Liyao
Ma
a,
Sonia
Mallet-Ladeira
b,
Julien
Monot
a,
Blanca
Martin-Vaca
*a and
Didier
Bourissou
*a
aCNRS/Université Paul Sabatier, Laboratoire Hétérochimie Fondamentale et Appliquée (LHFA, UMR 5069), 118 Route de Narbonne, 31062 Cedex 09 Toulouse, France. E-mail: blanca-maria.martin-vaca@univ-tlse3.fr; didier.bourissou@univ-tlse3.fr
bUniversité de Toulouse III Paul Sabatier, Institut de Chimie de Toulouse, ICT, UAR 2599, 118, route de Narbonne, F-31062 Toulouse, France
First published on 18th April 2024
Cycloisomerization reactions involving C–P bond formation have been overlooked for the synthesis of P-heterocycles. In this work, we developed a simple, efficient and versatile route to synthesize benzophosphole oxides by reacting ortho-alkynyl secondary phosphine oxides with 5 mol% AgSbF6. Mechanistic investigations revealed a radical-chain mechanism involving phosphinoyl radicals as key intermediates and rare 5-endo-dig cyclization as a key step, rather than the π-activation of the CC triple bond. The transformation is both efficient and versatile. It effectively complements alternative intermolecular approaches. It works with a wide diversity of substitution patterns (alkynyl, benzo and phosphorus moieties) and enables the exquisite control of regioselectivity. Post-functionalization via direct C–H vinylation of the C2 position is also substantiated.
In this regard, benzophosphole oxides (BPOs) are primary targets owing to their application in organic electronics, including light-emitting devices,3 photovoltaics4 and cell-imaging dyes.5 The cycloisomerization route has attracted much attention because of its selectivity and substrate scope, but surprisingly, it has only been much rarely considered thus far and remains largely underdeveloped. Accordingly, the preparation of BPOs through the cyclization of ortho-alkynyl secondary phosphine oxides (SPOs) has only been reported once under basic conditions.6 Typically, heating 1a at 70 °C for 24 hours in DMSO in the presence of tBuOK (20 mol%) was found to afford BPO 2a with 79% yield (Scheme 1). The reaction is simple to operate and does not require a TM-based catalyst, but it is limited in scope. It works only for substrates featuring an internal alkyne substituted by an aryl group.
![]() | ||
Scheme 1 Cycloisomerization of the (2-alkynylphenyl) phosphine oxide 1a into benzophosphole oxide 2a: unique precedent under tBuOK catalysis and the AgI-promoted route reported here. |
With the aim to apply and develop a TM-catalyzed cycloisomerization approach for the synthesis of P-heterocycles, such as 2a, we screened various complexes reported to be efficient in C–O, C–N and C–C bond-forming transformations (mainly Pd and Au complexes).7 As a result, we discovered that AgSbF6 alone efficiently promotes the cycloisomerization of 1a into 2a. This finding prompted us to in-depth investigate AgI-promoted cycloisomerization route to benzophosphole oxides and we hereafter discuss this transformation in terms of reaction optimization, mechanistic investigations, scope, comparison with alternative methods and post-functionalization.
Standard conditions | Conva (%) | Yielda (%) | |
---|---|---|---|
>96 | >96 (91)b | ||
Deviation from standard conditions | |||
a Estimated by 31P NMR spectroscopy. b Isolated yield in parentheses. c Yield in phosphaisocoumarin 3a in parentheses.7 | |||
Reaction conditions | 120 °C, 2 h | >96 | 93 |
20 mol% AgSbF6, 4 h | >96 | >96 | |
2 mol% AgSbF6, 0.8 M, 20 h | 95 | 80 | |
100 mol% AgSbF6 | >96 | 27 (49)c | |
NO AgSbF6 | 9 | 7 | |
Under air | 87 | 67 (15)c | |
Silver salt | 5 mol% AgNTf2, 9 h | 51 | 49 |
5 mol% AgOTf, 9 h | 61 | 60 | |
5 mol% AgBF4, 9 h | 44 | 44 | |
5 mol% AgNO3, 9 h | 100 | 20 | |
5 mol% AgOAc, 72 h | 100 | 20 | |
5 mol% Ag2CO3, 72 h | 100 | 20 | |
Solvent | Benzene, 9 h | >96 | >96 |
t BuPh, 9 h | 85 | 83 | |
1,2-Dichlorobenzene, 22 h | >96 | 84 | |
CH3CN, 33 h | >96 | 86 | |
DMF, 9 h | 61 | 48 | |
DCE, 47 h | >96 | 82 |
Changing toluene for more polar solvents such as 1,2-dichlorobenzene, DCE, DMF or CH3CN had no benefit, rather the opposite, while similar results were obtained with benzene. Using the optimized reaction conditions (AgSbF6 5 mol%, toluene, 80 °C), the reaction was then scaled up to 10 mmol of 1a, operating at 0.8 M to reduce both the time and the quantity of solvent. The transformation was complete in 5 hours, and 2a was obtained in 90% isolated yield (2.72 g).7
Given the state-of-the-art and literature precedent, two mechanistic scenarios can be a priori envisioned for the AgI-promoted cycloisomerization of 1a. On the one hand, AgI salts are known to activate alkynes via π-coordination and to promote the addition of pro-nucleophiles to the CC triple bond (Fig. 1a).8,9 Conversely, AgI salts may act as oxidants towards secondary phosphine oxides R2P(O)H [and phosphonates (RO)2P(O)H] to generate phosphinoyl radicals, which can then undergo radical addition to alkynes (Fig. 1b).8,10 On the other hand, the AgI salt is used in stoichiometric amount or an excess of oxidant is added to regenerate AgIin situ.11
To try to distinguish between these two paths, a series of experiments were performed. First, we assessed the impact of additives that may foster the “π-activation” route (Fig. 2a), i.e. weak bases such as tBu2Py, K2CO3 or Et3N (10 mol%) to activate the pro-nucleophile, PPh3 to stabilize AgI,12 and hydrogen-bond donors such as Ph2P(O)OH, C6H3(OH)3 or HFIP to favor protodemetalation upon H-shuttling.13 In most cases, the conversion of 1a was significantly lowered. Complete consumption was only observed with HFIP as an additive, but the benzophosphole oxide 2a was obtained in low yield (11%). Conversely, the concomitant use of AgSbF6 and TEMPO (5 mol% each) drastically reduced the conversion of 1a (15%) and the yield in 2a (10%) (Fig. 2b). Furthermore, the addition of 5 mol% of TEMPO after 4 hours of reaction under standard conditions considerably slowed down further transformation (the conversion of 1a stopped at 70–80% conversion and the yield in 2a did not exceed 65–70% after 5 additional hours, Fig. S3‡).7 These experiments favor the radical pathway over the π-activation route. To further substantiate the formation of the phosphinoyl radical A upon oxidation of 1a with AgI, spin trapping with a nitrone (DMPO) was performed (Fig. 2c).7 The reaction was conducted in tBuPh with 15 mol% of AgSbF6. The mixture was stirred at 80 °C for 1 hour to initiate the reaction, and then cooled to room temperature (to prevent direct reaction of 1a with the nitrone)14 before the addition of DMPO (15 mol%). ESR analysis showed the formation of a nitroxide radical (the pattern is very similar to that reported for the trapping of the phosphinyl radical Ph2P(O)˙ by DMPO and to that we obtained ourselves by reacting Ph2P(O)H instead of 1a with AgSbF6, Fig. S15‡).7,15–17 This experiment further supports the radical pathway as the operating mechanism in the AgI-promoted cycloisomerization of 1a.
![]() | ||
Fig. 2 Experiments performed to discriminate the two mechanistic paths, namely π-activation and radical addition. |
Based on the gathered information, we propose the radical-chain mechanism displayed in Fig. 3 to account for the cycloisomerization of 1a into 2a promoted by AgSbF6. The initiation would involve the generation of the phosphinoyl radical A upon oxidation of the λ5-P(O)H/λ3-P(OH) moiety of 1a/1a′ by the AgI cation. The propagation phase would then involve cyclization of Avia intramolecular 5-endo-dig radical addition to the CC triple bond, to give the vinyl radical B. Finally, hydrogen atom transfer (HAT) would deliver the benzophosphole oxide 2a. To decipher the H atom source, we resorted to D-labeling experiments using either 1a–D deuterated at the P atom or toluene-D8 as the solvent (Fig. 2d).7 Inspection of the 1H NMR signal for the vinylic
C–H of 2a showed that the SPO substrate indeed acts as a H donor towards B, enabling chain propagation. The absence of deuterium incorporation in 2a when operating in toluene-D8 indicates that the solvent does not take part in the HAT, in line with the similar results observed using toluene or benzene as the solvent.
![]() | ||
Fig. 3 Proposed radical-chain mechanism to account for the AgI-promoted cycloisomerization of 1a into 2a. |
Of note, radical 5-endo-dig cyclization with P(O)-centered radicals has not been previously reported to the best of our knowledge. With other types of radicals, this kind of cyclization is challenging and rare, but not unprecedented. Experimental evidence was first reported with a Si-centered radical,18 then with C-centered radicals19 and more recently with N-20 and Ge-centered radicals.21
Following mechanistic studies, we assessed the scope of the transformation with respect to the substitution pattern of alkynyl, benzo and phosphorus moieties (Fig. 4). The bifunctional substrates were prepared in few steps from ortho-bromo, iodo-benzene derivatives upon sequential introduction of alkyne and SPO moieties (by the Sonogashira coupling and ionic coupling with a dichlorophosphine followed by hydrolysis, respectively).7 Electron-enriched alkynes with a para-Me or para-OMe phenyl substituent (1b,c), an alkyl substituent (nBu, 1d), and a silyl group (SiEt3, 1e), were cyclized more rapidly than 1a and the corresponding benzophosphole oxides were obtained in high yields (in particular 2b–2d, 81–95%). As for the silyl-substituted BPO 2e, it was obtained in the mixture with the desilylated benzophosphole oxide 2o (2/1 ratio) under the standard conditions (5 mol% AgSbF6), but reducing the AgI loading to 2 mol% enabled us to increase the isolated yield of 2e to 72%. Me and MeO substitution of the ortho and meta positions of the phenyl group (1f–1h) resulted in longer reaction times without impacting the yields (92–99%). The introduction of electron-withdrawing groups at the phenyl substituent of the alkyne moiety required longer reaction times than 1a (15–23 hours) without compromising the efficiency of the cycloisomerization (79–97% yields for 2i–2l). Compound 2l was actually characterized by single-crystal X-ray diffraction analysis,7 unambiguously confirming the benzophosphole oxide structure and the C2-substitution. The cyclization reaction also worked well with substrates bearing π-conjugated alkene or heterocyclic substituents at the alkyne moiety, such as cyclohexene or thiophene, as substantiated by the formation of 2m and 2n with 73–74% yield. A longer reaction time was required to cyclize the terminal alkyne 1o (20 hours), and the corresponding BPO 2o was obtained in a modest yield (40%) probably due to parasitic reactions between the C–H bond and the silver salt.22 However, we leveraged on the desilylation process observed with 1e to develop an alternative route to 2o. Installing a trimethylsilyl group at the alkyne (substrate 1p) and using 50 mol% of AgI salt, the parent benzophosphole oxide 2o was formed as very major product within only 1 hour and it could be isolated in 85% yield.23 Of note, efficient preparation of 2o is a challenge and only a few precedents exist, as recalled in Scheme 2. Desilylation of 2p with TBAF affords 2o in only 50% yield.24 Better results were obtained by ring-closing metathesis of a divinyl precursor with an Hoveyda–Grubbs second-generation catalyst25a or by direct cyclization of Ph2P(O)H with acetylene promoted by CuCl2 in the presence of an excess of tert-butyl peroxobenzoate (TBPB) as an oxidant.25b
Substitution of the phenyl ring linking the SPO and alkynyl moieties was then investigated (substrates 1q–t). In all cases, the corresponding benzophosphole oxides were obtained in high yields (89–97%). Here, the cycloisomerization approach inherently provides precise control of the BPO structure. The reaction proceeds at the phenyl ring bearing the alkynyl moiety and gives a single regioisomer, in contrast to the intermolecular variant involving Ar2P(O)H secondary phosphine oxides and alkynes (see below). No significant electronic bias was observed between electron-donating and -withdrawing substituents, the reaction times to obtain 2q (Me), 2r (OMe) and 2s (CF3) (13 hours) being essentially identical (12–15 hours). Interestingly, Me substitution of the position ortho to P proved more impactful. The formation of 2t proceeded faster (3 hours), which may be due to some buttressing effects favoring the 5-endo-dig cyclization. Finally, variation of the P substituent was explored. Introducing the electron-withdrawing group 3,5-(CF3)2Ph (1u) made no noticeable difference with the reference substrate 1a in terms of reaction time and yield of the obtained BPO (9 hours, 97%). Comparatively, electron-enriched substrates proved more difficult to cyclize. For the iPr-substituted substrate 1v, 33 hours were required to achieve full conversion, but the yield of 2v was not compromised (96%). Phosphinates (RO)2P(O)H are more challenging substrates than SPO in oxidative radical couplings.16 Consistently, harsher conditions were required to cyclize the ethoxy-substituted substrate 1w (120 °C, 16 h), but the corresponding BPO 2w was nonetheless obtained in 52% isolated yield.26
Overall, the cycloisomerization methodology reported here allows for the efficient preparation of a wide structural variety of benzophosphole oxides with electron-donating/electron-withdrawing substituents at C2, the benzo ring and/or the P atom (17 examples, 88% average yield). This transformation is versatile and complementary to the alternative intermolecular routes developed over the last decade (Fig. 5 and S7‡).7 The most studied route involves the dehydrogenative coupling of diaryl SPO and internal alkynes (Fig. 5a).17,27–29 It requires an oxidant (typically AgI, MnIII or K2S2O8) in stoichiometric amount or excess. Greener variants have been uncovered recently using an organic photocatalyst and a pyridinium salt as an oxidant,17 or even simply dioxygen.30 Another strategy relies on a one-pot multicomponent reaction involving Co/Ni-catalyzed migratory carbometallation of alkynes, Cu-catalyzed C–P coupling and phosphorus oxidation (Fig. 5b).31 A third method is based on the electrophilic annulation of SPO with internal alkynes in the presence of an excess of Tf2O and a base (Fig. 5c).32 All these routes use internal alkynes and thus give C2/C3-disubstituted benzophosphole oxides. Moreover, symmetric internal alkynes are largely preferred to prevent the formation of regioisomeric mixtures. The same limitation applies to the diaryl SPO substrates used in the first and third strategies, and symmetric SPOs are used routinely to prevent selectivity issues in the cyclization step. It is worth noting that the cycloisomerization approach reported here requires the preparation of alkynyl-SPO substrate 1, but it inherently proceeds with complete selectivity and it circumvents the formation of BPO mixtures. This is nicely illustrated by the selective formation of compounds 2q–t, where related intermolecular transformations suffered from the randomization of the regiochemistry of the “benzo” fragment (Fig. S8‡).7,17,27,29,32,33
![]() | ||
Fig. 5 Comparison of the main synthetic routes developed to access benzophosphole oxides: known intermolecular strategies (a–c) versus the cycloisomerization approach reported here (d). |
As mentioned above, another attractive feature of the AgI-promoted cycloisomerization route is to provide efficient access to the parent benzophosphole oxide 2o. Given the recent progress achieved in the post-functionalization of BPO26,34–37 we wondered about the possibility to derivatize 2o by C–H activation. In particular, we became interested in the installation of vinyl groups at C2 since it is challenging by other means. One option is to achieve Pd-catalyzed Mizoroki–Heck or Stille cross-coupling from the BPO bearing a bromine atom at C2 (Fig. S9‡).7,34
More attractive synthetically is the Pd-catalyzed and Ag-assisted C–H vinylation reported recently by Hirano et al. from the BPO 2x (obtained by electrophilic coupling of 1,1-diphenylethylene and phenylphosphinic acid) (Scheme 3).35,36 It was observed that under similar conditions, the parent BPO 2o was fully consumed but a complicated mixture of products was obtained. Intrigued by the influence of the Ph group at C3 on this transformation, we tested the functionalization of 2o under the same conditions (10 mol% Pd(OAc)2, 2 equiv. AgTFA, 2 equiv. NaHCO3, dioxane, 110 °C) using para-methyl styrene as the partner. 31P NMR monitoring indeed showed consumption of 2o but the reaction leveled off at 40% conversion after 7 hours. To drive complete conversion, more Pd(OAc)2 (10 mol%) and AgTFA (2 equivalents) were added and gratifyingly, the C2-vinylated BPO 2oa was thereby obtained in 90% isolated yield.7 These forcing reaction conditions were then applied to para-methoxy styrene and ethyl acrylate, affording the corresponding C2-functionalized BPO 2ob and 2oc in good yields (94 and 73%, respectively).
Besides the specific preparation of BPO, these results point out the synthetic potential of C–P bond-forming cycloisomerization reactions to access P-heterocycles. This strategy is illustrated here in a radical transformation, but ionic as well as TM-catalyzed variants can certainly be conceived and are worth investigating. This work also highlights the synthetic potential of silver salts. Long neglected compared to other TM, AgI species have found increasing applications as π-activators, halide abstractors38 as well as oxidants.39 In this work, AgSbF6 was used as a radical initiator, in catalytic amounts and without any external oxidant.
Footnotes |
† Dedicated to Prof. Denis Curran on the occasion of his 70th birthday. |
‡ Electronic supplementary information (ESI) available: Full experimental procedures, characterization data of all new compounds (including 1H and 13C-NMR spectra, ESR spectra and XRD data. CCDC 2332940. For ESI and crystallographic data in CIF or other electronic format see DOI: https://doi.org/10.1039/d4qo00552j |
This journal is © the Partner Organisations 2024 |