Compounds with transition metal–carbon triple bonds, namely metal carbyne complexes, have attracted much interest and attention since the seminal work by Fischer et al.1,2,3. The M≡C bonds have been shown rich reactivity with electrophiles and nucleophiles4,5,6,7,8. Therefore, metal carbyne complexes are valuable in synthetic chemistry as precursors for interesting organometallic compounds9,10,11, and can be used as homogeneous catalysts in the synthesis of organic compounds and polymers12,13,14,15,16,17,18. Alkynes containing C≡C bonds are also a class of compounds with triple bonds. They are used as versatile synthetic intermediates and are ubiquitous among complex organic molecules and polymers with important applications19,20,21,22,23,24,25.

The reactions of M≡C and C≡C bonds tend to afford cycloaddition intermediates or products on account of their high degree of unsaturation. For example, the alkyne metathesis undergoes metallacyclobutadiene intermediates process12,13,14, and [2+2], [2+2+1], and [2+2+2] cycloaddition products have been widely published26,27,28,29,30,31,32,33,34,35,36,37. Acyclic products, however, have never been reported. Therefore, construction of acyclic compounds from compounds with these two highly unsaturated triple bonds remains great challenge.

Herein, we report a reaction between metal carbynes and terminal alkynes in the presence of acid, which gives rise to acyclic addition products. This kind of reaction is highly efficient because of the extreme angle strain associated with the small carbyne bond angle in the cyclic metal carbynes. Meanwhile, these reactions show regio- and stereoselectivity, delivering exclusively trans products and anti-Markovnikov addition reactions. In contrast to common organometallic reactions that are sensitive to air and temperature, metal carbynes possess good functional group tolerance for alkynes and are robust in air and under ambient conditions.

The M≡C bond in this work is cooperative in metalla-aromatic systems in which the transition metal d orbits and the carbon p orbitals overlap to form dπpπ conjugated systems38. By taking advantage of this kind of reaction, the pπpπ conjugation of the organic moiety and the dπpπ conjugation of organometallics can be linked in a single conjugated system, which endows these unique acyclic addition products with potential applications especially in the field of optoelectronics. Because of the electron transport feature enhanced by osmium, these dπpπ conjugated complexes have been proved to be feasible as electron transport layers (ETLs) to enhance device performance. For example, PM6:Y6-based non-fullerene solar cells with complex 30 as an ETL have acquired a power conversion efficiency (PCE) as high as 16.28%. This value is not only much higher than that in devices lacking ETL, but is also dramatically elevated compared with commonly used ETLs, such as the PDINO discussed below. This reveals that the dπpπ conjugated complexes have great potential as interfacial layer materials, especially when combined with good electron transport capability and solubility in alcohol to facile orthogonal solvent processing, which could open a window into the capabilities of these newly developed dπpπ conjugated complexes.


Reactions of alkynes with osmapentalynes

The osmapentalyne reactants such as 1a (Fig. 1a) were synthesized and characterized (See Supplementary Figs. 4043)39. A solution of the complex (1a) was treated with HCl·Et2O solution in dichloromethane (DCM), producing a carbyne triple bond shifted product (2a) (Fig. 1a). The clear nuclear magnetic resonance (NMR) signal changes accompany this process (Supplementary Figs. 6063). And the structures of 1a and 2a were confirmed by X-ray diffraction. As shown in Fig. 1b, the Os1≡C1 triple bond length is 1.840 Å and the Os1–C7 single bond length is 2.061 Å in 1a. In 2a in contrast, the Os1≡C7 triple bond length is 1.855 Å and the Os1–C7 single bond length is 2.029 Å. These data indicate a tautomeric shift of the Os≡C triple bond from the initial ring to a different five-membered ring after the addition of acid. The similar tautomerization of the π-system from 1a to 2a has been discussed in detail in our previous work40. The Os1–C7–C6 metal-carbyne angle in 2a is 128.6°, smaller than the corresponding angle in 1a (130.9°). Both of them deviate by nearly 50° from the ideal sp-hybridized carbon angle (180°). The high ring strain associated with this small carbyne angle promoted us to investigate the reactivity of 1a and 2a with alkynes. In our previous work, only cycloaddition reactions based on similar osmapentalyne complexes were reported27,30,31.

Fig. 1: Reactions of alkynes with osmapentalynes.
figure 1

a Reaction of alkynes and osmapentalynes and the reaction of carbyne triple bond shifted reaction. b X-ray molecular structure for the cations of complex 1a, complex 2a, and complex 3. Ellipsoids are at the 50% probability level; phenyl groups in PPh3 are omitted for clarity.

In this work, we discovered that the osmapentalyne (2a) reacts readily with phenylacetylene at room temperature (RT) in the presence of excess HCl·Et2O, forming an acyclic addition complex (3) (Fig. 1). In view of the shifted carbyne triple bond results described above, we used complex 1a as a reactant in an attempt to simplify the operation. As expected, the same product (3) was obtained within 5 min in 99% isolated yield, showing that complex 1a can be a good precursor in this type of reaction. It is worth noting that the Os≡C triple bond of complex 2a can be slowly oxidized in air and the reaction is improved in an inert atmosphere. In contrast, the Os≡C triple bond of precursor 1a is more stable because of protection from the bulky triphenylphosphonium substituents. With precursor 1a as a reactant, the reactions can be carried out smoothly on gram scale in air and with commercial solvents such as dichloromethane, acetonitrile or 1,2-dichloroethane. This is different from general organometallic reactions that require oxygen-free and anhydrous conditions. Therefore, we used complex 1a as the reactant in the following investigations.

The structure of complex 3 was confirmed by X-ray single-crystal diffraction. As shown in Fig. 1b, it is an anti-Markovnikov addition product. Two building blocks, metallacycle and styrene, are connected by a C7–C12 bond of length 1.449 Å. The C12–C13 bond length of 1.357 Å suggests that it is a typical C=C double bond. The angles of C7–C12–C13 and C12–C13–C14 are 123.4° and 127.8°, respectively. The carbon–carbon bond lengths within the metallacycles (1.390–1.402 Å) are intermediate between typical single- and double-bond lengths. The Os1–C1 bond length (1.844 Å) in 3 is almost the same as the metal-carbyne bond length (1.840 Å) in complex 1a, and the Os1–C1–C2 metal-carbyne angle is 130.0°.

Complex 3 was also characterized by NMR spectroscopy, elemental analysis (EA) and high-resolution mass spectrometry (HRMS) (Supplementary Figs. 6871). Two singlets at 4.81 and -0.18 ppm, observed in the 31P NMR spectrum in CD2Cl2 can be assigned to CPPh3 and OsPPh3, respectively. In 1H NMR spectrum, the two singlets at 7.44 and 5.87 ppm are assigned to C12H and C13H, respectively. The coupling constant of the newly formed HC12=C13H group is 16.90 Hz, confirming the E-isomer configuration41. No resonance peaks associated with branched and Z-isomers were observed, indicating the regio- and stereospecifiity of this reaction. The 13C NMR spectrum in CD2Cl2 showed a signal from the carbyne carbon C1 at 316.2 ppm, and a signal from C7 at 218.8 ppm. Additionally, HRMS showed a peak at m/z 1347.3621, consistent with the theoretical structure of [C79H63ClOOsP3]+(1347.3387).

Mechanism study

To gain further insight into these reactions, we designed isotope labeling NMR experiments (Fig. 2a) in which DCl·Et2O was used instead of HCl·Et2O. We found that the hydrogen on the trans-olefins formed (C13H) was from the added acid. Deuterated phenylacetylene was also tested in the reaction, and as expected, the hydrogen atom of terminal alkyne of phenylacetylene was not involved in the reaction. The identifications of 3-d1 and 3-d2 were supported by NMR spectroscopy and HRMS (Supplementary Figs. 204211).

Fig. 2: Mechanistic studies.
figure 2

(a) isotope labeling NMR experiments and (b) verification experiment.

The carbene mechanism was proposed for the formation of a η3-vinylcarbene ligand via the reaction between the alkyne and the protonated carbyne ligand42. We designed a verification experiment to determine whether this reaction follows the same mechanism. In our previous work, we demonstrated the conversion of the Os≡C bond in osmapentalyne to an Os=C bond upon addition of the strong acid, HBF4·Et2O40. Herein, we prepared the osmapentalene (37) containing an Os=C bond using the same method (Fig. 2b). The structure of the complex (37) was further confirmed by X-ray diffraction (Supplementary Fig. 7). However, treatment of 37 with phenylacetylene failed to produce complex 31′. Indeed, 37 cannot be reacted with phenylacetylene. Therefore, a carbene intermediate in this reaction mechanism is excluded.

The participation of terminal alkynes in catalytic alkyne metathesis has been reported recently43,44. Then, we questioned if this kind of reaction involves a [2+2] cycloadditon intermediate which is the most common type of reaction between an Os≡C bond and a C≡C bond. We prepared the [2+2] cycloadditon complex (38) by reacting complex 2b with phenylacetylene without any added acid (Fig. 2b). However, complex 38 is not converted to complex 31 notwithstanding addition of HCl·Et2O or any other acid, indicating that the reactions fail to follow the [2+2] cycloaddition route.

Following these results, DFT calculations were carried out to shed light on the mechanism of these reactions. The computed Gibbs free energy profile of the key reaction steps is shown in Fig. 3. First, under the synergistic effect of acid, the Os≡C bond of osmapentalyne (2a) reacts with the C≡C bond of phenylacetylene to form the Z-isomer intermediate (Int1), in which the metal center has only 16 electrons. This process has an energy barrier of 18.8 kcal mol−1 and is exergonic by 47.6 kcal mol−1. Secondly, the styryl group coordinates with the osmium center in an η2 manner to form Int2, and subsequently converted into Int3 with an energy barrier of 10.8 kcal mol−1. Notably, the computed η2-complexes (Int2, Int3) are resemblance to η2-complexes of ruthenium, which showed as key intermediates of various unconventional addition reactions to alkynes45. Subsequently, the chloride anion approaches the hydrogen at C1 to form an E-isomer (Int4), further reducing the energy. Finally, Int4 can easily lose the proton on C1 to form the osmapentalyne (3). This last step, from Int4 to 3 has an energy barrier of 4.5 kcal mol−1 and is exergonic by 4.6 kcal mol−1. In the process, the metal center returns to a stabilized 18 electron structure. We have reported previously the possibility of conversion of a 16-electron osmapentaene to an 18-electron osmapentalyne40. The reaction is exremely fast and efficient and the 16-electron osmapentaene intermediate was undetectable. Protons from the acid play a particularly important role and the common reaction of Os≡C bond with the C≡C bond shows very unique results.

Fig. 3: Gibbs free energy profile for the DFT-calculated possible mechanism.
figure 3

The computed free energies are in kcal mol−1.

Substrate scope

To investigate the functional group tolerance of this kind of reaction, a broad species of alkynes and osmapentalyne complexes were investigated. With regard to the alkyne scope (Fig. 4), a series of substrates with either electron-withdrawing or electron-donating groups at different positions of their phenyl rings, perform well in this reaction, affording 3-21 with 90–99% isolated yields. All these reactions are essentially quantitative according to NMR data, and the slightly lower yields were mainly the result of purification steps necessitated by the highly polar triphenylphosphine substituents in the products. Common substituents, such as halogen (5-11), terminal olefin (12), methoxylcarbonyl (13), carboxyl (14), hydroxyphenyl (15), nitro (16), methoxyl (17), free amino (18), cyano (19), formyl (20), and pinacolborato (21) are all compatible. In addition to substituted phenyl groups, commercially available alkynes containing heterocycles such as thiophene (22), benzo[d]thiazole (23), pyridine (24), pyrimidine (25), quinoline (26) and imidazo[1,2-b]pyridazine (27) also provide the desired products in good isolated yields (85–99%). For non-aromatic substituted alkynes, we found only methyl propiolate (28) and propiolic acid (29) can undergo the reaction efficiently (See Supplementary Figs. 168175).

Fig. 4: Substrate scope.
figure 4

The substrate scope of the alkynes and osmapentalynes.

Concerning the scope of the osmapentalyne (Fig. 4), a series of reactant osmapentalynes (1b-1g) were synthesized by the treatment of carbon ligands (L2-L7) with multiyne chains with OsCl2(PPh3)3 and PPh3, the different substituents at different positions appear not to affect the reaction (3, 31-36). For instance, the R′ substituent can be styryl (3), phenyl (31-33), thienyl (34) or hydrogen (35-46), and the Y group can be CH2 (31, 34), O (32, 35) or C(COOMe)2 (33, 36). The structures of all the above complexes (3-36) were confirmed by NMR spectroscopy, EA and HRMS (Supplementary Figs. 68203). Some of the structures, such as 15 and 29 were further confirmed by X-ray diffraction (Supplementary Figs. 4 and 5). Based on these results, we were able to conclude that these reactions adapt well to substituents and tolerate functional groups well.

Construction of large dπpπ conjugated systems

Significantly, the Os≡C bond described here cooperates in dπpπ conjugation metalla-aromatic systems in which the transition metal d orbits and the carbon p orbits have good overlap38. Thanks to the efficiency of these reactions, the pπpπ conjugated organic systems and dπpπ conjugated metallacycle systems can be easily connected through C–C bond coupling, and as a result, the dπpπ conjugated systems could be further extended. Furthermore, the good group tolerance of this kind of reaction facilitates the molecular design of target products.

To prepare large π-systems containing dπpπ conjugation, we treated the osmapentalyne complex (1a) with one equivalent of 1-ethynylpyrene in DCM, adding HCl·Et2O for 15 min at RT in air (Fig. 5a). As expected, the complex (30) was isolated in 95% yield. The structure of 30 was characterized by NMR spectroscopy, EA and HRMS (Supplementary Figs. 176179). The single-crystal X-ray diffraction of 30′, a derivative of 30 with BF4 as the counter anion, clearly revealed the acyclic geometry structure that is analogous to complex 30 (Fig. 5a).

Fig. 5: Construction of large dπpπ conjugated systems.
figure 5

a X-ray molecular structure for the cation of complex 30′. Ellipsoids are at the 50% probability level, phenyl groups in PPh3 are omitted for clarity. b UV–Vis absorption spectra of phenylacetylene, 1-ethynylpyrene, complex 1a, complex 3 and complex 30 measured in DCM at RT (1.0 × 10-5 M). c Selected orbitals related to the excitation of complex 3 (isovalue = 0.02).

The UV/Vis absorption spectra can reflect the extent of π-conjugation. As shown in Fig. 5b, no absorption peak in the region over 600 nm is detected from either the reactant aromatic alkynes or the osmapentalyne (1a). In contrast, the products 3 and 30 show strong absorption in the low-energy absorption region. In comparation with the absorption spectra of reactant 1a and product 3, the characteristic energy absorption band of 3 is redshifted from 465 nm to 572 nm with the extended π-conjugated framework. The molar absorption coefficient of product 3 at 572 nm is as high as 4.3 × 104 M−1 cm−1 while that of the reactant 1a is only 1.0 × 104 M−1 cm−1 in the corresponding low-energy absorption region. With further extended dp π-conjugation, the characteristic energy absorption band of 30 is further redshifted compared to complex 3. It is noteworthy that complex 30 demonstrate broad and strong absorption in the entire visible region. Therefore, this kind of reaction provides a powerful method with which to construct large π-systems containing dπpπ conjugation.

To gain insight into these intriguing absorption spectra, we carried out time-dependent DFT (TD-DFT) calculations on the cationic part of 3 (Supplementary Table 7). The λmax = 557 nm of 3 is 96.7% ascribed to HOMO → LUMO transitions. This is a clear dπpπ π-conjugation system demonstrated by the simulated distribution of HOMO and LUMO orbitals (Fig. 5c). There are strong π-delocalization between the d orbitals of Os atom and p orbitals of carbon atoms. What is more, the other occupied molecular orbitals (HOMO-1 and HOMO-2, shown in Supplementary Fig. 14) also reflect the dp π-electron delocalization. The C = C double bond that is formed combines the pπpπ conjugated benzene or pyrene with the dπpπ conjugated aromatic metallacycles to form an extended π system containing dπpπ conjugation. We also carried out the TD-DFT calculation on the cation of 30, and found its λmax is also 98.5% ascribed to HOMO → LUMO transitions. Both the theoretical spectra of complexes 3 and 30 are consistent with the data from the experimental spectra. The essential reason of the redshifted characteristic energy absorption band is due to the decrease of HOMO–LUMO energy gaps (Eg). From DFT calculations, the Eg of compounds 1a, 3, and 30 are 2.93, 2.44 and 2.14 eV, respectively (Supplementary Table 8). Furthermore, the decrease of Eg with further extended dp π-conjugation is mainly due to the introduction of substituents with doner properties that increase the HOMO energy level of the system.

Applications in organic solar cells (OSCs)

The π-conjugated molecules have been widely utilized for interface engineering of OSCs46,47,48,49,50. With such unique dπpπ conjugated systems in hand, we explored the performance of these systems as the ETL in OSCs. Devices were fabricated with a conventional structure of ITO/PEDOT:PSS/PM6:Y6/ETLs/Ag (Fig. 6). The photovoltaic performance of devices was studied by employing common interfacial material (PDINO)51, and the newly synthesized complexes 3 and 30 as ETLs.

Fig. 6: Applications in organic solar cells.
figure 6

a Conventional structure of photovoltaic devices and chemical structures of (b) various molecules used as electron-transport layers (ETLs). c Polymer donor PM6. d Non-fullerene acceptor Y6.

As demonstrated in Table 1, the performance of devices based on 3 and 30 were both better than that of the common used ETL (PDINO). A control device with PDINO as ETL without thermal treatment exhibited a PCE of 15.06% which is comparable to the reported value52,53. With the extended dπpπ π-conjugation, the fill factor (FF) assumed a subsequently increasing trend, and the optimal PCE of OSCs based on 3 and 30 as ETLs were dramatically improved to 15.76% and 16.28% respectively, with enhanced FF and short-circuit current density (Jsc). To further identify the interface function of these complexes and exclude the effect from the alcohol solvent54, the performance of devices without ETL or using neat ethanol as the ETL was also investigated (Supplementary Table 3). Both JV curves (Supplementary Fig. 8) displayed lower performance with PCEs of only 10.47% and 11.02%, respectively. At the same time, compared to the performance without ETL, the champion efficiency based on 30 as ETL had an increase of 55%. In addition, to verify the generality of the application of complex 30 in organic solar cells we also tested the results based on the system of PTB7-Th:PC71BM (Supplementary Table 5) and PTB7-Th:IEICO-4F (Supplementary Table 6). Both PTB7-Th:PC71BM-based device and PTB7-Th:IEICO-4F based device still showed an enhancement of about 10% PCE compared to the PCE of controlled devices. All the JV curves and external quantum efficiency (EQE) spectra were consistent (Figs. 7a, b, Supplementary Figs. 810), and the light response of the device using 30 as ETL was strengthened, thus the Jsc was enhanced correspondingly. Therefore, complexes 3 and 30 were demonstrated as effective ETLs with orthogonal alcohol solvent in both fullerene and non-fullerene solar cells to further pump the solar conversions.

Table 1 Photovoltaic parameters of PM6:Y6-based OSCs with different ETLs.
Fig. 7: The characteristics of organic solar cells devices.
figure 7

The (a) JV and (b) EQE curves of OSCs based on PM6:Y6 with different ETLs respectively under 100 mW cm-2 AM 1.5 G irradiation. c The UV–Vis absorption spectra of blend film PM6:Y6 (d) JphVeff curves. e JV curves of electron-only devices. f Jsc vs Light intensity curves based on PDINO and complex 30, respectively.

To fully understand the effect of different ETLs on carrier transportation, devices using 30 and PDINO as ETLs were studied in parallel. According to the mechanism of OSCs, the normalized absorption spectra of active layer PM6:Y6 spin-coated on 30 and PDINO were tested, respectively (Fig. 7c), revealing negligible change in the photon absorbance. For the JphVeff curves plotted in Fig. 7d, the P(E, T) defined as Jph/Jsat of a device based on complex 30 increased to 0.98 while the increase with the PDINO-based device was only 0.94, indicating a higher probability of exciton dissociation in the presence of complex 30. Subsequently, the mobilities of electron-only devices were also calculated in terms of the space-charge-limited current SCLC (Fig. 7e, Supplementary Table 4). Based on complex 30, the electron mobility e) was ~5.7×10−4 cm2 V−1 S−1, which is more than three times the PDINO (μe = 1.7×10-4 cm2 V−1 S−1). Meanwhile, dependent on a function of Jsc PS55, the fitting slope in Fig. 7f of devices treated by complex 30 was 0.96, which is also higher than that of a PDINO-based device (0.94). These results indicate that device based on complex 30 will inhibit the bimolecular recombination, and eventually will demonstrate a superior device performance than a device based on PDINO. Therefore, the dπpπ π-conjugation systems formed by this newly discovered reaction, as an ETL material, could effectively improve the quality of OSCs, since more efficient carrier transport could be achieved with decreasing carrier recombination and improving carrier mobility.

In summary, we have demonstrated a reaction of M≡C and C≡C bonds that yields acyclic addition products. Based on our experimental observations, together with theoretical calculations, we propose the mechanism of this kind of reaction is the electrophilic addition of carbyne by the alkyne under the synergistic effect of H+. These reactions are highly efficient, regio- and stereospecific, with good functional group tolerance, and are robust under air at RT. Taking advantage of this reaction, two different π-conjugated systems, for example, organic and organometallic systems, can be easily connected through this reaction. The resulting extended π-conjugation systems demonstrate unusually broad and strong absorption in the ultraviolet/visible region. Moreover, they supported impressive applications as ETL materials in OSCs. Specifically, the champion PCE based on complex 30 was enhanced to 16.28%, benefitting from improved carrier transport and restrained carrier recombination as a result of the extended π conjugation. Other properties and applications of such unique dπpπ conjugated systems are under further study. In view of the high efficiency under mild reaction conditions and facile experimental operation, this efficient methodology opens significant opportunities for preparation of dπpπ conjugated systems for use in functional materials.


General information

All syntheses of were performed without using standard Schlenk techniques unless otherwise stated. Other reagents were used as received from Aldrich Chemical Co. Column chromatography was performed on alumina gel (200–300 mesh), silica gel (200–300 mesh) or polystyrene gel (Bio-Beads TMS-X3 Support, 200-400 mesh) in air. NMR spectra were recorded on a Bruker AVIII-500 (1H, 500.2 MHz; 13C, 125.8 MHz; 31P, 202.5 MHz) spectrometer, a Bruker Ascend III 600 (1H, 600.1 MHz; 13C, 150.9; 31P, 242.9 MHz) spectrometer at RT, or a Bruker AVIII-400 (1H, 400.1 MHz; 13C, 100.6 MHz; 31P, 162.0 MHz) spectrometer at RT. 1H and 13C NMR chemical shifts (δ) are relative to tetramethylsilane, and 31P NMR chemical shifts are relative to 85% H3PO4. The absolute values of the coupling constants are given in Hertz (Hz). Multiplicities are abbreviated as singlet (s), doublet (d), triplet (t), multiplet (m), and broad (br). High-resolution mass spectra (HRMS) experiments were recorded on a Bruker En Apex Ultra 7.0 T FT-MS. The theoretical molecular ion peak was calculated by Compass Isotope Pattern software supplied by Bruker Co. Elemental analyses were performed on a Vario EL III elemental analyzer. Absorption spectra were recorded on a SHIMADZU UV2550 ultraviolet–visible spectrophotometer. Details for the synthesis and characterization of all the above complexes are given in the Supplementary Information.

X-ray crystallographic analysis

All single-crystal X-ray diffraction data were collected on a XtaLAB Synergy, Dualflex, HyPix diffractometer with Cu Kα radiation (λ = 1.54184 Å). With Olex256, all the structures were solved using the ShelXT57 structure solution program using the intrinsic phasing method, and all the structures were refined with the ShelXL58 refinement package using least-squares minimization. Non-H atoms were refined anisotropically unless otherwise stated. Hydrogen atoms were introduced at their geometric positions and refined as riding atoms unless otherwise stated. All single crystals suitable for X-ray diffraction were grown from a solution of CH2Cl2 layered with hexane. Further details on the crystal data, data collection, and refinements are provided in Supplementary Table 1, Supplementary Table 2, and Supplementary Table 3.

Computational methods

All structures were optimized at the B3LYP59,60,61 level of functional theory. Frequency calculations were performed to confirm the characteristics of all the calculated structures as minima. All these structures evaluated were optimized at the B3LYP/6–31G*/ED = GD3BJ level of DFT with an SDD basis62 set to describe Os atoms; single-point energy calculations were then performed on the mechanism using the B3LYP/Def2-TZVP method with the SMD solvation method63 in DCM. Whereas the UV-Vis-NIR spectrum was used TD-DFT calculations64 at the B3LYP/6–31G* level of DFT with an SMD solvation model in DCM; an SDD basis set to describe Os atoms. In all calculations, the effective core potentials (ECPs) reported by Hay and Wadt with polarization functions were added for Os (ζ(f) = 0.886)65. All the optimizations were performed with the Gaussian 09 software package66.

Experimental methods and characterization of OSCs

Y6, PTB7-Th was synthesized in our laboratory. PM6 was purchased from the solar material company while PC71BM was purchased from the 1-Material Company. First, the ITO substrates were cleaned by ultrasonication. Second, after drying in an oven, these substrates were placed in UV-Ozone for 15 min. Third, a layer of PEDOT:PSS was spin-coated on the surface of ITO at 3000 rpm for 20 s. After annealing the substrates at 150 °C for 15 min and then cooling down, the active layer was prepared by spin-coating PM6:Y6 with D/A = 1:1.1 ratio in chloroform (CF) at 6 mg ml−1 with additives 0.5% (v/v) chloronaphthalene (CN). Then the PDINO was spin-coated, complexes 3 and 30 were dissolved in ethanol at 1 mg ml−1, at 3000 rpm for 20 s on the active layer as ETLs and subsequently by evaporating the Ag (100 nm) electrode. Through a 0.045 cm2 mask, thermal evaporation in a vacuum chamber requires a pressure of ~3×10-6 Pa. All the operations were performed in a nitrogen glove box.