Main

The stability of the pnictogen–oxygen bond in phosphine oxides has been used for over a century to drive chemical reactions such as those discovered by Wittig1, Mitsunobu2, Appel3 and Staudinger4. The electronic structure that gives rise to the stability of the P=O/P+–O− bond was once a topic of intense debate, but the currently accepted model features a single covalent bond between the P and O atoms strengthened by electrostatic attraction between the P+ and O− centres as well as donation from O-centred lone pairs into P–C antibonding orbitals5. As the Group 15 element increases in atomic number, however, the pnictogen valence orbitals become more diffuse, overlap with O-based orbitals decreases, and the pnictogen atom becomes increasingly able to expand its coordination sphere. These trends suggest that the heavier congeners of phosphine oxides could exhibit distinct and interesting reactivity6. The behaviour and properties of these heavier congeners would also provide a means of validating the bonding model currently used to describe the Pn=O/Pn+–O− bond, where Pn is a pnictogen5. For As, the variations from P are small enough, possibly as a result of the scandide contraction7, that arsine oxides are generally analogous to phosphine oxides: they are monomeric species with As=O/As+–O− polar covalent bonds. For example, oxidation of either Ph3P or Ph3As with H2O2 readily affords monomeric Ph3PnO (Pn = P, As; Fig. 1a). The situation changes substantially for Sb: no molecules containing an unperturbed Sb=O/Sb+–O− bond have ever been isolated.

Fig. 1: Synthesis of pnictine oxides.
figure 1

a, Oxidation of Ph3Pn yields monomeric Ph3PnO for Pn = P and As. b, Oxidation of Ph3Sb yields dimers or polymers. c, Oxidation of Mes3Sb yields trans-Sb(OH)2Mes3 (ref. 22). d, Lewis-acid-mediated disaggregation of (Ph3SbO)2 (ref. 13). e, Synthesis of an intramolecular stiborane-stabilized stibine oxide14. The chelating moiety (shown as a curved line) represents o-C6Cl4O2. f, Treatment of trans-Sb(OH)2Mes3 with sulfonic acids yields a hydroxystibonium salt and not a stibine oxide.

Full size image

A substance described as triphenylstibine oxide was first reported in 19388, and many other investigators subsequently purported to produce ‘Ph3SbO’ by treating Ph3Sb with H2O2 (ref. 9). Melting-point measurements and careful molecular-weight determinations showed these different substances to be dimeric or polymeric compounds and their structures were ultimately established with single-crystal X-ray diffraction and Sb extended X-ray absorption fine structure analysis (EXAFS; Fig. 1b)10,11,12.

Although Ph3SbO is not stable as a monomer, disaggregation of the polymer can be achieved with the Lewis acid B(C6F5)3 to afford the Lewis acid–base adduct Ph3SbOB(C6F5)3 (Fig. 1d)13. Another Lewis-acid-stabilized stibine oxide was obtained with a biphenylene-bridged system featuring a stibine oxide intramolecularly coordinated to a stiborane (Fig. 1e)14. A final example of Lewis-acid stabilization comprises [(3,5-F2C6H3)4SbOSbEt3][B(C6F5)4], which formed when a mixture of Et3Sb and [(3,5-F2C6H3)4Sb][B(C6F5)4] was exposed to oxygen15. In these compounds, interaction with a Lewis acid stabilizes the stibine oxide but also perturbs the Sb–O bonding interaction, preventing direct analysis of the periodic bonding trend across the pnictine oxides. These examples highlight that stibine oxides can be non-aggregated, but they raise the question of whether a monomeric stibine oxide is isolable in the absence of a Lewis acid interacting with and stabilizing the Sb=O/Sb+–O− bond. Matrix isolation studies afford evidence for the existence of monomeric H3SbO (Sb–O stretching frequency (νSbO) = 825 cm−1), but only in solid argon at 12 K (ref. 16).

We sought to explore a kinetic stabilization approach in which the reactive Sb=O/Sb+–O− bond is protected by sterically bulky groups, a strategy that has been used with great success in the stabilization of other reactive main-group bonds17,18,19,20,21. The bulky mesityl groups of Mes3Sb prevent polymerization upon treatment with H2O2, but not coordination sphere expansion: the product is the stiborane trans-Sb(OH)2Mes3 (Fig. 1c)22. Our re-investigation of reports of Mes3SbO showed that the reported species is, in fact, a hydroxystibonium cation (Fig. 1f)23,24. This work similarly called into question the previously reported (2,6-(MeO)2Ph)3SbO (ref. 25).

Results

Synthesis and characterization

We sought to prepare the even more sterically hindered stibine Dipp3Sb, 1a, where Dipp = 2,6-diisopropylphenyl. Although many R3Sb species are readily accessed from SbCl3 and either RMgBr or RLi, these strategies do not afford 1a. We therefore adapted a synthetic strategy developed by Sasaki and colleagues26,27, whereby the aryl group is installed on the Sb centre with an organocopper(I) species. In this way, 1a was isolated as a colourless, crystalline, air-stable solid (Fig. 2a). Although the 1H NMR spectrum of 1a indicates that rotation about the Sb–C and CAr–CiPr bonds is rapid on the NMR timescale at room temperature, the X-ray crystal structure highlights the extremely crowded environment around the Sb atom (Fig. 2b,c). The corresponding arsine (1b) and phosphine (1c) were similarly prepared (Fig. 2).

Fig. 2: Synthesis of sterically crowded pnictines 1a–c.
figure 2

a, Synthesis of 1a–c, Dipp = 2,6-diisopropylphenyl. b, Thermal ellipsoid diagram of 1a at the 50% probability level, with H atoms omitted for clarity. c, Space-filling diagrams of 1a–c from views rotated by 90° about the horizontal axis. Colour code: Sb, teal; As, purple; P, orange; C, grey; H, white.

Full size image

Addition of 1a to a suspension of PhIO in CH2Cl2 led to rapid consumption of the solid. Solvent was stripped from the reaction mixture and the residue was washed with pentane to yield a colourless solid, 2a (Fig. 3a). The infrared spectrum of 2a shows a new band at 779 cm−1, which we assign as a νSbO stretching frequency. This value is greater than any of the νSbO values of (Ph3SbO)2 (643/651 cm−1)10, trans-Sb(OH)2Mes3 (520 cm−1)22,28 or [Mes3SbOH][O3SPh] (612 cm−1)23. The 1H NMR spectrum of 2a is distinct from that of 1a and is consistent with a single Dipp environment with restricted rotation about the Sb–C bonds. Exchange spectroscopy (EXSY) and variable temperature (VT) NMR experiments confirmed the chemical exchange and reversible resonance coalescence (Supplementary Figs. 17 and 18).

Fig. 3: Oxidation of sterically crowded stibine 1a.
figure 3

a, Oxidation of 1a with PhIO to give 2a. b, Model compounds featuring different Sb–O bonding motifs. A, A dimeric stibine oxide; B, a dihydroxystiborane; C, a hydroxystibonium salt with X = O3SPh. c, Sb K-edge XAS spectra, with a green dotted line indicating where the derivative is maximal for 2a. Full normalized Sb K-edge XAS spectra are provided in Supplementary Fig. 109. d, Sb K-edge EXAFS (left) and Sb–C phase-corrected Fourier transforms (right). Experimental data are shown in blue and fits in red. The relevant EXAFS parameters are provided in Supplementary Table 7. A breakdown of the contributions of the different scatterers to the EXAFS and Sb–C phase-corrected Fourier transforms of A and 2a are shown in Supplementary Figs. 110 and 111, respectively.

Full size image

The oxidation state of 2a was probed with Sb X-ray absorption spectroscopy (XAS), which we have recently used to shed light on the structures of Sb-containing compounds29. The Sb K edge of 2a is 2 eV higher in energy than that of 1a (Fig. 3c). A similar shift was seen for a variety of Sb(V) compounds, including a dimeric stibine oxide (Ph3SbO)2 (A), a dihydroxystiborane trans-Sb(OH)2Mes3 (B) and a hydroxystibonium salt [Dipp3SbOH][O3SPh] (C, vide infra), indicating that 2a also contains Sb(V) (Fig. 3c). The K-edge EXAFS data were collected to high resolution to gain further insight into the structure of 2a. Similar data were collected from 1a, A, B and C for comparison (Fig. 3d). The Fourier transform of the data from A shows a distinct Sb···Sb scattering at 3.148(3) Å (superimposed on an outer-shell carbon backscattering), which is absent for B and C as well as 2a, indicating that 2a does not feature a dioxadistibetane. A detailed fit of the data from B shows two O scatters at 2.128(3) Å, whereas C is better fit by a single O scatterer at 1.905(1) Å. These values are in excellent agreement with the crystallographically determined structures of these compounds. In contrast, the data from 2a are best fit with a single O-atom scatterer at a distance of 1.837(2) Å, which is substantially shorter than the Sb–O bonds characterized for any other isolated materials. Fit of a C atom in the place of the short Sb–O gave a worse goodness-of-fit index (F = 0.335 for Sb–C versus 0.319 for Sb–O) and a physically unreasonable Debye–Waller factor (σ2 = 0.0010 Å2 for Sb–C versus 0.0021 Å2 for Sb–O).

Ultimately, we were successful in growing diffraction-quality single crystals of 2a. The asymmetric unit of the crystal structure features a single molecule of Dipp3SbO, which we unambiguously assign as the identity of 2a (Fig. 4a). Hirshfeld atom refinement (HAR) afforded a Sb–O bond length of 1.8372(5) Å, which is in excellent agreement with the EXAFS distance. The next-nearest Sb···O distance is 9.0791(4) Å; space-filling diagrams highlight the steric shielding provided by the Dipp groups (Supplementary Fig. 99). One of the iPr C–H units is directed at the stiboryl O atom with an O···H distance of 2.132(9) Å (note: HAR affords freely refined H-atom positions similar to those given by neutron diffraction)30. The C–H···O bond angle of 148.1(8)° suggests a strong electrostatic contribution to the interaction relative to weaker C–H···O H-bonds, in which isotropic van der Waals forces play a larger role31.

Fig. 4: The structure and bonding of 2a, a monomeric stibine oxide.
figure 4

a, Thermal ellipsoid plot of 2a at the 50% probability level. Colour code: Sb, teal; O, red; C, black; H, grey. b, Surface plots (isovalue = 0.05) depicting the Sb–O bonding NLMO (left) and overlap of O lone pair and Sb–C antibonding pre-orthogonalized NLMOs (right). Colour code: Sb, teal; O, red; C, grey; H, white. c, Contour plot of ρ overlaid with the gradient field lines of ρ for the Sb–O bond. d, Contour plot of ∇2ρ for the Sb–O bond, with positive values contoured with solid lines and negative values with dashed lines. e, Values of ρ (e− Å−3), ∇2ρ (e− Å−5) and ellipticity ε for 2a–c along the Pn–O bond paths, with Pn at the left and O at the right along the horizontal axis. The bond length is normalized to 1.00. The location of the (3, −1) critical point is shown with a dashed vertical line.

Source data

Full size image

The molecular geometry of 2a from our X-ray crystal structure is in excellent agreement with the one from a theoretical geometry optimization (PBE0/def2-TZVPP), which features an Sb–O bond length of 1.827 Å. Notably, the scaled theoretical νSbO of 781 cm−1 at the gas-phase optimized geometry is in excellent agreement with the experimental value for 2a (779 cm−1). These results combine to allow us to conclude that we have isolated an example of a monomeric stibine oxide. For comparison, we similarly synthesized and characterized the lighter congeners Dipp3AsO (2b) and Dipp3PO (2c).

Electronic structure

To gain insight into the nature of the Sb=O/Sb+–O− bonding motif, we analysed the topology of the theoretical electron density of 2a (DKH-PBE0/old-DKH-TZVPP) (Fig. 4c–e). This analysis shows the locations of critical points in the electron density (ρ), that is, points in space where the derivative of ρ is zero in three mutually orthogonal directions. These critical points are characterized with a pair of numbers (ω, σ), where ω is the number of non-zero eigenvalues of the Hessian and σ is the sum of the signs of those eigenvalues32. As expected, a (3, −3) critical point is present near the nuclear position of each atom. We also identified (3, −1) critical points, also known as bond critical points, between each of the covalently bonded atoms (Supplementary Fig. 76). Although the values of various real-space functions at a (3, −1) critical point are frequently used to describe the nature of that bonding interaction32, for polar covalent bonds, like the Sb+–O− bond in a stibine oxide, these functions are more informative when evaluated along the length of the bond path (Fig. 4e)33. For the Sb–O bond of 2a, ρ features a single minimum at 0.173 e−Å−3, approximately halfway along the bond path. In the valence bonding region, the Laplacian (∇2ρ) exhibits a single O-proximal minimum of −1.346 e−Å−5. The ellipticity (ε) is negligible along the length of the bond path. The decrease in ρ and |∇2ρ| in the internuclear space from 2c to 2b to 2a suggests a systematic weakening of the Pn+–O− bond as the Group 15 element increases in atomic number. We note that the topological analysis also located bond paths connecting the pnictoryl O atoms and the iPr C–H units, as suggested by the crystallographic data. The analysis of the bond critical points of these interactions (Supplementary Tables 12 and 13) indicates that hydrogen-bonding interactions are present and that they decrease in magnitude from 2a to 2b to 2c.

Further insight into the Sb–O bonding in 2a was obtained from molecular orbital analyses. The canonical molecular orbitals (CMOs) are, as expected, highly delocalized across the molecule (Fig. 5b). The frontier CMOs feature substantial π or π* character from the Dipp substituents. The nearly degenerate highest occupied molecular orbital (HOMO) and HOMO–1 feature a substantial contribution from the lone pairs on the O atom. The lowest unoccupied molecular orbital (LUMO) features a substantial amount of Sb–O σ* character. The Dipp groups block the lobe of this orbital that extends opposite the Sb–O bond, which probably contributes to the stability of this molecule, as designed.

Fig. 5: Variation in the electronic structure of the pnictine oxides 2a–c.
figure 5

a, Calculated orbital energies (DKH-PBE0/old-DKH-TZVPP//PBE0/def2-TZVPP) in electronvolts, with frontier molecular orbitals shown in colour (P, orange; As, purple; Sb, teal). b, Canonical molecular orbital diagrams of 2c, 2b and 2a. Colour code surfaces: red, positive; blue, negative (isovalue = 0.02). c, Electrostatic surface potential (ESP) mapped van der Waals surfaces of 2c–a (values in kcal mol−1). The value of the surface minimum is indicated. C, grey; H, white; O, red; P, orange; As, purple; Sb, teal.

Source data

Full size image

More detailed information was obtained by analysing the natural localized molecular orbitals (NLMOs) of 2a (Fig. 4b and Supplementary Figs. 91–93). An Sb–O bonding NLMO is present and is polarized 74:25 toward the more electronegative O atom, which uses a hybrid atomic orbital enriched in p character (79%) to interact with the Sb. The Sb–O antibonding orbital is correspondingly polarized toward the Sb and exhibits the large lobe opposite the Sb–O bond that was observed in the LUMO CMO. There are two O-centred lone pair natural bond orbitals (NBOs) with nearly pure p character and a second-order perturbation theory analysis uncovered donor–acceptor interactions that delocalize electron density from these lone pairs into Sb–C antibonding orbitals (Supplementary Table 14). Similar delocalizations were observed for 2b and 2c, and deletion calculations showed that the non-covalent interactions between the O and Dipp3Pn fragments decreased from 2c to 2b to 2a. These donor–acceptor interactions strengthen the Pn–O bonds, and the decreased delocalization in 2a affords the lowest Wiberg Pn–O bond order of the three, but the O atom consequently retains the greatest natural atomic charge (Supplementary Table 15). The variation in charge accumulation is also reflected in the magnitude of the electrostatic surface potential minimum, for which 2c < 2b < 2a (Fig. 5c). The decrease in Pn+–O− bond strength (PO > AsO > SbO), is also reflected in the Pn+–O− stretch force constants (Supplementary Fig. 75) and the ratio of ΔEorb/ΔEtotal from an energy decomposition analysis of O and Dipp3Pn fragments (Supplementary Tables 8–10). Deformation density analyses show a redistribution of electron density from the Dipp3Pn fragment to the O atom to an extent that decreases from Sb to As to P (Supplementary Fig. 90).

Donor–acceptor interactions were also observed from the O-centred lone pairs to the iPr C–H antibonding orbitals for 2a–c (Supplementary Fig. 93), consistent with the presence of the O···H bond paths noted above. Non-covalent interaction analysis of 2a (Supplementary Fig. 89) uncovered a region with a negative product of ρ and the sign of the second-largest eigenvalue of the Hessian of ρ, sign(λ2)ρ, between the O and iPr C–H; the value of sign(λ2)ρ was less negative for 2b and 2c, indicating that this interaction, which may help to stabilize the Sb+–O− bond, is present in 2a and weakens for 2b and 2c. The presence of this hydrogen-bonding interaction in 2a was further confirmed by NBO perturbation theory and deletion calculations (Supplementary Fig. 93).

Reactivity

With an isolated stibine oxide in hand, we next explored its chemistry. The bonding characteristics outlined above suggest that 2a should exhibit O-centred Lewis-basic behaviour. Cooling a solution of 2a in neat 4-fluoroaniline affords colourless blocks, which X-ray diffraction analysis confirmed to contain the stibine oxide-aniline hydrogen-bonded adduct 3 (Fig. 6(i)). In the HAR model, the hydrogen-bonding H atom of 3 is located on the N atom with a N–H distance of 1.04(2) Å. The N···O distance of 2.858(1) Å implies that the hydrogen-bonding interaction is of moderate strength. The Sb–O bond remains short at 1.8421(7) Å, but is statistically significantly lengthened as compared to 2a. Consistent with this bond lengthening, the Sb–O IR stretching frequency decreases slightly from 779 cm−1 for 2a to 762 cm−1 for 3. Neither 2b nor 2c affords a similar product, consistent with the lower nucleophilicity of the O atoms in these species.

Fig. 6: Reactivity of the stibine oxide 2a.
figure 6

(i) Reaction with an aniline to form a hydrogen-bonded adduct. (ii–iv) Reactions with copper(I) chloride, silver(I) triflate and triphenylphosphinegold(I) triflate to yield transition-metal coordination complexes. (v) Reaction with a sulfonic acid to yield a hydroxystibonium salt. (vi) Reaction with acetic acid to yield a cis-hydroxyacetatostiborane. (vii) Reaction with BF3 to yield a trans-difluorostiborane. (viii) Reaction with phenylsilane to yield stibine 1a. Thermal ellipsoid plots at the 50% probability level are shown next to products; non-polar H atoms are omitted for clarity. Sb, teal; O, red; C, black; H, grey; N, blue; Ag, purple; F, green; S, yellow; Au, gold; P, orange; Cu, mauve.

Full size image

We next sought to determine whether this Lewis basicity would also manifest in metal ion coordination. Combination of 2a with 1 equiv. of CuCl yielded the complex (Dipp3SbO)CuCl (4; Fig. 6(ii)), whereas combination with 0.5 equiv. of AgOTf yielded [(Dipp3SbO)2Ag][OTf] (5; Fig. 6(iii)). If ClAu(PPh3) was treated with AgOTf and 2a, the salt [(Dipp3SbO)Au(PPh3)][OTf] (6; Fig. 6(iv)) was isolated. All three of the complexes were crystallographically characterized, and all three exhibit significantly nonlinear Sb–O–M angles (Supplementary Table 18). Solution of the structure of a second polymorph of 6 showed, however, that the complex can also take on a rigorously linear configuration. The Sb–O–M bending is most probably driven by crystal packing forces. We note that, in all cases, the geometry about the metal centres in 4–6 is nearly perfectly linear, as expected. Neither 2b nor 2c was able to form analogous complexes; the NMR resonances of these lighter pnictine oxides exhibited only minor shifts upon mixing with the metal precursors (Supplementary Fig. 49–51). We note that the strength of the intramolecular CHiPr···O interaction decreases upon coordination of 2a (Supplementary Tables 13 and 15).

Room-temperature treatment of 2a with a strong Brønsted acid, PhSO3H, resulted in clean formation of the hydroxystibonium salt [Dipp3Sb(OH)][O3SPh] (7a; Fig. 6(v)). Crystallographic analysis of the salt confirmed protonation at the Sb-bound O atom, which lengthens the Sb–O bond to 1.9119(7) Å and decreases νSbO to 611 cm−1. Compound 2b can be similarly protonated to yield [Dipp3As(OH)][O3SPh] (7b; Supplementary Fig. 106). Compound 2c interacts much more weakly with PhSO3H, but titration with up to 10 equiv. of the acid results in a systematic shift in the NMR resonances of 2c. This behaviour may arise from reversible formation of a hydrogen-bonded adduct in equilibrium with the dissociated species (Supplementary Figs. 59 and 60).

We were surprised to find that, in contrast, acetic acid not only protonates the O atom of 2a, but adds across the Sb–O bond at room temperature, affording the neutral stiborane cis-Sb(OH)(OAc)Dipp3 (8; Fig. 6(vi)). This 1,2-addition chemistry highlights the unsaturated nature of the stiboryl (Sb=O/Sb+–O−) group. The cis isomer forms despite the expectation that the more sterically bulky Dipp groups would assume the less-crowded equatorial positions and that the more apicophilic hydroxy and acetoxy groups would assume the trans-disposed axial positions. An intramolecular hydrogen-bonding interaction is present between the hydroxy and acetoxy groups (O···O = 2.630(2) Å), which may be responsible for the cis configuration. Neither 2b nor 2c reacts in this manner with acetic acid (Supplementary Figs. 64–66). We have yet to observe any cycloaddition chemistry (Supplementary Fig. 74), but substrates continue to be explored.

Combination of 2a and BF3·OEt2 at −78 °C results in rapid and clean conversion to 9, which does not feature an 11B NMR signal, but does exhibit a single sharp 19F resonance at −74.35 ppm. X-ray diffraction analysis shows 9 to be the difluorostiborane trans-SbF2Dipp3 (Fig. 6(vii)). This reaction is the first in which we have observed cleavage of the Sb–O bond. The maintenance of the 5+ oxidation state of the Sb centre suggests that the reaction is an oxide transfer in which one oxide is exchanged for two fluoride groups, consistent with the fluorophilicity of organoantimony(V) Lewis acids34. We have not fully characterized the by-product, but boranes are known to form stable boroxines with O2− (ref. 35). Unlike 8, 9 features the apicophilic fluoro substituents in the expected trans geometry (Sb–F = 1.9673(34) and 1.9706(31) Å), most probably as a result of the increased electronegativity of F and a lack of hydrogen bonding. No deoxygenation was observed upon combination of 2b or 2c with BF3·OEt2.

Finally, we observed that PhSiH3 is able to abstract the O atom from 2a to cleanly afford 1a (Fig. 6(viii)). The reaction does not proceed at room temperature, but readily reaches completion within 1 h at 50 °C. Under these mild conditions, neither 2b nor 2c reacts with PhSiH3 (Supplementary Figs. 72 and 73).

Discussion

The isolation of a monomeric stibine oxide, 2a, was achieved using a kinetic stabilization approach in which the unsaturated Sb+–O− bond is protected by the sterically bulky Dipp groups bound to the Sb centre. The isolation of 2a has permitted the spectroscopic and crystallographic characterization of this functional group. In combination with these experimental measurements, theoretical calculations provide insight into the nature of the Pn–O bonding interaction, and the variation in this bonding as the pnictogen is varied from Sb to As to P. The increased accumulation of charge on the O atom confers upon 2a reactivity that differs notably from that of 2b and 2c. We have described examples of 2a acting as a hydrogen-bond acceptor, a transition-metal ligand and a Brønsted base. The unsaturated nature of the Sb+–O− bond also allows it to engage in addition chemistry, as exemplified by the reaction with acetic acid. Finally, the Sb–O bond can be cleaved, either with maintenance of the Sb(V) oxidation state, as in the reaction with BF3, or with reduction to Sb(III), as in the reaction with PhSiH3. We will continue to investigate in greater depth each of these classes of reactions, and others, with an emphasis on comparing and contrasting the reactivity of stibine oxides with that of phosphine and arsine oxides.