Design principleWe initiated our scientific study on the activation of B–H bonds and the annulation process using nido-carboranyl phosphine (1a) and diphenyl acetylene (2a) as model substrates (Fig. 2a). Following a thorough screening process, we obtained the optimized conditions: 10 mol% of Pd(OAc)2 as catalyst and DCE as solvent under an argon atmosphere for a duration of 6 hours at a temperature of 35 oC. These curated conditions led to a yield of 83% of desired product of 3aa. Notably, the 11B NMR spectra of 3aa exhibited a diagnosable resonance with an evident downfield shift (δ = 4.2 ppm), indicating the formation of an exopolyhedral B–C bond. This proposed structure for 3aa was further confirmed by single-crystal X-ray diffraction (SC-XRD) analysis. When we explored the reaction with alternative transition metal catalysts such as Co(OAc)2, [Cp*IrCl2]2, or [RuCl2(p-cymene)]2, no product was generated. Furthermore, different bases or raising the temperature to 60 °C failed to increase yield. The indispensability of DCE as the solvent was underscored by the diminishing yields of 3aa when alternative solvents like THF and DCM were used owing to the solubility issue.Fig. 2: Reaction discovery and mechanistic studies.a General reaction conditions: 1a (0.1 mmol, 1.0 equiv.), 2a (0.12 mmol, 1.2 equiv.), Pd(OAc)2 (10 mol%), DCE (2.0 mL), 35 oC, 6 h, Ar atmosphere, yield of isolated product. b HRMS detection of the palladacyclic intermediate A; synthesis and isolation of B(9) palladation intermediate B; verification of reactivity of B toward alkyne. c Proposed reaction mechanism. The hydrogen atoms including B–H–B bridging hydride are omitted for clarity.Mechanistic investigationOnce the optimized reaction conditions were established, we conducted mechanistic studies to elucidate the reaction pathway (Fig. 2b). To detect the formation of a Pd–B intermediate, we initially used the catalyst Pd(PPh3)2Cl2 instead of Pd(OAc)2 to slow down the reaction. Such a strategy only led to the observation of a diagnostic HRMS signal for an intermediate A, which could further react with alkyne to yield 3aa. In order to isolate such a crucial intermediate, we designed a stoichiometric reaction by mixing 1b, Pd(OAc)2 and the chelating ligand of 1,10-phenanthroline. Fortunately, a yellow complex B was isolated in 70% yield, and characterized by SC-XRD, multinuclear NMR, and HRMS. The crystal structure revealed the formation of a type of four-membered Pd–B–C–P palladacycle, generated by P(III)-directed palladium-mediated B–H activation at B(9) site. The chelating coordination of 1,10-phenanthroline assists in stabilizing such a structure and makes this provable species isolatable. Complex B exhibits the further transformation toward alkyne to lead to 3ba in 37% yield. This observation implies that a four-membered-ring palladacyclic species should be an intermediate. It is worth noting that previous studies have demonstrated the B–H functionalization of carborane via the formation of a kinetically favored five- or six-membered cyclometalated intermediate with a transition metal40,41,42,46,47,48,49,50. Here, we obtained the crystal structure of a rare four-membered Pd–B–C–P cyclometalated intermediate for verification of both B–H activation and its subsequent transformation. This finding provides insights into both understanding the reaction mechanism and the design of B–H activation modes for carborane functionalization.According to the above mechanistic studies and literature40,41,42,46,47,48,49,50,54,55,56,57, a plausible mechanism was proposed (Fig. 2c). The Pd(OAc)2 catalyst first binds to 1a and then leads to B(9)–H activation with the assistance of P(III)-directing to liberate the key intermediate I, which contains a cyclometalated Pd–B–C–P four-membered ring, similar to the detected or isolated intermediate A or B. Subsequently, ligand exchange between acetic acid and diphenyl acetylene occurs, yielding intermediate II. Next, followed by alkyne insertion into the Pd–B bond, intermediate III is inclined to undergo reductive elimination to give rise to the desired product 3aa and Pd(0) species. Lastly, the oxidation of Pd(0) by HOAc affords catalytic species Pd(OAc)2 and H2, as confirmed by GC analysis (Supplementary Fig. 50).Scope studyNext, we set out to expand this catalytic strategy to the exploration of the variety of carboranyl phosphines with alkyne 2a (Fig. 3a). The introduction of aryl (3ba) or alkyl (3ca) moiety to the other carbon site of the carborane cage led to comparable outcomes with 3aa. Furthermore, the group of PR2 anchored at one carbon site of the carborane cage proved to be crucial in driving this transformation. For example, cyclohexyl (3da) and isopropyl (3ea) could provide the desired products in moderate yields. However, the reaction hardly occurred when tBu was employed owing to the large obstruction imposed by PtBu2. Nevertheless, the situation was released by replacing one of the tBu groups with a phenyl group (3ga). Notably, the introduction of an electron-donating group, such as a methoxy substituent, to the phenyl group (3 ha) greatly enhanced the efficiency of the reaction, leading to a yield as high as 91%. Conversely, when an electron-withdrawing group, specifically –CF3 (3ia) was introduced, a moderate level of yield (53%) was generated. These observations demonstrate the susceptibility of the alkyne cycle-closing process to the electronic and steric factors at the phosphorus site. Furthermore, the methodology was expanded to different types of alkyne substrates, encompassing a multitude of diverse substituents, generating the corresponding boron cluster-extended phosphoniums with impressive yields (Fig. 3b). Diarylacetylene substrates carrying electron-donating groups on the aryl rings, such as –CH3 (3ab), –OCH3 (3ac), –tBu (3ad), –NH2 (3ae) and –Ph2N (3af), gave rise to higher yields (82–96%). However, the reaction was severely influenced by the steric environment of the alkyne, as evidenced by the fact that diarylacetylene bearing a methyl group in the ortho-position yielded only a moderate yield of 3ag. The diarylacetylene hosting multiple methyl substituents (3ah) is compatible with this method. Interestingly, substrates featuring fluoro groups in ortho-, meta-, and para-positions of the aryl alkyne, respectively, exhibited satisfactory performance (3ai, 3aj, and 3ak) in this reaction, yielding the desired products in good yields (68–78%). The structure of compound 3ai was further confirmed through SC-XRD analysis. Moreover, if substrates bear electron-withdrawing groups in the para position, including –CN (3ao), –NO2 (3ap), and –CF3 (3aq), we were delighted to isolate the desired products in yields ranging from 66 to 79% (3al–3aq). It is worth noting that substrates with reactive functional groups, such as –NH2 (3ae, 95%) and –CHO (3an, 76%), exhibited good tolerance. These intriguing results manifest the influence of the electronic and steric nature of the alkyne on the reactivity.Fig. 3: Substrate scope.General reaction conditions: 1 (0.1 mmol, 1.0 equiv.), 2 (0.12 mmol, 1.2 equiv.), Pd(OAc)2 (10 mol%), DCE (2.0 ml), 35 °C, 6 h, Ar atmosphere, yield of isolated product. aReaction time is 10 h. The hydrogen atoms of crystal structures are omitted for clarity. a Scope of carboranyl phosphines. b Scope of alkyne symmetrically substrates. c Scope of non-symmetrically disubstituted alkynes.We conducted further investigations on alkynes bearing polycyclic and heterocyclic substrates, which also led to satisfactory yields (64–75%, 4aa–4ae). However, the alkyne featuring two strongly coordinated pyridinyl groups (4af) gave rise to a modest yield, which should be ascribed to the disturbance of pyridinyl group on catalyst. The dialkyl-substituted internal alkynes such as hex-3-yne, oct-4-yne, and 1,4-dimethoxybut-2-yne exhibited moderate reactivity, producing acceptable yields of 66% (4ag), 63% (4ah), and 70% (4ai), respectively. In an effort to expand the repertoire of boron cluster-extended phosphoniums, we delved into the realm of non-symmetrically disubstituted alkynes (Fig. 3c). We employed the phenylpropyl alkyne carrying a comparable steric effect for methyl and phenyl. As anticipated, two isomers, 5aa1 and 5aa2, were obtained with a ratio of 3:5, and their structures were also confirmed by SC-XRD analysis (Supplementary Figs. 5 and 6). Alkynes containing two groups with distinct steric hindrance such as trimethylsilyl and phenyl resulted in the formation of two isomers, namely 5ab1 and 5ab2, with a ratio of (5:1). The larger trimethylsilyl group displays a tendency to position itself far away from the phosphonium unit. The precise structures of 5ab1 and 5ab2 were confirmed by SC-XRD analysis, aligning with the proposed structural formulation (Supplementary Figs. 7 and 8). In addition, we also examined electron-rich acetylenes (5ac1–5ad2), which reacted and yielded two isomeric products. Interestingly, terminal alkynes featuring aryl (5ae–5ag) and triisopropylsilyl (5ah) groups showed exclusive regio-selectivity in this annulation event owing to the sensitivity of annulation on steric hindrance.Synthetic applicationsThe compatibility of the reaction prompted us to explore its practicality (Fig. 4). Firstly, we incorporated photo-functional groups into the carboranyl phosphine (1a) by employing this synthetic strategy. Materials science has witnessed great interest in carbazole-based systems possessing pure π bonding structures owing to their one-of-a-kind light-emitting properties. To obtain carborane-based conjugates, the alkyne substrate 4,4’-bis(N-carbazoly)tolan (BCT) was used to lead to 6aa in a yield of 82%. Notably, BCT suffered from the ACQ effect, however, the conjugate 6aa exhibited distinct AIE property and showed high luminous efficiency of an absolute quantum yield of 70% in the solid state (Supplementary Figs. 19–21). Based on the same synthetic strategy, the acridine group was introduced to the carborane-based phosphine (1a) to generate two isomers (6ab1 and 6ab2). Secondly, we extended this methodology to the late-stage modification of drug molecules, such as Erlotinib, a small-molecule epidermal growth factor receptor tyrosine kinase inhibitor. Under the standard conditions, Erlotinib could directly react with 1a to afford the exclusive drug-modified molecule 6ac, which may be potential in the targeting drug for lung cancer on the basis of boron neutron capture therapy (BNCT)58,59. This protocol was also applicable to the modification of the drug tazarotene to produce 6ad1 and 6ad2. To our delight, dibenzocyclooctyne-COOH (DBCO-COOH), a classical reagent studied in biorthogonal chemistry, could be coupled with 1a to generate 6ae in a yield of 61%. Notably, 6ae is not only an instance bearing a large pentacyclic conjugation but also provides the possibility for further applications in materials and bioconjugation for BNCT owing to the reserved reactive carboxylic acid. Moreover, when 1a reacted with a polyalkyne such as 1,4-bis(phenylethynyl)benzene, 6af was produced containing two carborane-fused heterocycles, whose crystallographic structure was confirmed by SC-XRD. Thus, this protocol has indicated valuable application potential in materials and drug development.Fig. 4: The modification of functional molecules.General reaction conditions: 1a (0.1 mmol, 1.0 equiv.), 2 (0.12 mmol, 1.2 equiv.), Pd(OAc)2 (10 mol%), DCE (2.0 mL), 35 °C, 6 h, Ar atmosphere, yield of isolated product.Structural studyNext, we investigated the structural characteristics of the boron cluster-fused phosphoniums as they represent a type of framework bearing a B, P-containing heterocycle. Accordingly, the crystal structures of 3aa and the control compounds nido-K and Model-1 were carefully studied (Fig. 5a, b). 3aa shows a folded molecular structure with a dihedral angle of 35.07 (6)o between the generated five-membered ring (plane I) and the C2B3 plane of nido-carborane (plane II). In sharp contrast, Model-1 shows a nearly planar bicyclic structure with a dihedral angle of 2.05 (5)o. Such a large difference could be attributed to the  3D steric effect of the carborane cage. Furthermore, the carborane fusion also contributes to the structural deformation in comparison to Model-1 and nido-K. The size of the five-membered ring P1C1C2B9C8 in 3aa is larger than the corresponding P1C1C2C3C4 ring in Model-1, as demonstrated by the selected bond lengths (Fig. 5b). For example, B9–C8 (1.616 (3) Å) in 3aa is much longer than C3–C4 (1.398 (2) Å) in Model-1, and C2–B9 (1.576 (3) Å) in 3aa vs C2–C3 (1.447 (2) Å) in Model-1 is longer, which could be interpreted by the larger atomic radius of boron than carbon. In the case of planes II, in general, the deformation in 3aa vs nido-K is much less.Fig. 5: Structural investigations.a Crystal structures of 3aa, nido-K, and Model-1. Insets are the selected dihedral angles between planes I and II. b The selected bond lengths for planes I and the planes II of 3aa and the control compounds. The estimated standard deviation values are shown in parentheses. The H atoms in crystal structures are omitted for clarity. c Left, Comparison of 3aa and Model-1 enclosing geometries with NICS (top value, in ppm) and MCI (bottom value, in au); Middle, VDD charges (in a.u.); and right, molecular electrostatic potential isosurfaces (electronic density isovalue = 0.03 a.u.). d The chemical stability comparison of 3aa and the control compounds.To further understand the electronic structure of the fused system, DFT computations at the ZORA-BLYP-D3(BJ)/TZ2P were performed on 3aa, 1a, and Model-1 (see Supplementary information for additional computational details). The charges were evaluated with the Voronoi Deformation Density (VDD) method, but other methods gave the same trends (Supplementary Fig. 58). Firstly, the computed equilibrium geometries are in good agreement with the X-ray data (Fig. 5c and Supplementary Fig. 56). The carborane cage keeps its aromaticity when fused to the five-membered heterocycle (magnetic-based NICS in the center of the cage is –22.0 for 1a and –19.8 ppm for 3aa, respectively). The aromaticity of the cage is further confirmed by Wade-Mingos’ rule with a total of 26 valence electrons, thus following the 2·(11 + 2) rule for a nido cluster with 11 vertices. At difference, the B, P-heterocycle in 3aa is non-aromatic as indicated by both NICS (–1.9 ppm) and electronic-based MCI (0.001 a.u.), due to the localization of the 2π-electrons in the C–C double bond in this ring, with the P atom not allowing for the delocalization (Fig. 5c and Supplementary Fig. 52). Next, evaluation of the charges shows an important charge transfer from carborane to the heterocycle in 3aa (0.512 e) compared to Model-1 (0.170 e). Despite the charge transfer, the P atom remains positively charged, in part due to the electron-withdrawing phenyl substituents (Supplementary Figs. 53–58). The charge transfer due to the formation of the fused ring can be observed through the electrostatic potential isosurfaces (Fig. 5c).Study of propertiesThen we examined the stability of the fused framework. 3aa shows a higher decomposition temperature of 323.6 °C according to thermogravimetric analysis in comparison to 273.6 °C for Model-1 (Supplementary Fig. 18). In particular, 3aa demonstrates satisfactory chemical stability towards acid or base in air (Fig. 5d and Supplementary Figs. 12–17). However, the 2D aromatic ring-based phosphoniums of Model-1 and Model-2 underwent facile ring-opening via C–P bond cleavage53. This is attributed to the more reduced electron density at P site in 3aa vs Model-1 (Fig. 5c). Therefore, the 3D aromatic nido-carborane enables to stabilize P-containing heterocycle relative to the corresponding 2D π-conjugated aromatic system.Now we turn to investigate the photophysical properties associated with such a fused framework. Firstly, the UV-Vis spectra reveal an absorption peak around 370 nm for 3aa in contrast to 1a (Fig. 6a). This should be attributed to the charge transfer transition, as demonstrated by the DFT calculations (Supplementary Fig. 59). Thus, the electronic communication between the 3D carborane and the B, P-heterocycle was further experimentally confirmed. The computed UV-Vis spectrum is also consistent with the measured one. In contrast, 3aa shows undetectable emission in solution in sharp contrast to the strong luminescence observed for Model-1 (Fig. 6b). This must be caused by the facile molecular motions in the excited state triggered by the folded structure in contrast to a rigid aromatic structure in Model-1, thus leading to severe non-radiative decay in solution7,60,61. To our delight, 3aa exhibits typical AIE effect. As shown in Fig. 6c, no emissions were observed when the water fractions (fw) for the THF/water mixtures were less than 70%. When fw was reached to 80%, a faint emission could be observed due to the initial formation of molecular aggregates. At fw = 90% the greenish-yellow luminescence showed a sharp rise (Fig. 6d). This turn-on emission behavior should be attributed to the restriction of molecular motion caused by molecular aggregation7,60,61. On the other hand, the bulky size of carborane and the twisted molecular structure of the framework can effectively inhibit the π···π interactions in the aggregate state, thus further improving the aggregate-state luminescence62,63,64,65,66,67. To quantify the AIE effect, the αAIE = ϕ (fw = 99%)/ϕ (fw = 0%) value was used by measuring the quantum yields in THF solution and 99% water fraction, respectively. As a result, the αAIE value reached 3000 for 3aa, surpassing many classical 2D π-conjugated AIE systems7,60,61. Other carborane-based phosphoniums in this study exhibited more conspicuous AIE performance (Fig. 6e, and Supplementary Figs. 22–27, 34–45), in particular, 5aa2 showed αAIE > 7000. Hence these AIE molecules might be useful in the design of advanced molecular systems for bioimaging applications. Moreover, in the solid state, the color-tunable emissions ranging from blue to red have been achieved in the framework by simply altering the electronic effect of the substituents on the alkyne aryl groups, demonstrating the electronic sensitivity of the core framework on periphery environment (Fig. 6f and Supplementary Figs. 28, 29). The solid-state luminescence efficiency could reach up to 70% (3ad), in sharp contrast to the faint emission in Model-1 (Supplementary Figs. 22, 30–33) and non-emissive 1a. The distinctive photophysical phenomena between the 3D aromatic boron cluster-based phosphoniums and the 2D aromatic π-conjugated phosphoniums (i.e., Model-1) are attributed to the unusual electronic and geometric structures of the framework. These properties may have promising potentials in photo-functional materials.Fig. 6: Photophysical properties of the selected compounds.a The UV/Vis absorption spectra in THF solution (c = 10 μM). b The photoluminescence (PL) spectra in THF solution (c = 10 μM, λex = 380 nm). c The emission intensity of 3aa in a THF/H2O mixture with increased H2O volume fraction (fw) to 99% (c = 10 μM, λex = 380 nm). d Relative emission intensity of 3aa. e The αAIE of 3aa, 3ac, 4ad, 5aa2, 3ba, 3ea and Model-1. f The solid-state luminescence of the selected products. Insets are the absolute luminescence efficiency, CIE (Commission Internationle de l’Eclairage) coordinates, and emission peaks. g The chiral crystal structures of 3aa-Rp and 3aa-Sp. h The CD spectra of 3aa-Rp and 3aa-Sp in THF. i The CPL spectra of 3aa-Rp and 3aa-Sp in PMMA with a weight ratio of 1:100.Moreover, carborane clusters are an unconventional source for chirality investigations, including both cage and planar chirality. Until now, boron cluster-based CPL molecules are rarely documented68,69. In this study, the carborane-fused phosphoniums provide a promising avenue to investigate boron cluster-based CPL molecules. Interestingly, we have successfully separated two chiral carborane-fused phosphoniums, 3aa-Rp and 3aa-Sp, through chiral high-performance liquid chromatography, as confirmed by SC-XRD analysis (Fig. 6g). Their CD spectra showed a mirror-image relationship with the alternating positive and negative Cotton effect (Fig. 6h). Surprisingly, no CPL was observed for 3aa-Rp and 3aa-Sp in solution (Supplementary Fig. 46). However, in a PMMA film they exhibited enhanced CPL with a photoluminescence dissymmetry factor |gPL| of 9.9 × 10–3 (Fig. 6i), indicating an aggregation-induced enlargeable CPL property. Such a high value of |gPL| represents impressive performance among the reported CPL-active organic small molecules. Additionally, we have observed an unexpected phenomenon of isomer-dependent luminescence, detailed in Supplementary Figs. 62–64. These results also indicate the potential to develop a type of luminescent materials using the 3D boron cluster-fused framework.We have demonstrated a couple-close synthetic strategy to obtain a type of 3D aromatic boron cluster-fused phosphoniums through B–H activation of carboranyl phosphines with alkynes. The dehydrogenative process avoids the use of extra oxidant during the reaction. Mechanistic studies have revealed a rare four-membered palladacyclic intermediate, which is distinct from the previously reported five/six-membered cyclometalated intermediates for B–H activation. The resulting type of conjugated framework shows high thermal and chemical stability in contrast to 2D aromatic ring-based analogs, in particular, interesting photophysical properties, such as unusual AIE effect with a αAIE value of up to 7000, readily tuned emission color spanning the entire visible light region and highly efficient solid-state emission with Φ up to 70%, as well as AICPL with a gPL value of 1.0 × 10–2. This study is devoted to developing synthetic strategies, gaining structural insights, and exploring valuable physicochemical properties of the 3D boron cluster-based heterocyclic systems. We hope these results would expand scopes of both boron clusters and heterocycles and demonstrate the application prospect in materials and drug development.
