at room temperature, and single crystals were obtained in an analogous manner: mp > 300 °C. Spectra and crystallographic data are given in the Supporting Information.

Diferrocenyl-Tetrazenido Complex (5) and Azoferrocene (7)

A 50 mL round-bottom Schlenk flask equipped with a reflux condenser was charged with 15 mL of 1,2-dichloroethane, 0.200 g of azidoferrocene 2 (0.881 mmol, 2.0 equiv), and 0.102 g of freshly prepared (1,5-cyclooctadiene)(cyclopentadienyl)cobalt(I)⁶ (3; 0.440 mmol, 1.0 equiv) under protection from air by an argon atmosphere. After the mixture was refluxed overnight, a dark solution was obtained, containing black byproduct, pink azoferrocene 7, and green product 5. The reaction mixture was cooled to a heating bath temperature of 40 °C, and the solvent was removed in vacuo. Note: unreacted (1,5-cyclooctadiene)(cyclopentadienyl)cobalt(I) sublimes under these conditions and can thus be removed. For purification the reaction mixture was dissolved in 5 mL of dichloromethane and applied on a previously conditioned neutral aluminum oxide column (solvent: pentane/diethyl ether, v/v 3/2). Chromatographic details: 400 mL of solvent 1 (pentane/diethyl ether, v/v 3/2) to remove remaining (1,5-cyclooctadiene)(cyclopentadienyl)cobalt(I) (3), 300 mL of solvent 2 (pentane/diethyl ether, v/v 1/1) to enhance separation, 500 mL of solvent 3 (diethyl ether) to elute pink azoferrocene 7, 100 mL of solvent 4 (acetonitrile/diethyl ether, v/v 1/1) for washing, 500 mL of solvent 5 (acetonitrile) to elute green product 5. Note: the amount of dry, pink azoferrocene 7 was 0.026 g (0.065 mmol), 15% yield. Spectroscopic data of 7 concur with published values.^8c The green acetonitrile fraction was collected, and the solvent was removed on a rotary evaporator. Drying in vacuo afforded 0.053 g (0.096 mmol) of green powdery 5, representing a yield of 22%. Compound 5 is slightly light as well as air sensitive. It dissolves well in dichloromethane, 1,2-dichloroethane, and chloroform. ¹H NMR (300 MHz, CDCl₃): δ 4.24 (s, 10H, Fc-Cp), 4.43 (pseudo-t, 4H, J = 2.0 Hz, C3/C4 of substituted Fc-Cp), 5.01 (s, 5H, Co-Cp), 5.30 (pseudo-t, 4H, J = 2.0 Hz, C2/C5 of substituted Fc-Cp) ppm. ¹³C NMR (75 MHz, CDCl₃): δ 65.9 (C3/C4 of substituted Fc-Cp), 67.0 (C2/C5 of substituted Fc-Cp), 70.7 (Fc-Cp), 78.2 (Co-Cp), 115.4 (quaternary carbon of substituted Fc-Cp) ppm. MS (ESI pos, [m/z]): 551.00 [M + H]⁺. IR (ATR): 3078 (ν_C–H), 2923 (ν_C–H), 1725, 1448 (ν_C=C), 1411 (ν_C=C), 1251, 998, 813 (δ_op(C–H)), 523, 473 (ν_as(Co-ring)) cm^–1. UV/vis (CH₂Cl₂): λ_max 419 nm (ε = 6807 L mol^–1 cm^–1), λ_max 462 nm (ε = 7124 L mol^–1 cm^–1), λ_max 759 nm (ε = 3997 L mol^–1 cm^–1). Single crystals of 5 were obtained via diffusion crystallization in dichloromethane out of pentane: mp 197 °C (dec). Spectra and crystallographic data are given in the Supporting Information.

Azocobaltocenium Bis(hexafluoridophosphate) (6)

A 100 mL round-bottom flask was charged with aminocobaltocenium hexafluoridophosphate (8; 100.5 mg, 0.2879 mmol, 1 equiv), 30 mL of CH₂Cl₂, HgO (184.7 mg, 0,7200 mmol, 2.5 equiv), and I₂ (182.7 mg, 0.7200 mmol, 2.5 equiv). The dispersion was stirred for 24 h under ambient conditions without protection from air. For workup, 30 mL of diethyl ether was added and the dispersion was filtered through a paper filter. From this filter the product was eluted with acetonitrile until a yellow coloring of the solvent was no longer seen. The yellow solution was evaporated to dryness on a rotary evaporator and dried in vacuo to yield an orange powder. If ¹H NMR showed an incomplete reaction, the product could be used again in the same reaction with an equivalent amount of HgO and I₂ to yield 99% pure product. Yield: 86.9 mg (0.125 mmol, 87%). Compound 7 is air- and water-stable, soluble in acetonitrile, methanol, acetone, and nitromethane and moderately so in water. ¹H NMR (300 MHz, CD₃CN): δ 5.26 (s, 10H, Cp), 5.91 (pseudo-t, 4H, J = 2.0 Hz, C3/C4 of substituted Cp), 6.30 (pseudo-t, 4H, J = 2.0 Hz, C2/C5 of substituted Cp) ppm. ¹³C NMR (75 MHz, CD₃CN): δ 80.99 (C3/C4 of substituted Cp), 86.62 (C2/C5 of substituted Cp), 70.7 (Fc-Cp), 87.64 (Cp), 119.26 (quaternary carbon of substituted Cp) ppm. MS (ESI pos, [m/z]): 202.0062 [M]⁺/2 of dication; calcd, 202.0028. IR (ATR): 3124 (ν_C–H), 1419, 1407, 1032, 904, 813 (ν_P–F), 554 (ν_P–F), 541, 508, 466 (ν_as(Co-ring)), 428 cm^–1. UV/vis (CH₃OH): λ_max 292 nm (ε = 13200 L mol^–1 cm^–1), λ_max 356 nm (ε = 5220 L mol^–1 cm^–1), λ_max 427 nm (ε = 1400 L mol^–1 cm^–1). Single crystals of 6 were obtained from an acetonitrile solution: mp 180 °C (dec). Spectra and crystallographic data are given in the Supporting Information.

Supporting Information Available

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.organomet.8b00681.

• Spectra (¹H/¹³C NMR, IR, HRMS, UV–vis), cyclic voltammograms, DFT results, crystal data, and a mechanistic scheme (PDF)

Accession Codes

CCDC 1867090–1867092 contain the supplementary crystallographic data for this paper. These data can be obtained free of charge via www.ccdc.cam.ac.uk/data_request/cif, or by emailing data_request@ccdc.cam.ac.uk, or by contacting The Cambridge Crystallographic Data Centre, 12 Union Road, Cambridge CB2 1EZ, UK; fax: +44 1223 336033.

The authors declare no competing financial interest.