Boron-targeted H-atom transfer drives disproportionation of 2,1-benzazaborolyl radical anions beyond reversible Gomberg–Krause dimerization

Abstract

This contribution provides a detailed mechanistic insight into a remarkably complex reaction of neutral 1-Ph-2-tert-butyl-1H-1,2-benzazaborole (1) with elemental potassium in THF, yielding the known 10π-aromatic 1H-2,1-benzazaborolyl (2,1-Bab⁻) potassium salt (2), an isoelectronic species with indenyl potassium, along with a racemic mixture of a new potassium hydridoborate complex (3), formed together with 2 in ca. 1 : 1 molar ratio. Unlike derivatives of isoelectronic 1H-indene, which undergo various self-protonation processes under analogous conditions, for 1, we propose a completely different and rather complex, non-linear tandem mechanism. Key species are four enantiomeric pairs of short-lived, dimeric, diamagnetic intermediates, denoted α, β, ω₁ and ω₂, possessing structures related to Gomberg’s dimer and Krause’s adduct. Based on an exhaustive multinuclear NMR analysis of the reaction mixture, including experiments with C3-methylene deuterium-labelled starting compound 1-d_n (n ∈ {0–2}) and the observation of a primary AKIE (k_H/k_D ≫ 1), as well as a detailed analysis of full scan mass spectra and observed ions in ultrahigh-resolution LDI-MS data of the resulting mixture of isotopologues of the final product 3-d_n (n ∈ {0–4}), comprising multiple isotopomers, in combination with probabilistic calculations, electrochemical studies, and DFT calculations, we were able to reconstruct the full sequence of reaction steps and pathways. The presented results support the involvement of various H-atom transfer (HAT) processes within pairs of radical-anionic species K⁺1˙⁻, proceeding through different transition states, and, most notably, reveal an extremely rare boron-targeted HAT. These findings provide a basis for developing synthetic routes to unconventional boron hydrides.

---

Introduction

The structural and electronic diversity of boron-containing species has expanded rapidly in recent decades, encompassing (anti)aromatic ring systems,^1–5 or cationic boron centres,⁶ and, notably compounds in which boron acts as a spin carrier (Scheme 1A and D).^7–21 Particularly striking are advances in accessing highly unusual electronic states, including the formation of a boryl radical anion from a hexaaryldiboron(6) dianion²² (Scheme 1A′), as well as the isolation of nucleophilic boryl anions^23–26 (Scheme 1B) and the long-sought nucleophilic boryl dianion^27–29 (Scheme 1C). These developments highlight the increasing accessibility of electron-rich and unconventional boron species, opening new directions in main-group reactivity. Further expansion of this landscape can be achieved by introducing additional heteroatoms into the boron framework. In this regard, substitution of a C=C unit by an isoelectronic, isosteric, and isolobal B(sp²)–N(sp²) bond represents the oldest but powerful and widely employed strategy for modulating electronic structure and reactivity.³⁰ This concept has been broadly implemented across diverse molecular platforms,^31–49 including stable boron-centred radical systems featuring B(sp²)–N(sp²) motifs (Scheme 1D).^8,9,50–56

Scheme 1 Representative examples of the discussed classes of boron species (A – D) and Cp⁻ and Ind⁻ analogues containing a B(sp²)–N(sp²) bond (E–I). Summary of the electrochemical or alkali metal reduction of variously substituted IndH derivatives reported in the literature,^57–64 illustrated here for 2,3-disubstituted derivatives (J), together with the questioned reactivity of an isoelectronic 1,2-disubstituted 2,1-BabH derivative (K).

Besides these exotic species, a unique historical role is played by the 1,2-azaborolyl anion (Ab⁻) (Scheme 1E), a five-membered C₃BN ring representing a 6π-aromatic analogue of the cyclopentadienyl anion (Cp⁻). The chemistry of this system was pioneered by Schmid and co-workers in the 1980s,^65–68 while subsequent studies have focused primarily on transition- and rare-earth-metal complexes, particularly in the context of olefin polymerization^69–76 and asymmetric catalysis.^70,77

Beyond Ab⁻, several positional isomers of 10π-aromatic B(sp²)–N(sp²) congeners of the 1H-indenyl anion (Ind⁻) have emerged since the early 2000s, in some cases accompanied by their neutral IndH-like conjugated C–H acid. However, their reactivity and coordination chemistry remain largely unexplored. Among these, the most extensively studied systems are the 3a,7a-azaborindenyl anion⁷⁸ (Scheme 1F) and 1-bora-7a-azaindenyl anion (Scheme 1G).⁷⁹ In addition, several neutral 1H-1,2-benzazaboroles have been reported,^80,81 although the most prevalent B(sp²)–N(sp²) congener of IndH is a 1H-2,1-benzazaborole (hereafter 2,1-BabH; Scheme 1K).^82–88 We have recently developed a convenient synthetic route to 1,2,3-trisubstituted 2,1-BabH derivatives,^39,89 and subsequently reported the first examples of 1,2-disubstituted 2,1-Bab⁻ anions (Scheme 1I) as alkali metal salts M⁺2,1-Bab⁻ (M = Li, Na, K) obtained via mesolytic C(sp³)–C(sp³) bond cleavage in (3,3′)-bis(2,1-BabH) dimers (Scheme 1H) upon alkali metal reduction.^90,91

Given the unusual reactivity of these dimeric precursors⁹¹ and the lack of systematic mechanistic studies on B(sp²)–N(sp²) congeners of 1H-indene, we set out to investigate the reaction of a monomeric 1,2-disubstituted 2,1-BabH derivative with alkali metals (Scheme 1K). Owing to its isoelectronic relationship with IndH, a key question was whether its reactivity would follow the well-established pathways described for substituted CpH⁹² and IndH derivatives^57–64 under electrochemical or alkali metal reduction, illustrated here for 2,3-disubstituted-IndH derivatives (Scheme 1J). In these systems, single electron transfer (SET) to IndH generates radical anion (IndH˙⁻), which in aprotic media undergoes rapid “father–son” self-protonation⁹³ from (C1)H-acidic methine group of the parent substrate itself, ultimately yielding the indenyl anion (Ind⁻) and saturated indane (IndH₃) in syn-configuration around the C2(sp³)–C3(sp³) bond (Scheme 1J). In the absence of a suitable proton donor, further reduction of Ind⁻ or disproportionation of the two IndH˙⁻ leads to dianionic species (IndH²⁻), which act as Brønsted bases and reductants and are subsequently quenched via alternative protonation pathways,^64,93 affording the same two final products or indane IndH₃ with anti-configuration around the C2(sp³)–C3(sp³) bond. Overall, this sequence can be viewed as a Birch-type reduction of the C=C bond.

In contrast, analogous reduction of the B(sp²)–N(sp²) bond in the studied case of 1,2-disubstituted 2,1-BabH (Scheme 1K) is far from trivial. Formation of a B–H bond within R¹HB(sp³)←N(sp²)HR² fragment (Scheme 2; top) would require a protonation of an over-reduced nucleophilic boryl dianion (2,1-BabH²⁻) with a lone electron pair (LEP) at boron, an extremely rare and electronically demanding species (Scheme 1C).^28,29 Considering this, a more plausible pathway appeared to be simple deprotonation at the C3-methylene position, leading to formation of alkali metal salt M⁺2,1-Bab⁻ with concomitant H₂ evolution (Scheme 2; middle), analogous to the behaviour of unsubstituted IndH.⁶³

Scheme 2 Overview of the unexpected reaction outcome between 1 and potassium in THF, highlighting the final products and key observations, including the lack of reactivity of 1 toward hydride sources and of 2 toward H₂ (dashed arrows), thereby pointing to an alternative mechanism for the formation of 3. Only the 1S enantiomer of 3 is shown for clarity.

Unexpectedly, the observed reactivity deviates markedly from these scenarios. Herein, we report a previously unrecognized, non-linear tandem mechanism involving multiple thermodynamically metastable intermediates and featuring an exceptionally rare boron-targeted H-atom transfer. This study uncovers an unprecedented mode of reactivity at the interface of boron and organic radical chemistry.

Results and discussion

Initial reduction of 1H-2,1-Benzazaborole (1) with potassium: defining the scope of the study

To address the above question, a ca. 0.3 mol dm⁻³ solution of colourless 1-phenyl-2-tert-butyl-1H-2,1-benzazaborole (1; see the SI for synthesis) in THF-d₈ was treated with a potassium mirror (1 : 1 molar ratio) at room temperature under rigorously anaerobic and anhydrous conditions. Strikingly, no gas evolution (anticipated as H₂) was observed; instead, the reaction immediately produced a deep red solution (time (t) = 0). As the potassium metal was gradually consumed from the walls of the Schlenk tube (t = 45 min), the colour intensity increased. However, by the time of NMR analysis (t = 60 min), the solution had shifted to a blood-red colour. While NMR spectroscopy confirmed the formation of the expected blood red 2,1-Bab⁻ potassium salt^90,91 (2), it simultaneously revealed the formation of a second product. Unexpectedly, a racemic mixture of new potassium hydridoborate complex (3) was generated alongside 2 in approx. 1 : 1 molar ratio (vide infra) (Scheme 2), directly contradicting the anticipated reaction outcome and pointing to a fundamentally different reaction pathway.

Although monohydrogenated 3 was formed instead of the expected dihydrogenated 2,1-BabH₃ (Scheme 2, top) and the products were obtained in a 1 : 1 rather than 2 : 1 molar ratio, the N(LEP)–B⁻ fragment present in 3 remains isoelectronic and isolobal with the H–C2(sp³)–C3(sp³)–H fragment in indane (IndH₃) (Fig. 1). Consequently, the fundamental mismatch in polarity between the B(δ⁺)–H(δ⁻) bond in 3 and the C3(δ⁻)–H(δ⁺) bond in IndH₃ persists. This observation points to a mechanistic pathway for the reaction of 1 with potassium that differs fundamentally from that established for IndH.

Fig. 1 At first glance, the H–C2(sp³)–C3(sp³)–H and N(LEP)–B⁻ fragments appear analogous, suggesting a similar formation mechanism.

A direct comparison of 1 and 3 suggests that formation of 3 from 1 would formally require one equivalent of potassium hydride. However, compound 1 proved inert toward hydride sources (Scheme 2, see the SI for details). Conversely, formation of 3 from 2 would formally require one equivalent of dihydrogen. Thus, if H₂ were generated during the reaction alongside 2, it could, in principle, react in situ to give 3. Yet, no reactivity of 2 toward H₂ was observed, even at 3 bar (Scheme 2).

However, a striking feature of the reaction of 1 with potassium lies in the transient formation of multiple intermediates of unknown structure (vide infra), despite 2 and 3 being the final products. These species rapidly diminish over time and are fully converted into 2 and 3 within 90 min. Their transient nature and clean conversion into 2 and 3 strongly indicate a complex and unprecedented reaction mechanism (Scheme 2, “black box”).

Identification and characterisation of the final products 2 and 3

The identity of compound 2, a known potassium salt of the aromatic 2,1-Bab⁻ anion, was confirmed by an excellent match of the ¹H, ¹¹B, ¹³C, and ¹⁵N NMR data with those previously reported earlier^90,91 for isolated 2 prepared according to Scheme 1I. Notably, several signals render the NMR spectra of the 2,1-Bab⁻ anion in THF-d₈ highly characteristic (cf. C and D in Fig. 4).

The structure of the new potassium hydridoborate complex 3 formed alongside 2 was elucidated using 1D and 2D ¹H, ¹¹B, ¹³C, and ¹⁵N NMR correlation and exchange experiments (see SI for full NMR assignment). A cross-peak observed in the ¹H–¹⁵N HMBC spectrum at δ(¹⁵N) = −332.4 ppm (cf. −235.2 and −208.7 ppm for 1 and 2, respectively; referenced to CH₃NO₂, δ(¹⁵N) = 0.0 ppm; Fig. S5 and S18) falls within the range characteristic of B-bound N(sp³) atoms bearing a LEP.⁹¹ Together, the B and N atom formally constitute two stereogenic centres in the anionic part of 3 (vide infra). In contrast to 1, where the (C3)H₂ methylene bridge appears as a singlet in the ¹H NMR spectrum, it becomes diastereotopic in 3, giving rise to an AX spin system at 4.12 and 4.44 ppm with ²J(¹H, ¹H) = 13.0 Hz. The hydridic B–H proton resonates as a broad multiplet at 3.32 ppm, reflecting coupling to both ¹¹B and ¹⁰B nuclei. Importantly, the observed ¹J(¹¹B,¹H) coupling constant of 84.3 Hz corresponds to a sharp doublet in the ¹¹B NMR spectrum at −4.8 ppm (Fig. 4C), consistent with values reported for methylborate complexes of the same 2,1-Bab framework.⁹¹

Following work-up (see SI), complex 3 was isolated in high yield (80%) by fractional crystallization of its 18-crown-6 adduct and fully characterized. The solid-state structure of [3·(18-crown-6)·(THF)₄] was determined by single-crystal X-ray diffraction (Fig. 2). The potassium cation is coordinated by the 18-crown-6 ether ligand in the equatorial plane and by two THF molecules in axial positions, resulting in effective spatial separation from the hydridoborate anion. However, a NOE cross-peak between the tBu protons and the –CH₂–CH₂– units of 18-crown-6 in the 2D ¹H–¹H NOESY spectrum (Fig. S23) indicates that the cation remains in close proximity to the anion in the solution. Accordingly, [3·(18-crown-6)·(THF)₄] forms contact ion pairs in THF, consistent with the solid-state structure (Fig. 2), where the shortest H⋯H contact between the ionic components is 2.43 Å.

Fig. 2 Molecular structure of 1S-enantiomer of rac-(1R/1S)-[3·(18-crown-6)·(THF)₄] showing 40% probability thermal ellipsoids. In the solid state at 150 K, the nitrogen atom in the position 2 is pyramidalized, thus appearing as anti-configuration, i.e. (1S,2R)-, of the rac-(1S,2R)/(1R,2S)-configurations. Two co-crystalized molecules of THF are omitted for clarity. Selected bond lengths [Å], bonding angles [°] and torsion angle [°]: B1–N1 1.571(13), N1–C7 1.550(19), C7–C6 1.55(2), C6–C1 1.372(13), C1–B1 1.607(11), C1–B1–N1 106.5(6), B1–N1–C7 94.7(9), N1–C7–C6 121.8(13), C6–C1–B1 118.0(8), C1–B1–N1 106.5(6), C12–B1–N1–C8 119.6(9).

Despite crystallographic disorder arising from the positioning of B1, H1, N1, and C8 atoms (tert-butyl group) on special positions in the monoclinic space group C_c, the structure unambiguously reveals a racemic mixture of (1R,2S) and (1S,2R) enantiomers of the anionic fragment, with tetrahedral B1 and N1 centres. The B1–H1 bond length of 1.09(8) Å [Σ_rcov(B–H) = 1.17 Å (ref. 94)] is comparable to that in K[BH₄],⁹⁵ and the B–H stretching frequency (2324 cm⁻¹) is consistent with values reported for borohydride species.⁹⁶

Both boron and nitrogen atoms are formally sp³ hybridized, and the B–N bond therefore exhibits single-bond character, in contrast to compound 2.⁹⁰ This is reflected in the B1–N1 bond length of 1.571(13) Å [Σ_rcov(N–B) = 1.56 Å (ref. 94)]. Consequently, the C₃BN ring in 3 lacks π-electron delocalization, consistent with its colourless appearance, in contrast to the intensively coloured aromatic anion 2. In both enantiomers, the B-bound phenyl group and the N-bound tBu substituent adopt an anti-configuration. Although the presence of a stereogenic boron centre adjacent to a pyramidal nitrogen atom allows, in principle, for diastereomer formation, only a single set of NMR signals is observed (both in the isolated complex and in the reaction mixture), indicating rapid inversion at nitrogen on the NMR timescale.

Mechanistic insight into the early stages of the reaction

Upon addition of a colourless solution of 1 in THF-d₈ to a potassium mirror at −50 °C in 1 : 1 molar ratio, the reaction rate was sufficiently reduced to allow observation of three distinct colour changes during potassium consumption, followed by a less pronounced fourth transition thereafter. The first change occurred immediately (t = 0), as a deep red solution formed throughout the reaction mixture (A and A′ in Fig. 3).

Fig. 3 Observed colour changes during consumption of potassium metal (mirror) within a single batch (A–C) and in flame-sealed NMR tubes from various batches containing the final blood red solution of products 2 and 3 (D), formed over time from the initial deep red solution (A′). Notes: Images A′ and D′ correspond to different batches at comparable concentrations but at different stages of the reaction: the deep red solution immediately after complete consumption of potassium (A′) and the final blood red solution containing 2 and 3 (D′). Image D also shows a small amount of dark solid residue at the bottom of the NMR tubes precipitated over time.

The subsequent two changes of colour were observable only in the absence of stirring. When the potassium mirror was not fully submerged, a thin dark blue layer formed on its surface (Fig. 3B), likely corresponding to a radical-anionic species K⁺1˙⁻ generated by reduction of 1 (vide infra). This transient blue layer disappeared upon contact with the bulk solution, consistent with rapid reaction either with itself or with unreacted 1 still present in the mixture.

After a few seconds, a dark green species (λ_{max abs} = 636 nm; see Fig. S25) emerged at the interface with the submerged potassium (Fig. 3C) and diffused into the deep red solution. This species was observable only while unreacted potassium remained and rapidly vanished upon diffusion, again suggesting reaction with residual 1. Its behaviour is consistent with formation via over-reduction (vide infra).

Once the potassium was fully consumed (ca. 45 min under sonication at 0 °C), and the reaction mixture was allowed to warm to room temperature, a final, fourth colour change gradually occurred.⁹⁷ The deep red solution slowly faded over 90 min, yielding a blood red mixture containing exclusively 2 and rac-(1R/1S)-3 (Fig. 3D and D′), as confirmed by multinuclear NMR spectroscopy (Fig. 4C). Notably, this final mixture showed no further reactivity toward additional potassium.⁹⁸

Fig. 4 Expanded view of stacked ¹H NMR spectra (500.20 or 400.13 MHz, THF-d₈, 21 °C) and full ¹¹B NMR spectra (160.48 MHz, THF-d₈, 21 °C) recorded at various stages of the reaction between compound 1 and potassium metal in THF-d₈ (A–C). Spectrum (B) was acquired at 21 °C immediately after sample preparation. Data marked with # were previously reported.^90,91 Full spectra are provided in SI.

To characterize the species present in the deep red solution (Fig. 3A and A′), a sample from the low-temperature experiment was quickly analysed by NMR spectroscopy at room temperature immediately after complete consumption of potassium. Owing to the intrinsic instability and short lifetime (ca. 1 h) of these intermediates at room temperature, the NMR investigation was limited to basic 1D experiments. Nevertheless, the obtained spectra proved to be remarkably complex. For instance, the ¹H NMR spectrum (Fig. 4B) displayed several dozen sharp yet heavily overlapping signals spanning the entire region from aliphatic to aromatic chemical shifts (see Fig. S8 for the full spectrum).

In addition to these unidentified signals, minor resonances corresponding to the final products 2 and 3 were detected. Notably, these products were consistently present in an approximately 1 : 1 molar ratio, irrespective of the reaction progress (as defined by time). In contrast, no signals attributable to the starting material 1 were observed in any of the ¹H, ¹¹B, or ¹³C NMR spectra of the deep red solution (cf. ¹¹B NMR spectra A and B in Fig. 4).

Particularly notable was the crowded region of the ¹H NMR spectrum between 3.0 and 6.3 ppm, which contained dozens of unresolved, doublet-like resonances (Fig. 4B). Based on their non-equivalent relative intensities (1 : 0.98 : 1.45 : 1.08),⁹⁹ we proposed that these signals arise from at least four distinct spin systems, corresponding to a minimum of four structurally related species. At that stage of the research, we tentatively assigned these species, denoted α, β, ω₁ and ω₂, as products of dimerization of a radical anion K⁺1˙⁻ formed via reduction of 1 by potassium. We supposed that these species may adopt structures analogous to the classical Krause’s adduct, the dianionic dimer [Ph₃B-C₆H₅=BPh₂]²⁻ (Scheme 3), reported nearly a century ago.^8,100–102 This dimer is formed via Gomberg-type coupling of the triphenylboryl radical anion, reflecting the isoelectronic relationship between Ph₃B˙⁻ and the neutral trityl radical Ph₃C˙, which forms the well-known Gomberg’s dimer.^8,103–107 In the context of boron chemistry, a related dimerization process has more recently been described for an intramolecularly P-coordinated neutral boryl radical, which undergoes reversible Gomberg-like dimerization in solution (Scheme 3).^108,109

Scheme 3 Gomberg–/Krause-type dimers relevant to this study and their transient formation during the potassium-mediated reduction of 1, ultimately yielding a mixture of 2 and 3. In the case of K⁺1˙⁻, only the dominant resonance forms are shown (for plots of the SOMO and the spin-density map, see Fig. 6; for a list of spin densities, see Fig. S67). Redox potential is referenced to SCE. (#) Each of the intermediates α, β, ω₁ and ω₂ is formed as a racemate; thus, a total of eight stereoisomeric intermediates are present. For clarity, only one enantiomer is shown in each case. (#) – Varies from batch to batch. The origin of this phenomenon is discussed further below.

The plausibility of radical-anion dimerization in the reaction of 1 with potassium, leading to intermediates α, β, ω₁ and ω₂, was supported by several observations. Although the ¹³C{¹H} APT NMR spectrum of the deep red solution contained more than one hundred signals (Fig. S10–S12), eight resonances (in addition to those of 2 and 3) corresponding to tBu and methylene (C3)H₂ groups could be readily identified. These signals could be grouped into two sets with slightly different chemical shifts, giving rise to four signals for each substituent.

Furthermore, despite the overall complexity of the ¹¹B NMR spectra (Fig. 4B), two distinct types of signals were discernible in markedly different spectral regions. One appeared as an intense, sharp singlet at −2.2 ppm, while the other consisted of a set of overlapping broad signals in the range from 30 to 34 ppm. These regions are characteristic of tetrahedral (T_d) and trigonal planar (TP) boron centres coordinated to nitrogen,⁶⁵ respectively. This observation suggests that at least some (if not all) of the intermediates α, β, ω₁ and ω₂ contain both T_d- and TP-coordinated boron atoms within their structure, analogous to the Krause’s dimer (Scheme 3).

Importantly, all four intermediates retain the methylene (C3)H₂ moiety, as evidenced by the ¹³C{¹H} APT NMR spectra. Despite these insights, experimental structure determination remained highly challenging due to the spectral complexity and the inherent nature of these species, namely their short lifetimes and extreme sensitivity to trace amounts of air and moisture. Nevertheless, deuterium labelling ultimately enabled their elucidation (vide infra).

Subsequent reactions after initial reduction

The reaction sequence is initiated by a single-electron transfer (SET) from potassium to neutral compound 1, affording the radical-anionic species K⁺1˙⁻. In contrast to the radical anions of 2,3-disubstituted 1H-indenes,^77–82,104 this species undergoes rapid dimerization to yield dianionic dimers with two potassium counter-cations, namely α, β, ω₁ and ω₂. Notably, this dimerization is sufficiently fast that lowering the reaction temperature to −50 °C does not significantly affect the rate of formation of the deep red dimers in solution, but only slows their subsequent transformation into the final products 2 and 3.

Importantly, multiple experiments demonstrated that the deep red solution is EPR silent,¹¹⁰ while providing well-resolved NMR spectra (Fig. 4B), indicating that all four dimers α, β, ω₁ and ω₂, are most likely closed-shell singlet species. These observations initially suggested that reverse monomerization of the dimers is not thermodynamically favoured under the given conditions; however, EXSY NMR experiments and DFT calculations reveal the opposite behaviour (vide infra).

In addition to the initial SET between 1 and potassium, the pronounced colour changes observed during the early stages of the reaction suggested the involvement of further redox processes. Surprisingly, cyclic voltammetry revealed only a single cathodic redox event within the accessible potential window (down to −3.2 V; Fig. 5), corresponding to the reduction of 1 to the transient radical anion 1˙⁻. This process, with E⁰(1/1˙⁻) = −2.87 V vs. SCE, was quasi-reversible only at scan rate of 20 V s⁻¹ and became fully irreversible at lower scan rates. Such behaviour is consistent with rapid consumption of 1˙⁻ via dimerization to α, β, ω₁ and ω₂. Given the highly negative reduction potential, it is not surprising that attempts to reduce 1 using potassium amalgam were unsuccessful.^111,112

Fig. 5 Cyclic voltammogram of 1 recorded under rigorously anaerobic and anhydrous conditions in vacuum-transferred THF at 21 °C (vs. SCE). A carbon fibre microelectrode (BASi MF-2007, 11 μm) was used as the working electrode, with 0.1 M nBu₄N⁺PF₆⁻ as the supporting electrolyte. Scan rate 20 V s⁻¹. The horizontal arrow indicates the direction of the forward potential sweep.

The DFT-optimised structure of K⁺1˙⁻ (M062X/D3/def2-TZVP, PCM (THF)) exhibits an essentially planar geometry at the boron centre (Σ < 358.2°, see SI for atomic coordinates), consistent with the planar geometry of the kinetically stabilized trimesitylborane radical anion (Mes₃B˙⁻), which does not undergo Gomberg- (or Krause-) type dimerization.¹¹³

In contrast to Mes₃B˙⁻, where the C–B bonds are elongated by ca. 0.02 Å relative to neutral Mes₃B, the C7a–B bond length in K⁺1˙⁻ remains essentially unchanged, while the (i-C)–B bond is shortened by approximately 0.04 Å. This behaviour can be attributed to delocalisation of spin density (vide infra) through p orbitals, which induces co-planarization of the B-phenyl group with the 2,1-BabH ring upon reduction of 1. In the neutral precursor, these ring systems are oriented nearly perpendicular to each other.

Analysis of the bonding pattern in K⁺1˙⁻, including the alternation of C–C and C=C bonds, indicates that both the B-bound phenyl group and the C₆ ring of the 2,1-BabH core lose the aromatic character present in 1. The most pronounced structural change relative to 1 is, however, the complete loss of double-bond character in the B–N bond. The nitrogen atom adopts a formal sp³ hybridization and coordinates to the potassium cation, resulting in elongation of the B–N bond from 1.409 Å in 1 to 1.530 Å in K⁺1˙⁻.

Calculated Mülliken/NBO spin density distributions (Fig. 6) reveal that the unpaired electron in K⁺1˙⁻ is extensively delocalized, with the largest contributions located on the boron atom (0.429/0.466) and on several carbon centres, including p-Ph (0.233/0.183), o-Ph (0.209/0.143), o′-Ph (0.163/0.108), C5 (0.123/0.096), and C7 (0.135/0.090), while the nitrogen atom carries negligible spin density (−0.034/0.009). These results identify the most probable sites for the proposed dimerization (Scheme 3 and vide infra).

Fig. 6 Plots of the SOMO (left) and the Mülliken spin-density map (right) of K⁺1˙⁻ calculated at the M062X/D3/def2-TZVP (PCM, THF) level of theory.

As noted above, K⁺1˙⁻ is too short-lived to be directly observed in solution. Nevertheless, the pronounced colour changes observed during the initial stages of the redox process (Fig. 3B) may reflect its transient formation. In particular, the thin, dark blue layer temporarily formed on the potassium mirror may correspond to this elusive species. To support this hypothesis, time-dependent DFT calculations were performed (see section 5.1 in SI).

While the dark blue colour is most likely associated with K⁺1˙⁻, the dark green colour (Fig. 3C) is tentatively attributed to the formation of over-reduced species derived from the deep red dimers α, β, ω₁ and ω₂ upon further reduction by potassium metal (details in SI in sections 2.6–2.8). Notably, regardless of whether formation of the dark green species was observed across different batches, the final outcome of the reaction remained unchanged, consistently yielding compounds 2 and 3 in an approximately 1 : 1 molar ratio.

Deuteration experiments

Given the limited mechanistic insight at this stage, we formulated three key questions that needed to be addressed: (i) what are the structures of the dimers α, β, ω₁ and ω₂?; (ii) how are these intermediates converted into the final products 2 and 3?; and (iii) what is the origin of the hydridic B–H hydrogen in 3? The working hypothesis was that this hydrogen most likely originates from the (C3)H₂ methylene unit of 1. However, alternative sources, such as the solvent or other C–H fragments of 1, as well as subsequent hydrogen migration processes, could not be excluded.

To address the latter question, a strategy based on C3-deuterium-labelled 1 was devised. Using a total synthetic approach, an isotopic mixture comprising 93.7% C3-di-deuterated 1-d₂, 5.9% C3-mono-deuterated 1-d, and 0.4% non-deuterated 1 (based on NMR analysis) was obtained, consistent with the isotopic composition of the starting materials (Scheme 4; see sections 2.9–2.11 in the SI for details).

Scheme 4 Total synthesis of C3-di-deuterated 1H-2,1-benzazaborole 1-d₂. Deuterium enrichment was determined by ¹H NMR integration. Reagents and conditions: (@) synthesis according to ref. 114 (i) 4-methylpyridine, K₃PO₄, DMSO, D₂O (99.9% D); (ii) N,N-dimethyl-4-nitrosoaniline. Isolated yield is given in parentheses.

With C3-deuterium-labelled 1 in hand, all key mechanistic questions outlined above could, in principle, be addressed. Most importantly, reduction of the isotopologue mixture 1-d_n (n ∈ {0, 1, 2}) with one equivalent of potassium in THF-d₈ afforded, despite the differing isotopic compositions, the same reaction outcome as observed for non-deuterated 1, namely the formation of isotopologues of 2 and 3. The consequences of deuterium incorporation are clearly reflected in the NMR spectra; the corresponding discussion on this matter can be found in section 2.12 of the SI.

For the deuterium-labelled product 2-d_n, containing the aromatic 2,1-Bab⁻ anion, ∼99.6% of the material was C3-deuterated, while a minor fraction of non-deuterated 2 (∼0.4%) was inevitably formed due to the presence of 1-d (5.9%) and 1 (0.4%) in the starting mixture. In addition to the expected mono-C3-deuterated isotopologue 2-d, which predominated (∼81% and ∼85% for batches A and B, respectively), a significant fraction corresponded to the di-deuterated isotopologue 2-d₂ (∼19% and ∼15% for batches A and B, respectively), in which deuterium is incorporated not only at the C3-position but also at para-phenyl carbon (p-Ph-C).

These species were further classified using the “_X_X” notation (X = H/D; Fig. 7), denoting their relevant isotopomers, i.e. 2-d as _D_H and 2-d₂ as _D_D. The partial incorporation of deuterium at the unexpected p-Ph–C position in 2-d_n (n ∈ {0, 1, 2}) (where 2-d₁ ≡ 2-d, while 2-d₀ ≡ 2) represents a key mechanistic insight (vide infra). The predominance of _D_X isotopomers (i.e. both 2-d and 2-d₂) over the non-deuterated isotopologue 2 (_H_H) is evident from the absence of the (C3)H singlet at 6.25 ppm in the ¹H NMR spectra (Fig. S36D and E at page S35; batches A and B), and corroborated by the presence of a broad (C3)D broad signal at 6.28 ppm in the ²H NMR spectrum (reaction performed and recorded separately in non-deuterated THF; batch C; Fig. S36F at page S35). In the ¹³C{¹H} NMR spectrum, the (C3)D group gives rise to a triplet at 96.4 ppm (¹J(¹³C,²H) = 26.5 Hz; Fig. S52).¹¹⁵ Deuteration incorporation at the p-Ph position in the _D_D isotopomer of the 2-d₂ isotopologue is reflected in a reduced intensity of the corresponding (p-Ph)H triplet of triplet at 6.98 ppm in the ¹H NMR spectrum (Fig. S36D and E at page S35; batches A and B), while the corresponding ²H signal of the (p-Ph)D deuteron is broad and of low intensity (Fig. S36F at page S35; batch C).

Fig. 7 Definition of the “_X_X” and “XXXX” notation (X = D/H) used to describe individual isotopomers of the isotopologues 2-d_n (n ∈ {0–2}) and 3-d_n (n ∈ {0–4}), respectively. The “_X_X” notation refers to substitution at the C3-methylene carbon and p-Ph carbon positions, whereas “XXXX” notation corresponds to substitution at the boron, C3-methylene, and p-Ph carbon positions. For simplicity, diastereotopic C3-methylene H/D atoms in XHDX and XDHX for 3-d_n are not distinguished (i.e. XHDX ≡ XDHX).

The situation is even more complex for the second product, the deuterium-labelled potassium borate 3-d_n. The major isotopologues are the C3-di-deuterated hydridoborate 3-d₂ (two deuterium atoms in total) and the deuteridoborate 3-d₃ (three deuterium atoms in total), while smaller amounts of a full series of isotopologues 3-d_n (n ∈ {0, 1, 2, 3, 4}) are also observed (with 3-d₁ ≡ 3-d and 3-d₀ ≡ 3; see Table 1). Notably, the presence of a relatively abundant tetra-deuterated isotopologue 3-d₄ (9.7% in batch B based on LDI-MS analysis) was unexpected and pointed to a more complex mechanistic scenario.

Table 1 List of isotopologues and corresponding isotopomers, including elemental compositions, experimental and theoretical monoisotopic m/z values, and associated mass errors for ions detected in positive- and negative-ion mass spectra of the studied systems. Data are shown for 1 and 1-d_n (n ∈ {0–2}), and for isolated 3 and 3-d_n (n ∈ {0–4}) from batch B, obtained as polycrystalline samples of [3·(18-crown-6)·(THF)₄] and [3-d_n·(18-crown-6)·(THF)₄], respectively, using ultrahigh-resolution LDI-MS. Comparison with NMR data is provided for 3 and 3-d_n (n ∈ {0–4}) analysed directly in the reaction mixture (batch B), which also contains 2 and 2-d_n (n ∈ {0–2}). Relative abundances of isotopologues in the analysed mixtures were determined from the intensities of the corresponding monoisotopic peaks in the full-scan mass spectra

a Contains deuterium only at natural abundance.b Differences between isotopic distributions obtained by LDI-MS and NMR are discussed in the SI.c For simplicity, diastereotopic C3-methylene H/D atoms in XHDX and XDHX for 3-d_n are not distinguished (i.e. XHDX ≡ XDHX). (N.A.) not available by the given analytical technique. The most abundant isotopomer for each isotopologue is highlighted in grey.

---

Instead of three anticipated sites of deuterium incorporation in 3-d_n (C3 and boron), four distinct positions were identified, with the additional site corresponding to the p-Ph–C position. Detailed analysis of ultrahigh-resolution LDI-MS data, supported by NMR studies of isolated [3-d_n·(18-crown-6)·(THF)₄] crystals (see SI, sections 3.3–3.8), revealed that multiple isotopomers can contribute to a given isotopologue 3-d_n (summarized in Table 1). For example, the 36.7% fraction of 3-d₂ consists predominantly of two isotopomers, HDDH (33.0%) and DHDH (3.7%), while the 46.1% fraction of 3-d₃ is mainly composed of DDDH (38.9%) and HDDD (7.2%), using the “XXXX” notation (X = H/D; Fig. 7).

Although the same batch of starting material 1-d_n was used, slight variations in the molar ratio of 2-d_n : 3-d_n were observed across experiments (0.70 : 1 to 0.97 : 1), mirroring the behaviour of non-deuterated 1. These variations also influenced the distribution of individual isotopologues and isotopomers within 3-d_n. For instance, the proportion of p-Ph-C-deuterated isotopomers (XXXD) differed slightly between batches (19% vs. 15% for A and B, respectively), while more pronounced differences were observed for deuteridoborates (DXXX; 32.7% vs. 52.4%, respectively; cf. ¹H and ¹¹B NMR spectra of batches A and B shown in Fig. S36D and E at page S35 and Fig. 8).

Fig. 8 Detailed expansion of the ¹¹B and ¹¹B{¹H} NMR spectra (160.48 MHz, THF-d₈, 21 °C) of different batches of the final reaction mixtures, focused on the signal of borate complex 3-d_n (black traces), showing the distribution of BD (red) and BH (blue) species obtained by signal deconvolution. The proton-coupled ¹¹B spectra were deconvoluted using a doublet splitting of 84.3 Hz. (*) BD is present at natural abundance in this case. Full spectra are provided in the SI.

These discrepancies (in both 2-d_n : 3-d_n molar ratio and D-atom enrichment) likely arise from experimental ambiguities associated with the heterogeneous nature of the reaction. Importantly, however, within a given batch, the fraction of p-Ph deuterated isotopomers is identical for both 2-d_n (_X_D) and 3-d_n (XXXD) (19 vs. 15% for batches A and B, respectively), indicating the presence of a common intermediate in at least one reaction pathway (vide infra).

To enable reliable probabilistic analysis of the reaction mixture—particularly the composition of [3-d_n·(18-crown-6)·(THF)₄] (n ∈ {0–4})—all analytical data were therefore obtained from a single batch. As batch B was selected for this purpose, the following discussion of the reaction mechanism is primarily based on these data.

Probabilistic analysis of deuterium distribution in the final products (batch B)

Regarding the consequences of deuteration within the isotopologues of 3-d_n (n ∈ {0–4}), the weighted average deuterium content in this borate complex was determined to be 2.574 D atoms per molecule for batch B (based on LDI-MS analysis; see Table 1 and SI, sections 3.3–3.6). This value deviates significantly from the three deuterium atoms expected for the idealised tri-deuterated species 3-d₃ (DDDH), assuming a simplified reaction stoichiometry (eqn (1)):

2eq. 1-d₂ + 2eq. K → 2-d (\_D\_H) + 3-d₃ (DDDH) (1)

based on the assumption of only one and three available deuterium incorporation sites in 2-d_n and 3-d_n, respectively. The observed value also differs markedly from the four deuterium atoms corresponding to the tetra-deuterated isotopologue 3-d₄ (DDDD).

This discrepancy can be rationalised by four closely related factors. First, incomplete C3-di-deuteration in the starting material results in an average of only 1.933 D atoms per molecule of 1-d_n (n ∈ {0–2}), rather than ideal value of 2. Second, partial di-deuteration of the product 2-d_n (n ∈ {0–2}) leads to the formation of both 2-d (_D_H) and 2-d₂ (_D_D), which together account for ∼99.6% of all 2-d_n isotopologues. The presence of ∼15% of 2-d₂ (_D_D) effectively reduces the amount of deuterium available for incorporation into 3-d_n by approximately 0.15 D atoms per molecule of 2-d_n. Third, the experimentally observed deviation of the molar ratio 2-d_n : 3-d_n from ideal 1 : 1 value (0.85 : 1.00 in batch B; i.e. normalised ratio 0.459 : 0.541) further contributes to the imbalance. Assuming that one molecule of 1-d_n is required for the formation of each molecule of both 2-d_n and 3-d_n, the resulting discrepancy between the initial deuterium content and that accounted for in the final products is approximately 0.014 D atoms per molecule of 1-d_n. Finally, this small deficit can be attributed to the formation of minor by-products, which remove deuterium from the system, for example as mono-deuterated benzene (C₆H₅D) and other unidentified species formed via benzene elimination (see SI for detailed calculations).

Importantly, the slightly higher deuterium content in the starting material compared to that found in the final products 2-d_n and 3-d_n, indicates that all deuterium atoms present at the “_X_X” sites in 2-d_n and the “XXXX” sites in 3-d_n originate exclusively from 1-d_n, rather than from the deuterated solvent (THF-d₈). This conclusion is supported by experiments performed in non-deuterated THF (batch C), in which the full set of deuterated isotopomers (DXXX, XHDX, XDDX, and XXXD) is still observed (Fig. S36F at page S35).

Structural elucidation of dimeric (deuterated) short-lived intermediates α-d_n, β-d_n, ω₁-d_n and ω₂-d_n (n ∈ {0–4})

In the case of the deuterated system, reduction of 1-d₂ with potassium generates the radical-anionic species K⁺1-d₂˙⁻, which is expected to undergo dimerization to give the tetra-deuterated dimers α-d₄, β-d₄, ω₁-d₄ and ω₂-d₄. However, as the starting material 1-d_n (n ∈ {0–2}) is not fully di-deuterated, the resulting dimers comprise a distribution of isotopologues α-d_n, β-d_n, ω₁-d_n and ω₂-d_n (n ∈ {0–4}), each associated with a larger set of isotopomers dictated by combinatorial statistics. For clarity, the following discussion focuses on the tetra-deuterated representatives (α-d₄, β-d₄, ω₁-d₄ and ω₂-d₄), which are statistically the most abundant.

Remarkably, deuteration enables experimental structural elucidation of these dimers, as their lifetimes are significantly prolonged compared to their non-deuterated counterparts. While α, β, ω₁ and ω₂ are detectable by NMR spectroscopy for only ca. 90 minutes at room temperature, the deuterated species α-d_n, β-d_n, ω₁-d_n and ω₂-d_n remain observable for up to 41 h under identical conditions, with conversion to 2-d_n and 3-d_n reaching only 84%. This pronounced lifetime extension upon C3-di-deuteration indicates the involvement of a primary kinetic isotope effect (KIE) (vide infra). Although quantitative determination of k_H/k_D is precluded by experimental limitations associated with the heterogeneous and tandem nature of the process, the magnitude of the effect (k_H/k_D ≫ 1) clearly supports a primary KIE.

Importantly, C3-di-deuteration does not affect the rate of dimerization of K⁺1-d₂˙⁻ (i.e., the second stage), which remains extremely fast irrespective of isotopic substitution. The observed apparent KIE therefore pertains exclusively to the subsequent transformation of α-d_n, β-d_n, ω₁-d_n and ω₂-d_n into the final products 2-d_n and 3-d_n (the third stage), in which migration of the C3 methylene H/D atom is both the initiating and rate-determining step (vide infra). Notably, the relative molar ratio of the dimers (1 : 0.98 : 1.45 : 1.08 at 21 °C) is identical for both deuterated and non-deuterated systems and remains essentially constant during their conversion to the final products.

The structures of the deuterated intermediates α-d₄, β-d₄, ω₁-d₄ and ω₂-d₄ (Scheme 5), which are equally applicable to the non-deuterated analogues and intermediate isotopologues (Scheme 3), were unambiguously established by 2D NMR homo- and heteronuclear NMR correlation experiments and ¹H–¹H EXSY/NOESY measurements (Fig. S36–S49). These dimers arise from Krause–Gomberg-type radical-anion dimerization of K⁺1-d₂˙⁻ via B(sp³)–C(sp³) coupling.

Scheme 5 Overview of the reduction of C3-di-deuterated 1H-2,1-benzazaborole derivative 1-d_n (n ∈ {0–2}) by potassium metal in THF/THF-d₈, depicting the proposed structures of the short-lived intermediates α-d₄, β-d₄, ω₁-d₄ and ω₂-d₄, and the formation of the final products 2-d_n (n ∈ {0–2}) and 3-d_n (n ∈ {0–4}). (#) Each of the intermediates α-d₄, β-d₄, ω₁-d₄ and ω₂-d₄ is formed as a racemate; thus, a total of eight stereoisomeric intermediates are present. For clarity, only one enantiomer and the most abundant isotopologue (d₄) are shown in each case. (@) Lower isotopologues of each dimer (e.g. α-d₃, α-d₂, α-d and α) are also formed, reflecting the initial isotopic composition of the monomer K⁺1-d_n˙⁻ (n ∈ {0–2}). Furthermore, multiple isotopomers are expected for each isotopologue based on combinatorial considerations (cf. ¹H NMR spectra in Fig. S36A and C at page S35).

In contrast to the classical Krause dimerization of the achiral Ph₃B˙⁻ radical-anion (Scheme 3), which yields essentially a single dimer,¹⁰² the presence of two distinct aromatic coupling sites and the prochiral nature of K⁺1˙⁻ (and K⁺1-d₂˙⁻) leads to the formation of four structurally distinct dimeric dianions. Each structure contains two distinct boron environments, denoted as a tetrahedral (T_d) and a planar boron unit, their respective coordination geometries (Scheme 5).

The exclusive formation of B(sp³)–C(sp³) coupled dimers, involving the pyramidalized boron atom of one monomer and either the C5 position of the dearomatized 2,1-Bab ring (α-d₄ and β-d₄) or the para-position of the dearomatized B-phenyl substituent (ω₁-d₄ and ω₂-d₄) of the second monomer of K⁺1-d₂˙⁻, can be rationalised by two factors: minimization of steric repulsion and localization of SOMO alpha-spin density at these sites (Fig. 6). Formation of B(sp³)–B(sp³) coupled dimers is likely disfavoured by steric congestion arising from the B-phenyl-N-tert-butyl substitution pattern. Similar steric constraints are known to prevent formation of hexaphenylethane in Gomberg-type systems,¹⁰³ except in highly specialised or constrained cases.^116–118

Alternative C–C coupling pathways (e.g., so-called p,p′-dimers)¹¹⁹ are also not observed (see section 3.1 in SI). While such species have been proposed in related systems,¹⁰² their absence here is likely due to thermodynamic destabilization associated with simultaneous dearomatization of two C₆ rings,¹¹⁸ in contrast to the observed dimers, which involve only a single dearomatized ring.

From a thermodynamic perspective, the relative molar ratio of the dimers reflects a balance between the Gibbs free energies of formation and the activation barriers (ΔG^‡) for dimerization of K⁺1˙⁻. DFT calculations (M062X/D3/def2-TZVP (PCM, THF)) identified transition states TS1-α, TS1-β, TS1-ω₁ and TS1-ω₂ (spin multiplicities M = 1; Fig. 9), with ΔG^‡ in the range 7.29–12.15 kcal mol⁻¹, consistent with kinetically accessible processes at room temperature. However, the experimentally observed ratio (1 : 0.98 : 1.45 : 1.08) does not directly correlate with calculated values [α (−16.45) < β (−15.34) < ω₂ (−3.44) < ω₁ (−3.41) kcal mol⁻¹], suggesting the involvement of additional dynamic factors.

Fig. 9 DFT-calculated energy profile (M062X/D3/def2-TZVP (PCM, THF)) for the overall non-linear tandem reaction of 1 upon reduction with potassium metal based on various H-atom transfers (HATs). All transition states (TS), dimeric intermediates, and the final product 3 are formed as pairs of enantiomers; for clarity, only one enantiomer is shown. (#) For the transformation of INT1 into 3 via TS4, see Scheme 8. Cartesian coordinates of all optimised structures are provided in the SI (xyz file).

Indeed, interconversion between dimers—particularly within the ω₁/ω₂ pair—is likely operative. The negligible (0.03 kcal mol⁻¹) and moderate barriers for reverse monomerization (ΔG^‡ ≈ 13.6–15.6 kcal mol⁻¹) indicate facile equilibration between ω₁ and ω₂ via reversible dissociation and recombination. This behaviour is supported by ¹H–¹H EXSY/NOESY experiments at 295 K, which reveal exchange between ω₂ and ω₁ [apparent equilibrium constant K′(ω₂-d₄ ⇌ ω₁-d₄) = 1.34; mixing time 0.9 s; Fig. S40 and S41].

The observed relative molar distribution of dimers in NMR spectra therefore most likely reflects a dynamic equilibrium between the radical-anionic precursor K⁺1˙⁻ and the ensemble of dimeric species (Scheme 6). The absence of an EPR signal for the deep red solution, combined with well-resolved NMR spectra, suggests either an extremely low steady-state concentration of K⁺1˙⁻ or its association into ion-pair clusters, analogous to behaviour reported for M⁺Ph₃B˙⁻ (the Krause’s Adduct monomer; M = alkali metal) in ethereal solvents.^8,120

Scheme 6 Proposed dynamic equilibrium between (non-)deuterated metastable dimers and the corresponding radical-anionic species K⁺1-d_n˙⁻ (n = 0 or 2) in THF.

From the NMR perspective, the characteristic signals observed in the deep red solution arise from the quinoid structures of the dearomatized benzene rings, namely the 2,1-Bab moiety in α and β, and B-phenyl rings in ω₁ and ω₂. These structural motifs give rise to two four-membered and two five-membered spin systems, respectively, which are responsible for the distinctive doublet-like signals in the ¹H NMR spectra in the range of 3.3–6.3 ppm (Fig. 4 and Fig. S36A and C at page S35).

As ω₁ and ω₂ are structurally related to Krause’s Adduct (KA, Scheme 3),¹⁰² the corresponding ¹H NMR patterns of their quinoid rings are in good agreement [cf. p-H: 3.39 (ω₁) and 3.33 (ω₂) vs. 4.03 ppm for KA; m- and m*-H: 4.96–5.08 ppm (ω₁ and ω₂) vs. 5.39 ppm for KA; o- and o*-H: 6.17–6.73 ppm (ω₁ and ω₂) vs. 6.58 ppm for KA]. Due to near-complete di-deuteration at the C3 and C3′ positions, the eight methylene AX spin system systems (corresponding to these positions in the four dimers), which appear for non-deuterated α, β, ω₁ and ω₂ in the range 3.5–4.6 ppm (cf. Fig. S36A vs. C), are almost entirely suppressed in the ¹H NMR spectra of α-d_n, β-d_n, ω₁-d_n and ω₂-d_n (n ∈ {0–4}) (see section 2.12.3 in SI for full assignment).

Notably, NMR spectroscopy enables a surprisingly detailed structural characterisation of the deuterated dimers in solution. In addition to ¹H–¹H NOESY experiments (Fig. S40 and S41), the high sensitivity of ¹⁵N chemical shifts to subtle structural and electronic changes proved particularly informative. Upon B(sp³)–C(sp³) coupling, a boron–carbon double bond is established in the planar boron unit of each dimer, i.e., either C7a = B1 (α-d_n and β-d_n) or i-C=B (ω₁-d_n and ω₂-d_n). As a consequence, the boron p-orbital is no longer fully available for pπ–pπ interaction with nitrogen, leading to partial disruption of the B(sp²)–N(sp²) double-bond character present in the starting material 1-d_n [δ(¹⁵N) = −238.3 ppm].

This redistribution of bonding results in partial pyramidalization of the nitrogen atom in the planar boron unit. Indeed, the ¹H–¹⁵N HMBC spectrum (Fig. S49) reveals, in addition to signals for the final products 2-d_n and 3-d_n at δ(¹⁵N) = −209.5 and −333.3 ppm, respectively, eight additional signals corresponding to the four dimers. These can be grouped into two distinct sets, both within the typical range for pyramidal amines: δ(¹⁵N) = −341.6, −342.0, −342.3, and −341.6 ppm (N′ atoms in the T_d boron units) and −294.2, −294.4, −290.3, and −290.6 ppm (N atoms in the planar boron units) for of α-d_n, β-d_n, ω₁-d_n and ω₂-d_n, respectively.

These values indicate full pyramidalization at the nitrogen atoms of the tetrahedral boron units, whereas the nitrogen atoms in the planar boron units remain only partially pyramidalized, consistent with residual pπ–pπ interaction with the adjacent boron centre. This bonding picture is fully supported by DFT-optimised structures. In all four dimers, the N–B bond length within the planar boron unit (1.473, 1.461, 1.479, and 1.472 Å, respectively) is significantly shorter than that in the tetrahedral unit distance (1.560, 1.562, 1.613, and 1.612 Å, respectively; cf. Σ_rcov(N–B) = 1.56 Å (ref. 94)).

DFT calculations further confirm the quinoid, dearomatized nature of the C₆ rings in both the 2,1-Bab moiety (α and β) and the B-phenyl substituent (ω₁ and ω₂), as evidenced by alternating C–C bond lengths and partial C=B double-bond character. In ω₁ and ω₂, the (i-C)=B bond length is 1.501 Å (cf. Σ_rcov(C–B) = 1.60 Å (ref. 94)), while in α and β, the (C7a)=B1 bond is shortened from 1.564 Å in 1 to 1.489 Å (α) and 1.494 Å (β) (cf. Σ_rcov(C=B) = 1.45 Å (ref. 94)).

Finally, the central B(sp³)–C(sp³) bond, which connects the tetrahedral and planar boron units in all dimers, is only slightly elongated (1.675–1.687 Å) relative to the sum of covalent radii (Σ_rcov(C–B) = 1.60 Å (ref. 94)), consistent with the observed propensity for reversible monomerization.

Conversion of dimeric intermediates α-d_n, β-d_n, ω₁-d_n and ω₂-d_n into 2-d_n and 3-d_n: third stage of the reaction

Under the assumption that the H/D distribution at the boron atom, the C3 carbon, and the p-Ph carbon in the final products 2-d_n (n ∈ {0–2}) and 3-d_n (n ∈ {0–4}) remains invariant, it is, in principle, possible to reconstruct the third stage of the tandem mechanism, i.e., the transformation of the dimers into the final products. This reconstruction is based on experimental data obtained for batch B and on the following seven key observations.

First, starting compound 1-d_n (n ∈ {0–2}) is no longer present at this stage of the reaction, excluding its direct involvement in reactions with the dimeric intermediates α-d_n, β-d_n, ω₁-d_n and ω₂-d_n (n ∈ {0–4}). Second, a primary kinetic effect (KIE) is observed for the conversion of the C3- and C3′-deuterated dimers into 2-d_n and 3-d_n, indicating that H/D transfer from the C3/C3′ methylene positions (formally; vide infra) is both the initiating and rate-determining step of the third stage, and thus of the overall mechanism. Third, from a stoichiometric perspective, 3-d_n contains one H/D atom more than the radical-anionic precursor K⁺1-d_n˙⁻, whereas 2-d_n contains one fewer, meaning that the two products differ by two H/D atoms.

Fourth, the distribution of deuterium at the C3 methylene in 3-d_n, reflected by the relative abundance of XHHX (∼0.6%), XHDX (∼10.6%), and XDDX (∼88.8%) isotopomers (derived from LDI-MS analysis), closely mirrors that of the starting material 1-d_n (0.4% 1, 5.9% 1-d, and 93.7% 1-d₂, based on NMR integration; Table 1), and is thus inherited from the corresponding radical-anionic species K⁺1-d_n˙⁻. Fifth, although the formation of multiple isotopomers of 3-d_n (n ∈ {0–4}) is expected from the isotopic composition of 1-d_n (n ∈ {0–2}), the relatively high proportion of hydridoborates HXXX (47.7% from LDI-MS) compared to deuteridoborates DXXX (52.3%) cannot be explained solely on this basis. Instead, it points to the involvement of multiple competing pathways in the third stage. At least one non-trivial pathway must involve additional H/D migration steps leading to p-Ph deuterated _X_D isotopomer of 2-d_n and XXXD isotopomers of 3-d_n, where each D incorporation at the p-Ph position generates a corresponding H atom, contributing to the observed enrichment of HXXX species.

Sixth, the observation that, in addition to boron, the p-Ph carbon serves as a destination for deuterium in both 2-d_n and 3-d_n suggests the presence of at least two parallel reaction pathways. Seventh, the identical proportion of p-Ph-deuterated _X_D and XXXD isotopomers in both products (15%) strongly indicates a shared intermediate or branching point in the mechanism.

Taken together, these findings support a paramagnetic rather than a diamagnetic reaction pathway for the transformation of dimers into 2-d_n and 3-d_n, despite the diamagnetic nature of both intermediates and products. In this scenario, the dimers undergo reversible monomerization to yield pairs of radical-anionic species K⁺1-d_n˙⁻, which, upon suitable mutual orientation, engage in intermolecular H/D transfer. Specifically, an H/D atom from the C3 methylene group of a donor unit is transferred either to the boron atom B1′ of an acceptor unit, leading to the formation of 3-d_n (more probable pathway), or to the p′-Ph carbon, followed by further H/D migrations (less probable pathway; vide infra).

Hydrogen atom transfer (HAT) generally involves the concerted transfer of a proton and an electron in a single kinetic step.^121,122 While examples of HAT between organic substrates and transition metal centres are known,^123,124 such processes have not previously been reported in connection with Gomberg-type dimers. Moreover, boron-targeted HAT represents a rare class of transformations. This type of process was studied in detail by Roberts et al. in the 1990s in the context of radical scrambling involving neutral amine–boryl radicals generated from amine-boranes.^125–127 Related systems include diradical dianionic boron compounds reported by Wang et al.,¹⁷ radical cationic diazadiborinines described by Kinjo,⁵¹ and diborane radical anions investigated by Furukawa, Lin, and Yamashita.¹²⁸ In most of these cases, however, the mechanistic details of hydrogen transfer were not the primary focus. More recent studies invoking HAT in boron chemistry typically involve boron as an indirect mediator rather than as the acceptor centre itself,^129–131 or describe H-atom abstraction from B–H bonds in NHC–borane complexes.^132–134

To evaluate the viability of the proposed mechanism for the third stage, DFT calculations were performed on the non-deuterated system to identify relevant transition states (TS) and intermediates (INT). A low-energy transition state TS2 (spin multiplicity M = 1, ΔG^‡ = 10.90 kcal mol⁻¹) corresponding to boron-targeted HAT was located (Fig. 9). This pathway provides a direct route to the formation of 2 and 3, although alternative pathways are also accessible (vide infra). In TS2, the H atom is transferred from the C3 carbon atom to the boron atom B1′, which are separated by 3.114 Å. This process is assisted by a potassium cation that bridges the donor unit via coordination to the nitrogen lone pair and the C3 carbon, and the acceptor unit via interaction with π-system of the B1′-phenyl group.

The C3⋯H⋯B1′ arrangement in TS2 is nearly linear (162.0°), closely resembling the geometry reported by Roberts et al. for boron-targeted HAT (>C⋯H⋯B = 159.9°).¹²⁵ Thus, in TS2, the near-linearity of this arrangement appears to be an intrinsic feature of the HAT process, only marginally perturbed by additional interaction between the bridging H atom and the potassium cation. The H atom also interacts with the potassium cation at a distance of 2.628 Å (cf. Σ_rvdW(K⋯H) = 3.85 Å;¹³⁵ Σ_rcov(K–H) = 2.28 Å (ref. 94)). A second potassium cation coordinates to B1′ from the opposite side, aligning with the C3⋯H⋯B1′ axis (>C3⋯B1′⋯K = 157.7°), thereby further stabilising the transition state. The driving force for this “intra-radical-anion-pair boron-targeted HAT” is likely a synergistic combination of aromatization of the five-membered C₃BN ring in the donor unit (leading to 2) and the thermodynamic stabilisation of the borate complex 3, as reflected in the large exergonicity of the process .

Importantly, the structure of TS2 implies that dimers α, β, ω₁ and ω₂ must first undergo reverse monomerization via homolytic cleavage of the C5–B1′ (α and β) or at (p-C)–B1′ bond (ω₁ and ω₂), regenerating two molecules of K⁺1˙⁻ through the corresponding TS1 transition states. These species can subsequently reorient to access TS2, enabling the boron-targeted HAT from C3 to B1′ (Fig. 9). However, because the associated activation barriers are comparable (ΔG^‡ = 10.90 kcal mol⁻¹ for TS2 vs. 7.29–12.15 kcal mol⁻¹ for TS1s), the re-formed radical-anion pair can either proceed toward product formation or recombine to regenerate any of the dimers with similar probability (Scheme 7).

Scheme 7 Diagrams related to Fig. 9 illustrating the transformation of TS1s states of non-deuterated dimers α, β, ω₁ and ω₂ into TS2 via transient formation of two K⁺1˙⁻ species, leading to the formation of final products 2 and 3, alongside the TS3 → TS4 pathway discussed later in this manuscript (A), and the outcome of the TS2 pathway for K⁺1-d₂˙⁻ (B).

This dynamic interplay renders the mechanism inherently non-linear (Fig. 9). In this context, the dimers act as a transient “buffer reservoir” for the highly reactive K⁺1˙⁻ species, prolonging its effective lifetime through rapid reversible interconversion and thereby extending the overall reaction timescale to hours.

When the deuterated precursor 1-d_n (n ∈ {0–2}) is employed, the pathway via TS2 is directly reflected in the isotopologue distribution of the final products. In the dominant case of 1-d₂ (93.7% content by NMR), formation of K⁺1-d₂˙⁻ followed by HAT through TS2 yields _D_H isotopomer of 2-d together with the DDDH isotopomer of 3-d₃ (Scheme 7B). Consistently, 3-d₃ (DDDH) is indeed both the most abundant isotopologue and isotopomer observed (Table 1). Its experimentally observed abundance (38.9% by LDI-MS), however, falls well below the idealized limit, which can be attributed not only to incomplete deuteration of the starting material but, more importantly, to the operation of additional competing pathways alongside TS2 (Scheme 7A; vide infra).

The second proposed HAT pathway, targeting not the boron atom but the carbon atom in the p′-Ph position of the H-abstractor monomer K⁺1˙⁻, was also found to be viable, as indicated by the calculated TS3 (M = 1, ΔG^‡ = 18.71 kcal mol⁻¹; Fig. 9). Similarly to TS2, the TS3 becomes accessible after reorientation of two K⁺1˙⁻ units regenerated from any of the TS1s pathways. HAT from C3 methylene group via TS3 leads to formation of the final product 2 along with intermediate INT1, with (Fig. 9).

The arrangement of the C3⋯H⋯(p′-Ph-C) atoms in TS3 is even closer to linear than in TS2 (>C3⋯H⋯(p′-Ph-C) = 171.0°), and the boundary atoms-distance is shorter (C3⋯(p′-Ph-C) = 2.883 Å in TS3 vs. C3⋯B1′ = 3.114 Å in TS2). Unlike in TS2, no potassium cation is bridging H-donor and H-abstractor molecule of the monomer K⁺1˙⁻ in the TS3, yet one potassium cation bridges the H-donor and H-abstractor units; however, one potassium cation η⁶-coordinates to the B1′-phenyl ring from the opposite side of the H-atom acceptor, aligned along the axis of the nearly linear C3⋯H⋯(p′-Ph-C) arrangement (>C3⋯(p′-Ph-C)⋯K = 168.0°).

Upon HAT to the (p′-Ph)CH fragment, the phenyl ring in the INT1 acquires a cross-conjugated diene character as a (p-Ph)CH₂ methylene bridge is formed, while the boron atom becomes negatively charged due to π-interaction with the ipso-phenyl carbon, giving a trigonal planar borate species. As INT1 is thermodynamically less stable than the final tetrahedral borate complex 3 by 18.10 kcal mol⁻¹, yet represents its positional isomer, it is subsequently converted into 3 via intermolecular, self-propagating cycle involving TS4 (Scheme 8A).

Scheme 8 DFT-calculated energy profile (M062X/D3/def2-TZVP (PCM, THF)) for the isomerization of INT1 to 3 via an intermolecular self-propagating, boron-targeted HAT cycle, leading to propagation of K⁺1˙⁻ (colour change from black to magenta) (A). Corresponding scheme for INT1-d₃, yielding hydridoborate 3-d₂ (HDDH) and propagating tri-deuterated K⁺1-d₃˙⁻, assuming 1-d₂ as the starting compound (B). Plausible pathways rationalising formation of 3-d₄ (DDDD) and 2-d₂ (_D_D) via reversal of the boron-targeted D-atom transfer (DAT) through TS2 (C, right).

This step corresponds to a boron-targeted HAT from the (p-Ph)CH₂ methylene group of INT1 to the boron atom of another K⁺1˙⁻ unit (ΔG^‡ = 19.44 kcal mol⁻¹). In this process, INT1 reacts with K⁺1˙⁻ to regenerate a new K⁺1˙⁻ species, effectively propagating the reactive intermediate pool and further contributing to the non-linear character of the overall mechanism. In TS4, the key geometric parameters closely resemble those in TS2; however, in this case both potassium cations are positioned along the extension of the (p-Ph-C)⋯H⋯B1′ axis.

The isomerization of INT1 via TS4 is crucial for rationalizing the formation of p-Ph-deuterated isotopomer, namely _D_D in 2-d₂ and XXXD in 3-d_n, when deuterated starting compound material 1-d_n is employed (Table 1). In the simplified case of 1-d₂ leading to K⁺1-d₂˙⁻, the TS3 pathway generates a (p-Ph)CH,D methylene unit in tri-deuterated INT1-d₃ (Scheme 8B). During subsequent conversion via TS4 in the presence of K⁺1-d₂˙⁻, primary KIE governs the intermolecular competition, favouring transfer of the weaker C–H bond over C–D. This results in formation of the key p-Ph-deuterated radical-anion K⁺1-d₃˙⁻ along with the HDDH isotopomer of 3-d₂.

The formation of K⁺1-d₃˙⁻ in turn explains the origin of the _D_D isotopomer of 2-d₂, as this species can react with K⁺1-d₂˙⁻ via the energetically preferred TS2 pathway, yielding 2-d₂ (_D_D) together with 3-d₃ (DDDH) (Scheme 8C). Conversely, reversing the direction of deuterium transfer through TS2 also accounts for the formation of the tetra-deuterated isotopologue 3-d₄ (DDDD), with 2-d (_D_H) as the corresponding co-product.

Because the p-Ph-deuterated isotopomers of both products (i.e., _X_D in 2-d_n and XXXD in 3-d_n) are formed in identical proportions (15% in batch B), the directionality of deuterium transfer between tri-deuterated K⁺1-d₃˙⁻ and di-deuterated K⁺1-d₂˙⁻ must be statistically equivalent. The validity of the TS3 → TS4 → TS2 sequence is further supported by the experimental observation that these p-Ph-deuterated species indeed constitute a minor fraction (∼15%) of the product distribution.

Consistently, 3-d₂ is the second most abundant isotopologue (36.7% by LDI-MS), and its HDDH isotopomer is likewise the second most abundant individual isotopomer (33.0%) within the entire 3-d_n (n ∈ {0–4}) distribution (Table 1). This agrees with the dominance of the TS2 pathway (ΔG^‡ = 10.90 kcal mol⁻¹), which directly yields 3-d₃ (DDDH) without requiring additional steps, whereas formation via TS3 necessitates subsequent transformation through higher-energy pathways.

A more detailed discussion of minor isotopomers, as well as deviations in hydridoborate/deuteridoborate ratios, non-ideal 2-d_n : 3-d_n stoichiometry, and formation of benzene/mono-deuterated benzene as a negligible side product, is provided in the SI (sections 3.6–3.10).

Although the preceding discussion emphasized the role of dimeric intermediates α, β, ω₁ and ω₂, the DFT energy profile (Fig. 9) indicates that their involvement is not essential. Two molecules of K⁺1˙⁻ formed after the first stage can undergo HAT via TS2 or TS3 directly, without prior dimerization (TS1s) and subsequent monomerization. This is enabled by a shared region on the potential energy surface, where TS1s, TS2 and TS3 originate from the same pair of K⁺1˙⁻ pair and exhibit comparable activation barriers.

Experimental observations support this scenario. A relatively high degree of conversion to 2 and 3 is observed immediately after complete consumption of potassium, even at low temperature, when the concentration of dimers is maximal. This behaviour is inconsistent with the relatively slow rate of dimer transformation (third stage), particularly evident in deuterated systems (Fig. S36C and section 3.2 in the SI).

Moreover, the relative ratio of α, β, ω₁ and ω₂ (1.00 : 0.98 : 1.45 : 1.08) remains essentially constant during the reaction as it only changes to 1.00 : 0.93 : 1.57 : 1.01 after 41 h (for batch A), indicating similar apparent rate constants and comparable access of all dimers to TS2 and TS3 (followed by TS4).

Finally, the key factor governing all three HAT processes (TS2–TS4) is the distribution of SOMO alpha-spin density in K⁺1˙⁻ (Fig. 6). The highest contribution is located at the boron atom (relevant for TS2 and TS4), while the second highest resides at the p-Ph carbon (relevant for TS3). Notably, the absence of significant spin density at the C3 methylene position (the H-donor) of K⁺1˙⁻, or at the (p-Ph)CH₂ unit of diamagnetic INT1, does not preclude HAT, as such localisation is not a general requirement for this mechanism.¹²¹

Conclusions

We have carried out a comprehensive mechanistic investigation of the anomalous reduction of 1-Ph-2-tert-butyl-1H-1,2-benzazaborole (1) with potassium in THF. Although 1 is formally isoelectronic with 1H-indene derivatives and furnishes isoelectronic products upon reduction, we demonstrate that the underlying reaction pathways are fundamentally different. Whereas substituted 1H-indenes follow a self-protonation pathway, compound 1 undergoes a radical disproportionation process mediated mostly by a rare, boron-targeted hydrogen atom transfer (HAT), a rarely documented elementary step and one that is previously unrecognized in such systems. This process yields the known aromatic 2,1-Bab⁻ salt^90,91 alongside a previously unknown hydridoborate complex 3.

Mechanistically, the reaction proceeds through an unusual non-linear tandem network. Initial single-electron transfer generates the elusive radical anion K⁺1˙⁻, which rapidly dimerizes into four Gomberg-type (or Krause-Adduct-type) intermediates α, β, ω₁ and ω₂. These dimers are not productive endpoints but metastable reservoirs of K⁺1˙⁻. Their reversible monomerization regenerates radical-anion pairs that undergo selective HAT from the C3 methylene position to either boron (dominant pathway) or the para-phenyl carbon (minor pathway). The latter proceeds via intermediate INT1 and a subsequent intermolecular HAT isomerization cycle. The overall transformation is strongly exergonic, driven by aromatization of the C₃BN ring and formation of a thermodynamically stable borate complex. Reversibility, rapid reorientation, and partial propagation of K⁺1˙⁻ collectively render the mechanism intrinsically non-linear.

Deuterium-labelling experiments were decisive, enabling the experimental elucidation of the structures of the dimeric intermediates α, β, ω₁ and ω₂ and confirming the C3 methylene group as the main source of the hydrogen atom delivered to boron via HAT in the formation of 3. Critically, they also uncovered a second, otherwise hidden HAT pathway targeting the para-phenyl position, which would remain obscured in the absence of isotopic labelling due to hydrogen scrambling. The observed selectivity is consistent with the computed spin-density distribution in K⁺1˙⁻, with HAT occurring preferentially at sites of highest spin density.

Finally, this work establishes that HAT processes can operate within Gomberg-type dimers and Krause-like adducts, a reactivity mode not previously recognized. These findings redefine the mechanistic landscape of such systems and open new perspectives for the rational design of unconventional boron hydrides and related radical processes.

Author contributions

M. H. conceived and supervised the project, designed and performed the synthetic experiments, characterised the compounds, performed the probabilistic calculations, proposed the mechanism, and wrote the manuscript. L. D. acquired funding, contributed to project supervision, and revised the manuscript. O. M. performed the initial mechanistic DFT and TD-DFT calculations. A. R. acquired funding and collected and solved the sc-XRD data. A. L. measured and analysed the ²H NMR data. T. M. performed and analysed the electrochemical measurements. R. J. performed and analysed the LDI-MS measurements. M. A. S. designed and carried out the computational studies, contributed key mechanistic insight, and revised the manuscript. All authors have approved the final version of the manuscript.

Conflicts of interest

There are no conflicts to declare.

Data availability

The authors confirm that the data supporting the findings of this study are available within the article and its supplementary information (SI). Supplementary information includes: Experimental details, synthetic procedures, NMR spectra and their detailed assignments, UV–Vis spectra, mechanistic investigations, kinetic data, LDI-MS data with assignments of fragment-ion structures, mathematical analysis of deuterium distribution in the final products, details of single-crystal X-ray diffraction analyses and details of DFT calculations (pdf.). Optimized geometries of all calculated structures (xyz.). See DOI: https://doi.org/10.1039/d6dt01280a.

The raw data supporting this study are available from the corresponding author upon request, as no institutional data repository is currently available at our university.

CCDC 2043448 contains the supplementary crystallographic data for this paper.¹³⁶

Acknowledgements

The Grant Agency of the Czech Republic (project no. GA21-02964S) is acknowledged for the financial support.

References

1. A. Fukazawa, J. L. Dutton, C. Fan, L. G. Mercier, A. Y. Houghton, Q. Wu, W. E. Piers and M. Parvez, Chem. Sci., 2012, 3, 1814–1818.
2. J. He, F. Rauch, M. Finze and T. B. Marder, Chem. Sci., 2021, 12, 128–147.
3. C. Präsang, M. Hofmann, G. Geiseler, W. Massa and A. Berndt, Angew. Chem., Int. Ed., 2002, 41, 1526–1529.
4. T. Kupfer, H. Braunschweig and K. Radacki, Angew. Chem., Int. Ed., 2015, 54, 15084–15088.
5. H. Braunschweig, V. Dyakonov, J. O. C. Jimenez-Halla, K. Kraft, I. Krummenacher, K. Radacki, A. Sperlich and J. Wahler, Angew. Chem., Int. Ed., 2012, 51, 2977–2980.
6. W. E. Piers, S. C. Bourke and K. D. Conroy, Angew. Chem., Int. Ed., 2005, 44, 5016–5036.
7. W. Kaim, N. S. Hosmane, S. Záliš, J. A. Maguire and W. N. Lipscomb, Angew. Chem., Int. Ed., 2009, 48, 5082–5091.
8. Y. Su and R. Kinjo, Coord. Chem. Rev., 2017, 352, 346–378.
9. H. Budy, J. Gilmer, T. Trageser and M. Wagner, Eur. J. Inorg. Chem., 2020, 2020, 4148–4162.
10. R. Bertermann, H. Braunschweig, R. D. Dewhurst, C. Hörl, T. Kramer and I. Krummenacher, Angew. Chem., Int. Ed., 2014, 53, 5453–5457.
11. H. Braunschweig, I. Krummenacher, L. Mailänder, L. Pentecost and A. Vargas, Chem. Commun., 2016, 52, 7005–7008.
12. W. Yang, K. E. Krantz, L. A. Freeman, D. A. Dickie, A. Molino, G. Frenking, S. Pan, D. J. D. Wilson and R. J. Gilliard, Jr., Angew. Chem., Int. Ed., 2020, 59, 3850–3854.
13. R. Feng, W. Yang, W. Wang, Y. Zhao, G. Tan, L. Zhang and X. Wang, Chem. Commun., 2019, 55, 12908–12911.
14. X. Jia, J. Nitsch, Z. Wu, A. Friedrich, J. Krebs, I. Krummenacher, F. Fantuzzi, H. Braunschweig, M. Moos, C. Lambert, B. Engels and T. B. Marder, Chem. Sci., 2021, 12, 11864–11872.
15. R. Kinjo, B. Donnadieu, M. A. Celik, G. Frenking and G. Bertrand, Science, 2011, 333, 610–613.
16. P. Bissinger, H. Braunschweig, A. Damme, C. Hörl, I. Krummenacher and T. Kupfer, Angew. Chem., Int. Ed., 2015, 54, 359–362.
17. L. Wang, Y. Fang, H. Mao, Y. Qu, J. Zuo, Z. Zhang, G. Tan and X. Wang, Chem. – Eur. J., 2017, 23, 6930–6936.
18. A. Maiti, F. Zhang, I. Krummenacher, M. Bhattacharyya, S. Mehta, M. Moos, C. Lambert, B. Engels, A. Mondal, H. Braunschweig, P. Ravat and A. Jana, J. Am. Chem. Soc., 2021, 143, 3687–3692.
19. K. K. Hollister, W. Yang, R. Mondol, K. E. Wentz, A. Molino, A. Kaur, D. A. Dickie, G. Frenking, S. Pan, D. J. D. Wilson and R. J. Gilliard, Jr., Angew. Chem., Int. Ed., 2022, 61, e202202516.
20. Y.-J. Lin, W.-C. Liu, Y.-H. Liu, G.-H. Lee, S.-Y. Chien and C.-W. Chiu, Nat. Commun., 2022, 13, 7051.
21. B. Y. E. Tra, Y. Zhao, A. Molino, Y. Ouyang, C.-L. Deng, C. McAloon, N. D. McMillion, H. Kim, C. Zhang, P. Müller, R. G. Griffin and R. J. Gilliard, Jr., J. Am. Chem. Soc., 2025, 147, 18431–18437.
22. S. Li, F. Shiri, G. Xu, S.-M. Yiu, H. K. Lee, T. H. Ng, Z. Lin and Z. Lu, J. Am. Chem. Soc., 2024, 146, 17348–17354.
23. J. Gilmer, H. Budy, T. Kaese, M. Bolte, H.-W. Lerner and M. Wagner, Angew. Chem., Int. Ed., 2020, 59, 5621–5625.
24. J. Monot, A. Solovyev, H. Bonin-Dubarle, É. Derat, D. P. Curran, M. Robert, L. Fensterbank, M. Malacria and E. Lacôte, Angew. Chem., Int. Ed., 2010, 49, 9166–9169.
25. Y. Segawa, M. Yamashita and K. Nozaki, Science, 2006, 314, 113–115.
26. Y. Segawa, Y. Suzuki, M. Yamashita and K. Nozaki, J. Am. Chem. Soc., 2008, 130, 16069–16079.
27. E. Bernhardt, V. Bernhardt-Pitchougina, H. Willner and N. Ignatiev, Angew. Chem., Int. Ed., 2011, 50, 12085–12088.
28. J. Landmann, J. A. P. Sprenger, R. Bertermann, N. Ignat'ev, V. Bernhardt-Pitchougina, E. Bernhardt, H. Willner and M. Finze, Chem. Commun., 2015, 51, 4989–4992.
29. J. Landmann, F. Keppner, D. B. Hofmann, J. A. P. Sprenger, M. Häring, S. H. Zottnick, K. Müller-Buschbaum, N. V. Ignat'ev and M. Finze, Angew. Chem., Int. Ed., 2017, 56, 2795–2799.
30. M. J. S. Dewar, V. P. Kubba and R. Pettit, J. Chem. Soc., 1958, 3073–3076,  10.1039/JR9580003073.
31. M. J. D. Bosdet and W. E. Piers, Can. J. Chem., 2009, 87, 8–29.
32. P. G. Campbell, A. J. V. Marwitz and S.-Y. Liu, Angew. Chem., Int. Ed., 2012, 51, 6074–6092.
33. G. Bélanger-Chabot, H. Braunschweig and D. K. Roy, Eur. J. Inorg. Chem., 2017, 2017, 4353–4368.
34. A. J. V. Marwitz, M. H. Matus, L. N. Zakharov, D. A. Dixon and S.-Y. Liu, Angew. Chem., Int. Ed., 2009, 48, 973–977.
35. M. M. Morgan, E. A. Patrick, J. M. Rautiainen, H. M. Tuononen, W. E. Piers and D. M. Spasyuk, Organometallics, 2017, 36, 2541–2551.
36. Y. Zhang, W. Li, R. Jiang, L. Zhang, Y. Li, X. Xu and X. Liu, J. Org. Chem., 2022, 87, 12986–12996.
37. M. J. D. Bosdet, W. E. Piers, T. S. Sorensen and M. Parvez, Angew. Chem., Int. Ed., 2007, 46, 4940–4943.
38. J. Li, C. G. Daniliuc, G. Kehr and G. Erker, Angew. Chem., Int. Ed., 2021, 60, 27053–27061.
39. M. Hejda, A. Lyčka, R. Jambor, A. Růžička and L. Dostál, Dalton Trans., 2014, 43, 12678–12688.
40. H. V. Ly, T. D. Forster, A. M. Corrente, D. J. Eisler, J. Konu, M. Parvez and R. Roesler, Organometallics, 2007, 26, 1750–1756.
41. H. V. Ly, T. D. Forster, D. Maley, M. Parvez and R. Roesler, Chem. Commun., 2005, 4468–4470,  10.1039/B508152A.
42. K. E. Krahulic, G. D. Enright, M. Parvez and R. Roesler, J. Am. Chem. Soc., 2005, 127, 4142–4143.
43. H. V. Ly, H. M. Tuononen, M. Parvez and R. Roesler, Angew. Chem., Int. Ed., 2008, 47, 361–364.
44. H. V. Ly, J. Moilanen, H. M. Tuononen, M. Parvez and R. Roesler, Chem. Commun., 2011, 47, 8391–8393.
45. D. Wu, L. Kong, Y. Li, R. Ganguly and R. Kinjo, Nat. Commun., 2015, 6, 7340.
46. Y. Su, Y. Li, R. Ganguly and R. Kinjo, Angew. Chem., Int. Ed., 2018, 57, 7846–7849.
47. Y. Su, D. C. Huan Do, Y. Li and R. Kinjo, J. Am. Chem. Soc., 2019, 141, 13729–13733.
48. H. V. Ly, J. Konu, M. Parvez and R. Roesler, Dalton Trans., 2008, 3454–3460,  10.1039/B802930J.
49. Y. Mu, Y. Dai, D. A. Ruiz, L. L. Liu, L.-P. Xu, C.-H. Tung and L. Kong, Angew. Chem., Int. Ed., 2024, 63, e202405905.
50. Y. Su, Y. Li, R. Ganguly and R. Kinjo, Chem. Sci., 2017, 8, 7419–7423.
51. B. Wang, Y. Li, R. Ganguly, R. D. Webster and R. Kinjo, Angew. Chem., Int. Ed., 2018, 57, 7826–7829.
52. H. Wang, J. Zhang, Z. Lin and Z. Xie, Organometallics, 2016, 35, 2579–2582.
53. Y. Aramaki, H. Omiya, M. Yamashita, K. Nakabayashi, S.-i. Ohkoshi and K. Nozaki, J. Am. Chem. Soc., 2012, 134, 19989–19992.
54. J. Seufert, E. Welz, I. Krummenacher, V. Paprocki, J. Böhnke, S. Hagspiel, R. D. Dewhurst, R. Tacke, B. Engels and H. Braunschweig, Angew. Chem., Int. Ed., 2018, 57, 10752–10755.
55. N. Yuan, T. Zhu, J. Ma, Y. Xue and Y. Liu, Dalton Trans., 2025, 54, 2257–2260.
56. L. Zhu, Z. Feng and R. Kinjo, J. Am. Chem. Soc., 2024, 146, 20945–20950.
57. G. Farnia, G. Sandoná, F. Marcuzzi and G. Melloni, J. Electroanal. Chem. Interfacial Electrochem., 1989, 264, 297–300.
58. G. Capobianco, G. Farnia, G. Sandona, F. Marcum and G. Melloni, J. Electroanal. Chem. Interfacial Electrochem., 1982, 134, 363–367.
59. G. Farnia, F. Marcuzzi, G. Melloni and G. Sandona, J. Am. Chem. Soc., 1984, 106, 6503–6512.
60. G. Farnia, G. Sandonà, F. Marcuzzi and G. Melloni, J. Chem. Soc., Perkin Trans. 2, 1988, 247–254,  10.1039/P29880000247.
61. S. Gotta, M. Marchetti, M. Branca and G. Melloni, J. Chem. Soc., Perkin Trans. 2, 1994, 151–155,  10.1039/P29940000151.
62. G. Farnia, F. Marcuzzi, G. Melloni, G. Sandona and M. V. Zucca, J. Am. Chem. Soc., 1989, 111, 918–923.
63. A. Bosch and R. K. Brown, Can. J. Chem., 1964, 42, 1718–1735.
64. G. Farnia, J. Ludvík, G. Sandonà and M. G. Severin, J. Chem. Soc., Perkin Trans. 2, 1991, 1249–1254,  10.1039/P29910001249.
65. A. J. Ashe, in Comprehensive Heterocyclic Chemistry III, ed. C. A. R. A. R. Katritzky, E. F. V. Scriven and R. J. K. Taylor, Elsevier, 2008, pp. 1189–1224,  DOI:10.1016/B978-008044992-0.00417-X.
66. G. Schmid, in Comprehensive Heterocyclic Chemistry II, ed. A. R. Katritzky, C. W. Rees and E. F. V. Scriven, Pergamon, Oxford, 1996, pp. 739–766,  DOI:10.1016/B978-008096518-5.00075-7.
67. G. Schmid, D. Zaika, J. Lehr, N. Augart and R. Boese, Chem. Ber., 1988, 121, 1873–1880.
68. G. Schmid, W. Meyer-Zaika, R. Boese and N. Augart, Angew. Chem., Int. Ed. Engl., 1988, 27, 952–953.
69. A. J. Ashe and X. Fang, Org. Lett., 2000, 2, 2089–2091.
70. S.-Y. Liu, I. D. Hills and G. C. Fu, J. Am. Chem. Soc., 2005, 127, 15352–15353.
71. Z. Liu, J. Xu, W. Ruan, C. Fu, H.-J. Zhang and T.-B. Wen, Dalton Trans., 2013, 42, 11976–11980.
72. X. Fang and J. Assoud, Organometallics, 2008, 27, 2408–2410.
73. Y. Chen, X. Fang and W. Dan, Organometallics, 2016, 35, 15–19.
74. X. Fang, Y. Deng, Q. Xie and F. Moingeon, Organometallics, 2008, 27, 2892–2895.
75. X. Fang, X. Li, Z. Hou, J. Assoud and R. Zhao, Organometallics, 2009, 28, 517–522.
76. L. Xie, J. Zhang and C. Cui, Chem. – Eur. J., 2014, 20, 9500–9503.
77. S.-Y. Liu, M. M. C. Lo and G. C. Fu, Tetrahedron, 2006, 62, 11343–11349.
78. A. J. Ashe, H. Yang, X. Fang and J. W. Kampf, Organometallics, 2002, 21, 4578–4580.
79. M. M. Morgan, J. M. Rautiainen, W. E. Piers, H. M. Tuononen and C. Gendy, Dalton Trans., 2018, 47, 734–741.
80. M. Yamashita, Y. Aramaki and K. Nozaki, New J. Chem., 2010, 34, 1774–1782.
81. C. P. Manankandayalage, D. K. Unruh and C. Krempner, Chem. Commun., 2021, 57, 12528–12531.
82. M. Beguerie, C. Faradji, L. Vendier, S. Sabo-Etienne and G. Alcaraz, ChemCatChem, 2017, 9, 3303–3306.
83. M. Beguerie, C. Dinoi, I. del Rosal, C. Faradji, G. Alcaraz, L. Vendier and S. Sabo-Etienne, ACS Catal., 2018, 8, 939–948.
84. H. E. Dunn, J. C. Catlin and H. R. Snyder, J. Org. Chem., 1968, 33, 4483–4486.
85. M. Lauer and G. Wulff, J. Organomet. Chem., 1983, 256, 1–9.
86. A. M. Genaev, S. M. Nagy, G. E. Salnikov and V. G. Shubin, Chem. Commun., 2000, 1587–1588,  10.1039/B003999N.
87. A. Rydzewska, K. Ślepokura, T. Lis, P. Kafarski and P. Młynarz, Tetrahedron Lett., 2009, 50, 132–134.
88. R. Zhang, Y. D. Zhang, L. X. Wang, C. H. Ge, Z. Y. Ma, J. P. Miao and X. D. Zhang, Inorg. Chem. Commun., 2016, 74, 52–57.
89. M. Hejda, A. Lyčka, R. Jambor, A. Růžička and L. Dostál, Dalton Trans., 2013, 42, 6417–6428.
90. M. Hejda, R. Jambor, A. Růžička, A. Lyčka and L. Dostál, Dalton Trans., 2014, 43, 9012–9015.
91. M. Hejda, A. Lyčka, T. Mikysek, R. Jambor, A. Růžička, J. Vinklárek, C. Wilfer, A. Hoffmann, S. Herres-Pawlis and L. Dostál, Chem. – Eur. J., 2016, 22, 15340–15349.
92. G. Farnia, F. Marcuzzi and G. Sandonà, J. Electroanal. Chem., 1999, 460, 160–175.
93. C. Amatore, G. Capobianco, G. Farnia, G. Sandona, J. M. Saveant, M. G. Severin and E. Vianello, J. Am. Chem. Soc., 1985, 107, 1815–1824.
94. P. Pyykkö and M. Atsumi, Chem. – Eur. J., 2009, 15, 186–197.
95. R. L. Luck and E. J. Schelter, Acta Crystallogr., Sect. C:Cryst. Struct. Commun., 1999, 55, IUC9900151.
96. V. D’Anna, L. M. Lawson-Daku and H. Hagemann, J. Alloys Compd., 2013, 580, S122–S124.
97. At −80 °C, the deep red reaction mixture remains stable for at least 24 h, whereas at 50 °C no intermediate colour changes are observed and only the final mixture of 2 and 3 is formed. The reaction also proceeds solvent-free upon melting neat 1 with potassium (see the SI).
98. The reaction is severely disrupted when a PTFE-coated stirring bar is used, leading to formation of a black insoluble residue (likely from PTFE reduction) and a complex mixture of decomposition products. A glass-coated stirring bar or ultrasonic agitation is therefore required. As 2,1-Bab⁻ reacts only slowly with PTFE and [3·(18-crown-6)·(THF)₄] is unreactive, these observations point to the involvement of highly reactive, short-lived intermediates.
99. The reported relative abundances were obtained by averaging multiple integrals for each species.
100. E. Krause, Ber. Dtsch. Chem. Ges., 1924, 57, 216–217.
101. E. Krause and H. Polack, Ber. Dtsch. Chem. Ges., 1926, 59, 777–785.
102. J. J. Eisch, T. Dluzniewski and M. Behrooz, Heteroat. Chem., 1993, 4, 235–241.
103. H. Lankamp, W. T. Nauta and C. MacLean, Tetrahedron Lett., 1968, 9, 249–254.
104. M. Gomberg, J. Am. Chem. Soc., 1900, 22, 757–771.
105. M. Gomberg, J. Am. Chem. Soc., 1901, 23, 496–502.
106. M. Gomberg, J. Am. Chem. Soc., 1914, 36, 1144–1170.
107. M. Gomberg, J. Am. Chem. Soc., 1902, 24, 597–628.
108. A. J. Rosenthal, M. Devillard, K. Miqueu, G. Bouhadir and D. Bourissou, Angew. Chem., Int. Ed., 2015, 54, 9198–9202.
109. A. J. Rosenthal, S. Mallet-Ladeira, G. Bouhadir, E.-D. Sosa-Carrizo, K. Miqueu and D. Bourissou, Organometallics, 2018, 37, 755–760.
110. Our experimental setup did not allow direct generation of the radical anion within the EPR spectrometer. Attempts to generate 1˙⁻ in situ and rely on its diffusion into the resonator cavity were unsuccessful, most likely due to its extremely rapid dimerization.
111. The redox potentials of alkali metal amalgams are approximately 1 V less negative than those of the corresponding pure alkali metals (E⁰ (M⁺/M⁰) ≈ −3 V), as reported in the literature cited below.
112. J. Balej, Electrochim. Acta, 1976, 21, 953–956.
113. M. M. Olmstead and P. P. Power, J. Am. Chem. Soc., 1986, 108, 4235–4236.
114. X. Li, S. Wu, S. Chen, Z. Lai, H.-B. Luo and C. Sheng, Org. Lett., 2018, 20, 1712–1715.
115. The theoretically expected value of ¹J(¹³C,¹H) = 172.6 Hz for methine (C3)H group in 2, based on the gyromagnetic ratio γ(²H)/γ(¹H) = 26.7519/4.1066, was indeed observed as a doublet satellite in the ¹H NMR spectrum.
116. M. Stein, W. Winter and A. Rieker, Angew. Chem., Int. Ed. Engl., 1978, 17, 692–694.
117. S. Grimme and P. R. Schreiner, Angew. Chem., Int. Ed., 2011, 50, 12639–12642.
118. T. Takeda, Y. Uchimura, H. Kawai, R. Katoono, K. Fujiwara and T. Suzuki, Chem. Lett., 2013, 42, 954–962.
119. Y. Uchimura, T. Takeda, R. Katoono, K. Fujiwara and T. Suzuki, Angew. Chem., Int. Ed., 2015, 54, 4010–4013.
120. T. L. Chu, J. Am. Chem. Soc., 1953, 75, 1730–1732.
121. J. M. Mayer, Acc. Chem. Res., 2011, 44, 36–46.
122. J. J. Warren, T. A. Tronic and J. M. Mayer, Chem. Rev., 2010, 110, 6961–7001.
123. B. de Bruin, W. I. Dzik, S. Li and B. B. Wayland, Chem. – Eur. J., 2009, 15, 4312–4320.
124. D. C. Eisenberg, C. J. C. Lawrie, A. E. Moody and J. R. Norton, J. Am. Chem. Soc., 1991, 113, 4888–4895.
125. H.-S. Dang, V. Diart, B. P. Roberts and D. A. Tocher, J. Chem. Soc., Perkin Trans. 2, 1994, 1039–1045,  10.1039/P29940001039.
126. I. G. Green and B. P. Roberts, J. Chem. Soc., Perkin Trans. 2, 1986, 1597–1606,  10.1039/P29860001597.
127. J. N. Kirwan and B. P. Roberts, J. Chem. Soc., Chem. Commun., 1988, 480–482,  10.1039/C39880000480.
128. H. Asakawa, K.-H. Lee, K. Furukawa, Z. Lin and M. Yamashita, Chem. – Eur. J., 2015, 21, 4267–4271.
129. D. Wang, C. Mück-Lichtenfeld and A. Studer, J. Am. Chem. Soc., 2019, 141, 14126–14130.
130. D. Wang, K. Jana and A. Studer, Org. Lett., 2021, 23, 5876–5879.
131. S. Shi, F. Salahi, H. B. Vibbert, M. Rahman, S. A. Snyder and J. R. Norton, Angew. Chem., Int. Ed., 2021, 60, 22678–22682.
132. S.-H. Ueng, A. Solovyev, X. Yuan, S. J. Geib, L. Fensterbank, E. Lacôte, M. Malacria, M. Newcomb, J. C. Walton and D. P. Curran, J. Am. Chem. Soc., 2009, 131, 11256–11262.
133. J. C. Walton, M. M. Brahmi, L. Fensterbank, E. Lacôte, M. Malacria, Q. Chu, S.-H. Ueng, A. Solovyev and D. P. Curran, J. Am. Chem. Soc., 2010, 132, 2350–2358.
134. J. C. Walton, Angew. Chem., Int. Ed., 2009, 48, 1726–1728.
135. M. Mantina, A. C. Chamberlin, R. Valero, C. J. Cramer and D. G. Truhlar, J. Phys. Chem. A, 2009, 113, 5806–5812.
136. CCDC 2043448: Experimental Crystal Structure Determination, 2026,  DOI:10.5517/ccdc.csd.cc26lcpw.

---