Harnessing Sulfinyl Nitrenes: A Unified One-Pot Synthesis of Sulfoximines and Sulfonimidamides

Sulfoximines and sulfonimidamides are promising compounds for medicinal and agrochemistry. As monoaza analogues of sulfones and sulfonamides, respectively, they combine good physicochemical properties, high stability, and the ability to build complexity from a three-dimensional core. However, a lack of quick and efficient methods to prepare these compounds has hindered their uptake in molecule discovery programmes. Herein, we describe a unified, one-pot approach to both sulfoximines and sulfonimidamides, which exploits the high electrophilicity of sulfinyl nitrenes. We generate these rare reactive intermediates from a novel sulfinylhydroxylamine (R–O–N=S=O) reagent through an N–O bond fragmentation process. Combining sulfinyl nitrenes with carbon and nitrogen nucleophiles enables the synthesis of sulfoximines and sulfonimidamides in a reaction time of just 15 min. Alkyl, (hetero)aryl, and alkenyl organometallic reagents can all be used as the first or second component in the reaction, while primary and secondary amines, and anilines, all react with high efficiency as the second nucleophile. The tolerance of the reaction to steric and electronic factors has allowed for the synthesis of the most diverse set of sulfoximines and sulfonimidamides yet described. Experimental and computational investigations support the intermediacy of sulfinyl nitrenes, with nitrene formation proceeding via a transient triplet intermediate before reaching a planar singlet species.

Proof of concept was quickly obtained by reacting PhONSO SI-5 with 4fluorophenylmagnesium bromide at 0 °C and then adding morpholine to give sulfonimidamide 3a. However, isolated yields higher than 63% could not be obtained despite varying equivalents of morpholine, time and temperature of the second step, and the addition of acid (Table S1). Another issue was that this reaction set-up was not amenable to the selective synthesis of sulfoximines by sequential addition of two organometallic reagents, which was the primary aim of the project.

4-(5-Bromopyridine-3-sulfonimidoyl)morpholine 3ac
(5-Bromopyridin-3-yl)magnesium chloride lithium chloride complex was prepared according to a literature procedure. [11] Note: the Grignard reagent was prepared on 4x larger scale than the reaction with BiPhONSO as the magnesium-halogen exchange failed on small scale. One quarter of the resulting solution was used.

Notes on Order of Addition of Organometallic Reagents
• Aryl and small alkyl (methyl, n-alkyl and cyclopropyl) Grignard reagents may be added as the first or second nucleophiles with little or no effect on yields and side-product profile • Grignard reagents derived from bulkier alkyl groups such as isopropyl or tert-butyl should be added as the first nucleophile (i.e. before aryl or small alkyl Grignard reagents) to avoid formation of the N-alkylsulfinamide (by attack on the sulfinyl nitrene nitrogen) instead of sulfoximine • No evidence of N-arylsulfinamide formation from attack of aryl Grignard reagents on nitrogen has been observed, even when two bulky ortho-substituted aryl Grignard reagents are used (e.g. compound 2h).

Preparation and Reactions of N-H Sulfonimidate Esters [1,1'-Biphenyl]-4-yl 4-methoxybenzenesulfonimidate 5a
BiPhONSO Note: the product was unstable to silica and decomposed even upon standing over several days at -20 °C, likely due to self-condensation. This is consistent with a report on similar unprotected O-aryl-NH-sulfonimidate esters prepared using an alternative method. [14] Obtaining pure material was therefore challenging and the compound was characterised as a 55:45 mixture of the sulfonimidate ester and 4-phenylphenol, the by-product of selfcondensation. Only the peaks belonging to product 5a are reported below.

Addition of Radical Inhibitors
The divergent reactivity seen when tert-butylmagnesium chloride was added as the second nucleophile (formation of sulfinamide 4a instead of sulfoximine 2v) raised the possibility that a different mechanism could be operative. In order to test whether radicals may be involved, we repeated several of our reactions with the radical traps (2,2,6,6-tetramethylpiperidin-1yl)oxyl (TEMPO) and 1,1-diphenylethylene added prior to the Grignard or amine nucleophiles.
Although yields were generally 10-20% lower than the standard reactions, there was no evidence of adduct formation and the radical inhibitors were recovered unreacted. This indicates that the reaction does not involve radical species and that the sulfinamide product 4a is formed through a polar addition of the tert-butylmagnesium chloride to the nitrene nitrogen rather than sulfur (likely due to steric factors).

Computational Method
All calculations were performed using the Gaussian 16 C.01 software package. [15] Geometry optimisations and frequency calculations were obtained at the B3LYP-D3BJ/6-31G(d) level of theory.
[16] Solvent effects were accounted for with the SMD implicit solvent model using parameters appropriate for tetrahydrofuran at the same level of theory. [17] Vibrational frequencies were calculated to assign stationary points as either minima or transition In all cases, vibrational entropies were obtained using a quasi-harmonic approximation, treating vibrational modes below 100 cm -1 as free rotors and as rigid rotors above this cut-off, as first proposed by Grimme [18] and implemented in Goodvibes.py [19] . All Gibbs free energy profiles presented here use the values obtained at 195 K unless otherwise stated.
The minimum-energy crossing points (MECPs) between the singlet and triplet potentialenergy surfaces were located using the python script easyMECP.py, [20] based on a programme originally developed by Harvey et al. [21] Energies and gradients were calculated at the SMD(THF)-ωB97X-D/6-311++g(d,p) level of theory. A state-averaged vibrational analysis at the MECP was carried out with Glowacki's glowfreq code. [22] A natural bond orbital (NBO) population analysis was conducted on the optimised structures at the B3LYP-D3BJ/6-311++g(d,p) level of theory unless otherwise stated. The NBO population analysis was performed with NBO 6.0 integrated with the Gaussian 16 programme. [23] The structures generated are visualised using the software CYLView. [24] S59

Computational Benchmarking
To evaluate the effect of the level of theory on the geometries, crystal structures of magnesium diimidosulfinate compounds ( Figure S2), [25] which resemble the intermediates studied here, were obtained from the CSD database. They were optimised using five different DFT  Overall, all the functionals tested provided geometries in good agreement with X-ray data, as  Table S2. Comparison of mean absolute deviation in Angstroms (Å) considering seven relevant bond distances of optimised structures relative to X-ray crystallographic structures. In all cases, geometries were optimised using the 6-31g(d) basis set.

Structure of PhONSO
To reduce the computational cost, and assess the possible conformations, a model system was constructed in which sulfinylhydroxylamine reagent I (BiPhONSO) was modelled as PhONSO ( Figure S3). respectively. The Z-conformations were found to be lower in energy, with PhONSO-Z-P (referred to as simply PhONSO in Figure 4b and here on after), being preferred. This is in agreement with X-ray crystallography data of analogous compounds [28] and can be explained by the stabilising donation of the S lone pair into the N-O 2 σ* (6.9 vs 0.0 kcal mol -1 in PhONSO-E-P) and the N lone pair into the S-O 1 σ* (8.1 vs to 2.7 kcal mol -1 ) (Table S3)

Grignard Addition to PhONSO
The first step of our reaction design uses a sulfinylhydroxylamine (I, main text, Figure 1e), which upon combination with a Grignard reagent forms the negatively charged sulfinamide intermediate II.
Computationally, exploring reactions involving Grignard reagents is challenging, as these compounds can have a variety of coordination geometries and oligomeric states, which depend on the solvent, the halogen and the organic substituents on the magnesium atom. [29], [30] Crystallographic data have suggested the presence of monomeric, dimeric and tetrameric species; however, these structures may arise as a result of lattice packing effect and not necessarily represent the species present in solution. Spectroscopic, calorimetric, [31] and computational analyses have previously been used to explore the nature of these species in solution. For example, computational studies, in THF solvent, suggested that CH3MgCl favours a dimeric structure with bridging ligands and a preference for halide over alkyl in that position. [32] Additionally, recent computational studies have shown the possibility for parallel reaction pathways in solution. [33] Based on these observations, we first analysed the mechanism where a dimeric Grignard species is present (referred to as dimeric mechanism). However, later we also present the possibility of having a monomeric species (referred to as monomeric mechanism), as a possible alternative pathway which may operate in the reaction. Both mechanisms give qualitatively similar results; however, based on previous computational studies supporting a dimeric mechanism for methyl addition to carbonyl compounds, we focus our analysis on this pathway (Figure 4b & Figure S9). [26a, 33-34]

Dimeric Pathway
The relative stability of bridging Cl and methyl groups on the dimeric structure of CH3MgCl, has been previously explored, [32] suggesting that the former is thermodynamically favoured.
Our results show that the bridging Cl groups are only marginally preferred over methyl groups by 0.3 kcal mol -1 (D0 and D0a, Figure S4). The activation barriers for the formation of the N-phenoxy sulfinamide-Grignard complexes (D3, D3a & D3b), is also lower for the pathway having bridging Cl groups (path-a, Figure S4) compared to path-b and path-c respectively.
This suggests that a dimeric species with bridging Cl groups is likely to be the reactive species. Crystallographic data of complexes with Grignards often show the close association of solvent molecules such as diethyl ether (Et2O) or tetrahydrofuran (THF), including previous studies on the closely related magnesium diimidosulfinate complexes. [25] Therefore, most computational models have utilised explicit rather than implicit solvation in their calculations, [26a, 32-34] hence the effect of using an explicit solvent (in addition to the implicit solvation) on the energetics of path-a,b,c was evaluated ( Figure S5). solvation, comparable with previous calculations. [33] Similar to the results obtained with implicit solvent, path-a and path-c are found to be preferred and similar in energy, while path-b was found to be 2.8-3.1 kcal mol -1 higher in energy ( Figure S5). From these results, and to reduce the computational cost, we decided to only use an implicit solvent model to explore the Grignard addition and generation of the sulfinyl nitrene intermediate steps.

Sulfinyl Nitrene
As suggested in the main text (Figure 1e), the sulfinyl nitrene species III is likely to arise from a N-O bond fragmentation of D3, similar to that proposed for the formation of sulfonyl nitrenes. [27a] Sulfinyl nitrenes are expected to be highly reactive, and as a result they have only been observed by matrix-isolation from flash vacuum pyrolysis of sulfinyl azides. [27b, 27k] To gain insights on the nature of the sulfinyl nitrene, intermediate calculations were conducted on the free nitrene species ( Figure S6).
In agreement with previous computational studies on sulfinyl nitrene intermediates, [27b] the singlet state is more stable than the triplet by 21.0 kcal mol -1 (Nitrene-S1 compared to Nitrene-T1), due to the availability of the sulfur lone pair to donate electrons into the electron-deficient nitrene centre in the former. This contrasts with sulfonyl nitrenes, which predominantly favour a triplet ground state, [27a, 27f, 27h, 35] indicating the different chemistries of these nitrene species.   Table S4. Key bond lengths and Wiberg bond orders [36] in Nitrene in both the singlet (S1) and triplet state (T1 shown in parenthesis) at the ωB97X-D/6-311+g(d,p)//B3LYP-BJD3/6-31g(d) level of theory.
The minimum energy crossing point Nitrene-MECP from the triplet to the singlet nitrene is only 0.9 kcal mol -1 higher in energy than Nitrene-T1 with a triplet character and similar geometry. At the singlet state (Nitrene-S1), the S-N bond is shortened, while other bond lengths remain identical to the triplet. The MECP for sulfinyl nitrene bound to the magnesium complex (D4', D-MECP2 & D5, Figure 4b) also indicates that the triplet nitrene is likely to be a kinetically unstable species and preferentially convert to the more stable singlet nitrene D5 (III) with little influence of the counterion present.

Grignard-Sulfinyl Nitrene Complex
The singlet nitrene species in the presence of a magnesium complex D5, contains the phenoxide ligand as a non-bridging ligand. This is able to rearrange to a bridging ligand to form the species D5' (Figure 4b). Other possible anionic arrangements of D5' were investigated in order to assess their potential impact ( Figure S7 a & b).

S66
For species D7 it was found that having phenoxide as a bridging ligand was more favourable on average by 7.6 kcal mol -1 (D5' & D5'i vs D5'ii & D5'iii). Therefore, species with the chlorides as bridging ligands were not considered to have a substantial impact on the dimeric pathway.
From a singlet sulfinyl nitrene intermediate in the presence of a magnesium complex (D5) the possibility of an anionic sulfonimidate ester species (5a -) is possible. This is demonstrated in the experimental mechanistic investigations; when using 1 equivalent of Grignard followed by addition of TsOH, sulfonimidate ester 5a was isolated (main text, Figure 4a). The complexed anionic form of the sulfonimidate ester species (D8, Figure S7 c & d) is shown to be close in energy to the analogous complex with sulfinyl nitrenes ( Figure S7 a & b). Notably, sulfonimidate ester anion D8iii which is coordinated to magnesium via three heteroatoms is thermodynamically more stable than sulfinyl nitrenes D5' by 2.4 kcal mol -1 . The similar stability of these species means that formation of the sulfonimidate ester does not provide a thermodynamic sink for the reaction.

Second Grignard Addition
For the second methyl addition, we explore the possibility of having nitrogen or oxygen bound to magnesium, as well as methyl addition to either sulfur or nitrogen ( Figure S8). Due to the unsymmetrical nature of the magnesium complex, each transition state has two possible orientations of the sulfinyl nitrenes with regard to the bridging ligands. As both orientations were found to lead to transition states of similar energy, only the lowest energy pathways are presented below.

Monomeric Pathway
As mentioned before, crystallographic data have suggested the existence of monomeric Grignard complexes as well as the possibility of a parallel monomeric mechanism to operate. [33] For this reason, the likelihood of this species being present in solution was also explored to evaluate if the same mechanistic conclusions were obtained ( Figure S9).

S69
crossing of M4 to M5Cl-OPh via M-MECP2, is now 10.7 kcal mol -1 higher in energy than M4. The high energy cost associated to this crossing arises from the absence of η 2 hapticity in the singlet sulfinyl nitrene (M5) which would otherwise cause an increase in angle strain. As shown above, the sulfonimidate ester anions M9 and M9i are close in energy to M5Cl-OPh similar to the analysis of the dimeric complexes ( Figure S7). Alternative anions bound to the magnesium centre from Figure S9 were investigated but had little impact on the mechanistic conclusion. Analogous to the analysis for the dimeric pathway for the second methyl addition to the singlet sulfinyl nitrene, all possible species in equilibrium were first considered ( Figure S10). This shows that the most stable singlet nitrene is M5iCl-OPh with the chloride and phenoxide coordinated to the magnesium centre and the nitrenes nitrogen bound to magnesium. Hence the activation barrier of the second methyl addition is considered from M5iCl-OPh, where the first step is anion exchange to M5Cl-Me, M5iCl-Me, M5Me-OPh or M5iMe-OPh followed by the addition. with MOPh-TS5-7 giving the lowest activation energy. For methyl addition to sulfur, oxygen coordination to the magnesium is preferred over nitrogen, opposite to the dimeric pathway ( Figure S8). Moreover, addition of methyl to the nitrogen centre to form the sulfinamide product is kinetically less favourable comparing MCl-TS5i-8 with MCl-TS5-7 (2 kcal mol -1 higher), in agreement with the experimental results showing preference for sulfoximine formation in most cases. Therefore, the model for the monomeric pathway provides the same qualitative conclusion to the dimeric pathway when considering the selectivity of methyl addition to either the sulfur or nitrogen centre of the nitrene. It is worth noting that the dimeric mechanism provides a much larger difference in barrier height (5.7 kcal mol -1 ), which validates the use of the dimeric model, as experimentally most Grignard reagents produced the S-alkylation products in good to excellent yields.