Introduction

The [1,2] Brook rearrangement provides a transfer of the nucleophilic site from the oxygen atom to its adjacent carbon atom mediated by a Brønsted base, thus finding wide application in the unique methodology design and organic synthesis1,2,3. As these α-silyl carbinols can be readily obtained from carbonyl compounds, this rearrangement offers an effective means of polarity inversion at the carbon center of carbonyl groups. Therefore, since its development by Brook in the 1950s4, extensive research has not only hatched many transformations based on the carbanion but also established a clear understanding of the mechanism (Fig. 1a). General understandings recognize that the reaction involves a C-O-Si three-membered ring intermediate, followed by the C–Si bond cleavage to accomplish the rearrangement. Upon on these research foundations, one of its most important variants, the [1,2]-phosphonate-phosphate rearrangement, also known as the [1,2]-phospha-Brook rearrangement, has garnered considerable attention as a tool for generating carbon-based anionic nucleophiles in recent years5,6,7,8. This conversion often needs pentavalent phosphonates to accept oxygen ion attacks, while the existing P=O double bond effectively stabilizes the reaction intermediate during the process, facilitating the rearrangement through the C–P bond cleavage. Obviously, the absence of these characters makes the use of trivalent phosphines seem unreasonable in this process, thus, they cannot undergo the traditional [1,2]-phospha-Brook rearrangement (Int. A).

Fig. 1: Brook rearrangements and phosphonylation.
figure 1

a Brønsted base-mediated Brook rearrangements. b Lewis acid-mediated trivalent phospha-Brook rearrangements. c Deoxygenative phosphonylation of carbonyls. d The reported routes to access tertiary phosphine oxides involving deoxygenation.

Full size image

Mechanistically, despite trivalent phosphines in this structure not being able to be electrophilic sites, taking them as initiators of intramolecular charge transfer (ICT) processes for rearrangement reactions still appears highly intriguing. In such a case, we wondered that, unlike those traditional Brook reactions requiring a Brønsted base, it instead necessitates the assistance of a Lewis acid to active the C–O bond, thus transforming the initiator role from oxygen to phosphorus atom in this system (Fig. 1b). It probably forms a C-P-O three-membered oxirene structure with a C–P double bond, which could undergo a rearrangement via the C–O bond cleavage, leading to tertiary phosphine oxide formation. While some preliminary works have been reported9,10,11,12, they often suffered from limited applicability. Moreover, although an epoxide intermediate was once proposed10, the lack of thorough investigations renders the mechanism highly elusive and controversial, as well as leaving this trivalent phospha-Brook rearrangement unrecognized. The absence of systematic research in this field, both in practical applications and theoretical development, has significantly hindered the advancement of this chemistry.

In this work, we disclose a metal-free trivalent [1,2]-phospha-Brook rearrangement from carbonyl compounds to rapid access tertiary phosphine oxides (Fig. 1c). This current BF3-mediated deoxygenative phosphonylation can be easily carried out using trivalent diaryl or dialkyl phosphine reagents with easy-to-handle procedures. The reaction is applicable to both aldehydes and ketones, demonstrating broad substrate generality and functional group compatibility. Additionally, detailed mechanistic studies are conducted to provide clear insights into this unique rearrangement process. Regarding the facile accessibility of carbonyl compounds, our method offers a highly convenient approach for synthesizing these tertiary phosphine oxides, which are important compounds that are widely used in biochemistry or the precursor for ligand synthesis, and various strategies have been devised for their preparation. Phosphonylation involving deoxygenation using alcohols13,14,15,16,17, acids18,19,20,21, and esters22,23,24,25 as precursors has also been reported (Fig. 1d). However, these processes often require in situ or pre-prepared activation steps (such as through redox-active esters) to facilitate the subsequent C–P bond formation reactions under metal catalysis or photocatalysis, while deoxygenative conversion of ketones typically necessitates their prior transformation into hydrazone derivatives or relatively narrow scopes are observed26,27,28,29,30,31. In comparison, the current method offers many distinct advantages, including broad substrate applicability, high functional group tolerance, the use of readily available reagents, metal-free conditions, operational simplicity, and excellent atom-economy. These features make it highly promising for practical applications.

Results

Reaction optimization

Based on the above considerations, we initially employed compound 1 as the model substrate and systematically screened various Lewis acids (Fig. 2a). It was found that the use of BF₃·Et₂O in toluene at 120 °C could afford the desired product (2) with a 78% isolated yield. In the absence of this Lewis acid, no product was formed, and its analogs BBr₃ or BCl₃ did not yield any detectable compound 2. The commonly used B(C₆F₅)₃ also failed to promote this transformation. While a small amount of the desired product was obtained with TiCl₄, other Lewis acids, including Sc(OTf)₃, InCl₃, TMSOTf, AlCl₃, and NbCl₅, did not give any tertiary phosphine oxide. It is worth mentioning that the combination of BF₃·Et₂O and Lewis metal catalysts resulted in less favorable outcomes, whether at room temperature or under high-temperature conditions. Pentavalent phosphorus sources, such as HP(O)Ph2, HP(O)(OEt)2, cannot be utilized to proceed this deoxygenative phosphonylation, demonstrating the difference from the traditional phospha-Brook rearrangement. The conditions reported in previous works are insufficient or cannot effectively achieve this transformation, further highlighting the excellent efficiency and the real practical values in the organic synthesis of this metal-free approach (for details, see supplementary information).

Fig. 2: Reaction optimization and scope studies with respect to Ketones.
figure 2

a Reaction development. b The scope studies with respect to ketones under standard conditions. aConditions: ketone (0.2 mmol), HPPh2 (2.5 equiv.), BF₃·Et₂O (2 equiv.), toluene (1 mL), 120oC, 12 h, N2 atmosphere. The isolated yields were given. bMesitylene (1 mL) was used as solvent. cBF₃·Et₂O (3 equiv.) was used. d6 h.

Full size image

Scope studies

Under these remarkably simple conditions, a diverse array of ketones, comprising both acyclic (3–11) and cyclic candidates (12–19), either aliphatic or aromatic ones, were tested to prove the excellent scope of this method (Fig. 2b). Meanwhile, it also exhibits good functional group compatibility, which could further establish a solid foundation for the synthesis of tertiary phosphine-containing compounds. For instance, alkyl chlorides (3) and esters (5, 7) remain unaffected, despite the involvement of HPPh₂. Various heterocycles, such as indole (8), furan (8), pyrrole (9), thiazole (10), thiophene (11), tetrahydrofuran (12), and pyrrolidine (13), were efficiently incorporated into the substrates to produce the corresponding target products. Ketones bearing ferrocene skeleton can also be used to give tertiary phosphine oxides with this method (14). Cyclic ketones, regardless of the ring size (18, 19), are readily employed as substrates for the reaction. Even the highly rigid adamantly ketone could undergo this deoxygenative phosphonylation in very high yield (17). Moreover, complex molecules (20–24) bearing drug structures (e.g., Nabumetone, Pentoxifylline, Oxaprozin, Zaltoprofen) or natural products (e.g., Mesterolone) can be smoothly imbedded with phosphorous groups under these conditions, further underscoring the great potential of this method for drug modification and discovery. The diastereoselectivity of the C–P bond formation often depends on substrate control, allowing for exceptionally high dr values in some cases (23, 24).

Given the abundant availability of aldehydes, we then expanded our study to explore a broader range of aldehyde substrates (Fig. 3a). The results also showcased the robust applicability to both aliphatic (25–32) and aromatic (33–40) aldehydes. In the case of aliphatic aldehydes, candidates with α-secondary (25–27), tertiary (28–31), or quaternary (32) centers can be all successfully converted into the corresponding products. Additionally, this approach again demonstrated its remarkable tolerance for a variety of functional groups. In addition to ethers (25), imides (26) and amines (29) were tolerated, too. Notably, the Weinreb amide, a widely used precursor for ketone synthesis, remained intact in this reaction (27). Aryl halides, such as aryl chlorides (31) and bromides (35), which can be further coupled through the well-established transition metal-catalyzed strategies, were preserved in the process. Functional groups like SCF₃ (33), sulfonates (34), and sulfonamides (43) were compatible with this transformation. The presence of a trivalent phosphine group in the substrate with strong coordination ability did not impede the reaction, owing to the advantages of a metal-free system (36). Beyond the heterocycles mentioned earlier, this study further confirmed the compatibility with benzofuran (37), pyrazole (39), imidazole (40), and oxazole (42) under these conditions. Additionally, the impressive applicability of this method was further underscored by the success with more examples of complex molecules bearing an aldehyde group (41−48). Moreover, besides the commercially available HPPh₂, other disubstituted trivalent phosphines proved effective in synthesizing various tertiary phosphine oxides from carbonyl compounds (49–51), with a particular emphasis on the dialkyl-substituted target product (52) (Fig. 3b).

Fig. 3: Scope studies with respect to aldehydes and phosphines.
figure 3

a Aldehydes scope studies. b Reactions with various phosphine. Conditions: aldehyde or ketone (0.2 mmol), HPPh2 (2 equiv.), BF₃·Et₂O (1 equiv.), toluene (1 mL), 100oC, 12 h, N2 atmosphere. The isolated yields were given. aBF₃·Et₂O (2 equiv.) was used. bBF₃·Et₂O (3 equiv.) was used. c120oC.

Full size image

Application of products and the method

To further demonstrate the applicability and the practical use of this method, the large-scale reaction was first conducted (Fig. 4a). Taking compound 53 as an example, when the reaction was performed on a 3 mmol scale, the desired product (23) was still obtained with moderate efficiency. This product could subsequently be treated with Lawesson’s reagent to deliver the corresponding phosphine sulfide (54) in 83% yield. Additionally, the products of this deoxygenative phosphonylation reaction can be directly reduced to give tri-substituted phosphine compounds, which serve as highly valuable monophosphine ligands and some instances were listed in Fig. 4b. Furthermore, with this protocol, the dialdehyde compound 58 can be smoothly converted to product 59 and, after simple reduction and stabilization with borane, compound 60 with two phosphine coordination sites can be obtained (Fig. 4c). The presence of the C–Br bond in this compound offers a convenient pathway for synthesizing various tridentate metal catalysts, which exhibit unusual catalytic capability to achieve a range of challenging transformations.

Fig. 4: Reaction application.
figure 4

a Large-scale reaction. b The synthesis of monophosphine ligands. c The synthesis of tridentate phosphine ligands.

Full size image

Mechanistic studies

To elucidate the reaction mechanism, we performed a series of mechanistic experiments (Fig. 5). Upon employing DPPh2 (72% D), a total deuterium incorporation at the benzylic position of the product (62) in the reaction with aromatic aldehyde was observed (Fig. 5a). However, the reaction with aliphatic aldehyde exhibited deuterium incorporation not only at the α position of the phosphorus atom but also at the β position (64). A similar outcome was detected with ketone substrates (see supplementary information). Additionally, using an aldehyde (65) with an α-chiral carbon center resulted in the formation of a racemic phosphine oxide under standard conditions (Fig. 5b). Remarkably, even the recovered starting material underwent complete racemization within a short time. Further experiments revealed that, while the racemization process becomes less pronounced in the presence of HPPh2 at room temperature—likely due to the complexation between HPPh2 and BF332, as confirmed by B and P NMR spectra (see supplementary information), which weakens the enolization process—BF3 can dissociate at higher temperatures and continues to promote enolization. These comparation experiments at ambient temperature indicated that racemization occurs independently of the target product formation, indicating a highly possible enolization process that only needs the used Lewis acid. In parallel, we synthesized the deuterated substrate 63-D, and in the presence of compound 67, no significant intermolecular deuterium exchange was observed (Fig. 5b). This may be attributed to HPPh2, which, after participating in the enol exchange, rapidly undergoes intramolecular addition to the carbonyl group, preventing it from escaping the solvent cage (Fig. 5c). After the proton exchange with DPPh2, the corresponding D-labeled aldehyde substrates are in-situ formed that subsequently deliver products with β-D incorporation.

Fig. 5: Mechanistic studies.
figure 5

a Deuterated experiment. b Enolization process. c H/D exchange mechanism. d 16O/18O exchange experiment. e The proposed reaction mechanism.

Full size image

To probe bond cleavage and formation during this rearrangement reaction, we synthesized an 18O-labeled substrate 63-O (Fig. 5d). Surprisingly, although the oxygen in the product (64-O) is indeed derived from the carbonyl oxygen atom, we observed an unexpected crossover of oxygen atoms when two aldehydes were present simultaneously—a process that, to the best of our knowledge, has never been demonstrated before. This also proves that the reaction does not proceed via the traditionally proposed epoxide or the oxirene intermediate that we initially believed, thus providing clear insights into the mechanism of this reaction. Based on these findings, we propose that after the addition of HPPh2 to the carbonyls, the lone pair of the trivalent phosphine could then attack the C–O bond, which is activated by the Lewis acid, as previously hypothesized. However, contrary to the proposed three-membered ring transition state, the C–O bond undergoes complete cleavage, allowing it to escape the solvent cage and enabling the crossover reaction (Fig. 5e). The BF3-combined OH anion further interacts with the phosphonium ion and ultimately forms the target product via an enolization process.

In summary, by altering the valency of used phosphorous skeletons in traditional [1,2]-phospha-Brook rearrangements, we have developed a highly efficient metal-free deoxygenative phosphonylation of carbonyl compounds. The key to success lies in shifting from a Brønsted base-mediated approach to a Lewis acid-mediated pathway, resulting in a transformative change of the intramolecular electron transfer process. This transformation enables the selective break of the C–O bond rather than the conventional C–P bond. The numerous distinct advantages of this method make it a promising and practical route for the synthesis of tertiary phosphine oxide compounds. Detailed mechanistic studies have evidently demonstrated the mechanism underlying this Lewis acid-mediated trivalent phospha-Brook rearrangement. Through the vital 16O/18O exchange experiment, the unusual oxygen atom crossover negates the traditionally assumed epoxide or the aforementioned oxirene intermediates; instead, a phosphorus-facilitated α C–O bond cleavage process is proven to be involved. The current work represents another significant and innovative advancement of the classic Brook rearrangement, deepens our understanding of trivalent phosphorus chemistry, and expands the boundaries of research in this field.

Methods

Model procedure

In the nitrogen-filled glovebox, an oven-dried 10 mL Schlenk tube with a magnetic bar was charged with carbonyl substrate (0.2 mmol), diphenylphosphine (2.5 equiv. for ketones and 2 equiv. for aldehydes), and anhydrous toluene (1.0 mL). Then the tube was sealed and removed from the glovebox. BF3·Et2O (2 equiv.) was added to the Schlenk tube under a nitrogen atmosphere. The mixture was heated to 120 oC (for ketones, and 100 oC for aldehydes) and stirred for 12 h. Upon completion, the reaction mixture was quenched by saturated NH4Cl aqueous solution and extracted with EtOAc (2 mL x 3). The combined organic layers were dried over MgSO4, filtered, and concentrated under vacuo. The residues were purified by silica gel column chromatography with a gradient eluent of EtOAc:PE:MeOH to afford the target product.