Inclusion co-crystals based on Tröger’s base molecular box for thermally activated delayed fluorescence Authors: Yuan Wang, Kun Liu, Ya Li, Zhe Zheng, Yuna Shang, Bin Li, Hongming He, Chunju Li Published: 2026 ABSTRACT Designing thermally activated delayed fluorescence (TADF) emitters via a non-covalent through-space charge transfer (TSCT) is challenging, requiring a delicate balance between molecular geometries and electronic states. Herein, we present an inclusion co-crystal (ICC) strategy to achieve efficient TADF using non-covalent TSCT approach by supramolecularly engineered molecular box-based confined environments that provide sufficient binding ability to aromatic guests. Tröger’s base-derived molecular boxes (TB[2]) incorporating mesomeric (meso-TB[2]) and racemic (race-TB[2]) species are designed as electron-rich hosts, which can selectively encapsulate 1,3,5-benzenetricarbonitrile (BTN) and 1,2,4,5-tetra-cyanobenzene (TCNB) among six electron-withdrawing cyanobenzenes, forming three 1:1 binary inclusion CT co-crystals through the synergism of multiple non-covalent interactions. The BTN@meso-TB[2] and BTN@race-TB[2] co-crystals exhibit tunable yellow-green and bright cyan TADF emission due to strong host-guest complexation and weak TSCT interactions. Conversely, TCNB@meso-TB[2] is almost no luminescent as the strong intermolecular CT interaction promotes nonradiative decay processes. The present work offers an efficient strategy for designing TADF emitters by balancing the strength of CT and host-guest complexation. Introduction Since the pioneering work by Adachi and co-workers in 20121, thermally activated delayed fluorescence (TADF) materials2, have attracted increasing concerns in the fields of luminescent probes3–6, photodynamic therapy7–10, photocatalysis11–13, optoelectronic devices14–16, X-ray scintillation and imaging17–19. TADF emitters are capable of up-converting triplet excitons through reverse intersystem crossing (RISC), theoretically enabling an internal quantum efficiency of 100%. In principle, efficient TADF is decided by a narrow singlet-triplet energy gap (ΔEST, <0.2 eV) to promote the RISC process1,20,21, resulting in higher exciton utilization. For this purpose, the design strategy most frequently utilized is to separate the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO) in a donor-acceptor (D-A) structure. Conventional strategy to realize TADF is to design the planar molecules by intramolecular through-bond charge transfer (TBCT), in which the D and A are connected directly via single bond, resulting in rapid and efficient charge transitions. However, the TBCT-based molecules usually cause a strong CT effect and red-shifted emission loss (Fig. 1a)22–27. Additionally, intramolecular through-space charge-transfer (TSCT) system is an alternative approach toward high-efficiency TADF emitters28–30, wherein the D and A are placed on neighboring positions of a rigid bridge with a short spatial distance to form highly twisted molecules (Fig. 1a). This structural design not only reduces HOMO-LUMO overlap, realizing small ΔEST, but also enhances the radiative decay rate to ensure a high photoluminescence quantum efficiency (PLQY). However, the both covalent design principles for TADF emitters generally suffer from complicated chemical syntheses and aggregation-caused quenching31,32.Fig. 1Molecular design strategies for TADF materials.a General strategies of through-bond CT, through-space CT and organic small-molecule co-crystals for D-A type TADF materials. b Macrocycle-based inclusion co-crystals for TADF in this work. Molecular design strategies for TADF materials. a General strategies of through-bond CT, through-space CT and organic small-molecule co-crystals for D-A type TADF materials. b Macrocycle-based inclusion co-crystals for TADF in this work. Organic co-crystal engineering offers promising low-cost strategy for building photo-functional materials through the non-covalent (e.g., hydrogen bonds, halogen bonds, π–π stacking) assembly of two or more conformer components33–35. Specifically, organic charge-transfer (CT) co-crystals constructed by electron-rich D and electron-deficient A have garnered significant research owing to their molecular arrangements and tunable excited-state energy levels between D-A moieties that benefit from generating photonic and electronic properties compared to the individual synergistic constituent36,37. However, the co-crystals with TADF property are seldom in literature, presumably due to the following reasons: (i) the lack of insights into commonly reported co-crystal design38,39; (ii) the co-crystal structures often built by planar and rigid co-formers are ineffective to meet the stringent TADF process; (iii) the weak D-A interactions and strong CT promote nonradiative transition processes40–42. In supramolecular chemistry, macrocycles possessing polygonal topological structures and preorganized cavities are principal tools for recognition and assembly43–48. Recently, macrocycle co-crystals (MCCs) have rapidly developed into a cutting-edge field for building solid-state functional materials49,50, such as luminescence51–54, photothermal conversion55–57, adsorption/separation58–61 and vapochromism62–64, exploiting the exo-wall CT interactions or endocavity inclusion complexation by the marriage of macrocyclic chemistry and organic co-crystal engineering. MCCs should provide a potential for TADF materials that narrows the singlet-triplet energy gap through TSCT and therefore facilitates RISC by cavity-mediated host-guest complexation. To date, only few successful examples for this approach were recently reported by Chen’s group65–67, Chou’s group68 and Würthner’s group42 because the CT interaction between the host and guest generally quenches the fluorescence during complex formation69–71, and the construction of efficient TADF emitters through non-covalent CT interactions requires a balance between molecular geometries, interaction force and CT strength. Hence, designing macrocycles to build intermolecular D-A pair-mediated TADF and study the superstructure-property relationships are still a challenge. In this regard, box-type macrocycles offer a particularly promising option. By incorporating two parallel aromatic subunits, they form sequestered supramolecular cavities that strongly bind planar aromatic guests via robust π···π and C‒H···π interactions (Fig. 1b). Herein, a type of molecular boxes (TB[2]) bearing phenyl-sidewall Tröger’s base (TB) skeleton were synthesized as the donor, which contains one meso-isomer (meso-TB[2]) and one pair of enantiomers (race-TB[2]) due to TB’s inherently C2-symmetric chirality. They can selectively bind electron-deficient 1,3,5-benzenetricarbonitrile (BTN) and 1,2,4,5-tetra-cyanobenzene (TCNB) from six cyanobenzenes by host-guest inclusion complexation in solution with the highest association constant (Ka) for BTN beyond 105 M−1 in non-polar solvent (CDCl3/CCl4, v: v, 1:1) and for TCNB less than 102 M−1. Three sets of CT inclusion co-crystals (ICCs) with 1:1 stoichiometric ratios in the solid state, named BTN@meso-TB[2], BTN@race-TB[2] and TCNB@meso-TB[2], were obtained driven by the synergism of multiple π···π and C–H···π/N forces between TB[2] and BTN/TCNB. Intriguingly, the BTN@meso-TB[2] and BTN@race-TB[2] co-crystals exhibit different yellow-green and bright cyan TADF emission with similar PLQY of 21.3 and 21.6%, respectively, due to strong host-guest complexation and relatively weak TSCT interactions. Theoretical calculations reveal that the TADF mechanism proceeds via a “hot exciton” process through high‑level reverse intersystem crossing (hRISC) from higher‑lying triplet states to the S1 state. However, TCNB@meso-TB[2] is almost no luminescent as the strong intermolecular CT interaction between host and guest quenches the fluorescence and promotes nonradiative decay processes. Results Synthesis and crystal structure of TB[2] For the synthesis of molecular box, phenyl-sidewall TB was selected as the macrocyclic skeleton owing to its π-electron rich character and rigid clip-like structure72–74. Two 2,4-dimethoxyphenyl moieties were attached to the ends of TB moiety, giving that they are highly effective reaction modules for cyclization with aldehydes under Lewis acid catalysis, as well demonstrated in our prior studies75–78. Since the right-angle framework of TB, dimeric macrocycles with box-like structures are anticipated to form. The molecular boxes are expected to comprise one meso-isomer (meso-TB[2]) and one pair of enantiomers (race-TB[2]) due to the inherent C2-symmetry chirality79–82. The detail synthetic processes of TB[2] are illustrated in Fig. 2. First, the monomer TBM was obtained through the Suzuki–Miyaura coupling reaction of dibromo-substituted TBBr precursor and 2,4-dimethoxybenzeneboronic acid in 97% yield. Then, the one-step macrocyclization reaction was realized by mixing TBM and paraformaldehyde in CH2Cl2 under the catalysis of trifluoroacetic acid at room temperature for 35 min, affording meso-TB[2] and race-TB[2] with total yields of 73%. Moreover, the enantiomers of race‑TB[2] were successfully resolved by preparative chiral HPLC, and their optical purities were confirmed by circular dichroism (CD) spectroscopy, which exhibited clear mirror‑image Cotton effects in the range of 200‒350 nm for the two enantiomers (Supplementary Fig. 15). All chemical structures were characterized by 1H and 13C nuclear magnetic resonance (NMR), spectroscopy 2D 1H−1H COSY spectroscopy, and high-resolution mass spectrometry (HRMS) (Supplementary Figs. 1‒14).Fig. 2The synthetic route of TB[2].The synthetic route of TB[2] showing one meso-isomer (meso-TB[2]) and one pair of enantiomers (race-TB[2]). The synthetic route of TB[2]. The synthetic route of TB[2] showing one meso-isomer (meso-TB[2]) and one pair of enantiomers (race-TB[2]). Single crystals of meso-TB[2] and race-TB[2] suitable for X-ray crystallography were successfully obtained by slow evaporation solution growth in CH3CN. The resulting single-crystal structure showed that they crystallize in the triclinic P−1 space group (Supplementary Tables 1 and 2). As expected, the meso-TB[2] displays a rigid and box-like cavity with a size of approximately 10.7 × 6.0 Å, which is suitable for accommodating one small electron-deficient guest via π···π and C–H···π interactions (Fig. 3a). The dihedral angles (82.0°) of the two TB skeletons demonstrate the efficient stereoscopic cavity. The solid-state structure of race-TB[2] containing one pair of enantiomers (R2NR2N-TB[2] and S2NS2N-TB[2]) shows a twisted configuration because two opposite side dimethoxybenzenes are facing the same direction (Fig. 3b). In their crystal structures, on average four CH3CN molecules are observed for one host molecule, among which two CH3CN molecules are encapsulated inside the corners of TB clip with the same or different orientation by the C–H···π/N interactions and others reside in the lattice space by forming C–H···N hydrogen bonds with TB[2] (Fig. 3c, d and Supplementary Figs. 16 and 17). In addition, the facile formation of four solid-state inclusion complexes between meso-TB[2] and solvent molecules such as dichloromethane (DCM), 1,2-dichloroethane, chlorobenzene (CB) and 3-acetylpyridine (AP) by slow solution evaporation indicates its strong inclusion capability and crystallinity (Supplementary Tables 3–6 and Supplementary Figs. 18–21).Fig. 3Solid-state superstructures of meso-TB[2] and race-TB[2].Single-crystal structures (a, b) and stacking structures (c, d) of meso-TB[2] and race-TB[2]. Hydrogen atoms are omitted for clarity. All the above structure figures were prepared using PyMol (The PyMOL Molecular Graphics System, Version 1.8, Schrödinger, LLC). Solid-state superstructures of meso-TB[2] and race-TB[2]. Single-crystal structures (a, b) and stacking structures (c, d) of meso-TB[2] and race-TB[2]. Hydrogen atoms are omitted for clarity. All the above structure figures were prepared using PyMol (The PyMOL Molecular Graphics System, Version 1.8, Schrödinger, LLC). Host-guest complexation investigation The electron-rich and rigid cavity of TB[2] is ideal for binding planar guests. We selected six electron-withdrawing cyano-substituted benzenes (BZN, o-DCB, m-DCB, p-DCB, BTN, TCNB) differing in substituent number and position with appropriate molecular size to study the host-guest complexation with TB[2] in solution using ¹H NMR spectroscopy (Fig. 4a and Supplementary Figs. 22‒33). When physically mixing TB[2] and mono- or di-substituted cyanobenzene in CDCl3, no chemical shift changes were observed in the ¹H NMR spectra (Supplementary Figs. 22‒29), indicating weak binding in solution. We found that TB[2] exhibits significant binding affinity to tri-substituted cyanobenzene BTN. As depicted in Fig. 4, addition of meso-TB[2] or race-TB[2] (3.0 mM) to BTN (1.0 equiv) resulted in complete disappearance of BTN proton signals compared with free BTN (Supplementary Figs. 30 and 31). This observation confirms that guest molecules are engulfed inside the macrocycle cavity and suffers an inclusion-induced shielding effect. Simultaneously, the aromatic protons of TB[2] exhibited pronounced upfield/downfield shifts due to deshielding effects and phenyl-BTN π···π interactions. Notably, race-TB[2] exhibited much larger chemical shift alterations than meso-TB[2] when bound with BTN, suggesting stronger host-guest binding affinity. For comparison, the host-guest complexation of classic macrocyclic receptors (pillar[5]arene, calix[4]arene, calix[6]arene) with BTN was investigated (Supplementary Figs. 36‒38). 1H NMR spectra revealed no detectable inclusion complex formation, further validating the availability and size matching for the formation of box-type topology. For tetra-substituted cyanobenzene TCNB with larger electron-withdrawing ability, when mixing the colorless solution of meso-TB[2] or race-TB[2] and TCNB, obvious color changes to red were observed (inset pictures in Supplementary Figs. 63 and 64). The ultraviolet–visible (UV–Vis) spectroscopy demonstrated the existence of stronger CT interaction between TB[2] and TCNB than BTN (Supplementary Figs. 61 and 62). However, 1H NMR spectra displayed smaller chemical shift changes of TCNB and TB[2] than BTN (Supplementary Figs. 30‒33), indicating weaker host-guest interactions in solution.Fig. 41HNMR on host-guest complexation.a Chemical structures of TB[2] and six cyanobenzene guests. b Partial 1H NMR spectra (400 MHz, 298 K, CDCl3) of (Ⅰ) race-TB[2], (Ⅱ) race-TB[2] + BTN, (Ⅲ) BTN, (Ⅳ) meso-TB[2] + BTN, (Ⅴ) meso-TB[2] at 3.0 mM. 1HNMR on host-guest complexation. a Chemical structures of TB[2] and six cyanobenzene guests. b Partial 1H NMR spectra (400 MHz, 298 K, CDCl3) of (Ⅰ) race-TB[2], (Ⅱ) race-TB[2] + BTN, (Ⅲ) BTN, (Ⅳ) meso-TB[2] + BTN, (Ⅴ) meso-TB[2] at 3.0 mM. To quantify host-guest binding affinities, association constants (Ka) were measured via NMR titration (Supplementary Figs. 39‒60) and summarized in Table 1. Among these guests, the Ka values for mono- or di-substituted cyanobenzene of BZN and o-/m-/p-DCB with TB[2] are too small (<10 M‒1) to be determined. Although race-TB[2] shows a twisted conformation in the solid state, it displays the strongest binding ability for BTN with Ka value of (4.39 ± 0.58) × 103 M‒1 in CDCl3 (Supplementary Figs. 41 and 42), which is 8.8 times larger than that for meso-TB[2] [(5.01 ± 0.35) × 102 M−1] (Supplementary Figs. 39 and 40). This result implied that the binding BTN may induce the adaptive structural transformations of race-TB[2]. The meso-TB[2] and race-TB[2] have similar binding capacity of 67.44 ± 1.40 M‒1 and 59.98 ± 3.23 M‒1 for the stronger electron-deficient TCNB (Supplementary Figs. 57‒60). Solvent effects were then examined (Supplementary Figs. 34, 35, and 43‒56). The reduction of solvent polarity, by mixing non-polar CCl4, resulted in tremendous enhancement of binding affinities of one or two magnitude (Table 1, Entries 7‒13). For example, as the solvent polarity decreases from 8:1 to 2:1 (v:v) CDCl3/CCl4, the Ka value of BTN by race-TB[2] reaches (1.91 ± 0.49) × 104 M−1 and exceeds 105 M‒1 at the 1:1 (v:v) CDCl3/CCl4 ratio. The binding affinity is considerable high to neutral host-guest system in non-polar solvents.Table 1Association constants (Ka)a of TB[2] towards six cyano-substituted benzene moleculesEntryHostGuestSolventKa (M‒1)1meso-TB[2]BZN, DCBCDCl3b2race-TB[2]BZN, DCBCDCl3b3meso-TB[2]BTNCDCl3(5.01 ± 0.35) ×1024race-TB[2]BTNCDCl3(4.39 ± 0.58) ×1035meso-TB[2]TCNBCDCl367.44 ± 1.406race-TB[2]TCNBCDCl359.98 ± 3.237meso-TB[2]BTNCDCl3:CCl4, 2:1c(7.77 ± 0.96) × 1028meso-TB[2]BTNCDCl3:CCl4, 1:1c(3.94 ± 0.44) × 1039race-TB[2]BTNCDCl3:CCl4, 8:1c(7.00 ± 1.19) × 10310race-TB[2]BTNCDCl3:CCl4, 4:1c(6.89 ± 0.63) × 10311race-TB[2]BTNCDCl3:CCl4, 3:1c(9.63 ± 1.06) × 10312race-TB[2]BTNCDCl3:CCl4, 2:1c(1.91 ± 0.49) × 10413race-TB[2]BTNCDCl3:CCl4, 1:1c>105c14calix[4]areneBTNCDCl3b15calix[6]areneBTNCDCl3b16pillar[5]areneBTNCDCl3baKa values were determined by using 1H NMR titration methods.bHost-guest complexation was not observed.cThis value was too high (>105 M‒1) to be accurately determined by using 1H NMR titration methods. Table 1Association constants (Ka)a of TB[2] towards six cyano-substituted benzene moleculesEntryHostGuestSolventKa (M‒1)1meso-TB[2]BZN, DCBCDCl3b2race-TB[2]BZN, DCBCDCl3b3meso-TB[2]BTNCDCl3(5.01 ± 0.35) ×1024race-TB[2]BTNCDCl3(4.39 ± 0.58) ×1035meso-TB[2]TCNBCDCl367.44 ± 1.406race-TB[2]TCNBCDCl359.98 ± 3.237meso-TB[2]BTNCDCl3:CCl4, 2:1c(7.77 ± 0.96) × 1028meso-TB[2]BTNCDCl3:CCl4, 1:1c(3.94 ± 0.44) × 1039race-TB[2]BTNCDCl3:CCl4, 8:1c(7.00 ± 1.19) × 10310race-TB[2]BTNCDCl3:CCl4, 4:1c(6.89 ± 0.63) × 10311race-TB[2]BTNCDCl3:CCl4, 3:1c(9.63 ± 1.06) × 10312race-TB[2]BTNCDCl3:CCl4, 2:1c(1.91 ± 0.49) × 10413race-TB[2]BTNCDCl3:CCl4, 1:1c>105c14calix[4]areneBTNCDCl3b15calix[6]areneBTNCDCl3b16pillar[5]areneBTNCDCl3baKa values were determined by using 1H NMR titration methods.bHost-guest complexation was not observed.cThis value was too high (>105 M‒1) to be accurately determined by using 1H NMR titration methods. Association constants (Ka)a of TB[2] towards six cyano-substituted benzene molecules aKa values were determined by using 1H NMR titration methods. bHost-guest complexation was not observed. cThis value was too high (>105 M‒1) to be accurately determined by using 1H NMR titration methods. Solid-state superstructures of ICCs To gain structural insight into the guest encapsulation within the host cavity, three ICCs (BTN@meso-TB[2], BTN@race-TB[2] and TCNB@meso-TB[2]) were obtained by cocrystallizing TB[2] with BTN or TCNB through slow vapor diffusion of isopropyl ether into a CHCl3 solution (Supplementary Tables 7‒9). We found that the yellow co-crystals of BTN@meso-TB[2] and BTN@race-TB[2] emit strong yellow-green and bright cyan fluorescence under UV light, respectively, in contrast to TB[2] powders with wathet blue emission (Fig. 5a). However, the co-crystals of TCNB@meso-TB[2] display dark red, only weak fluorescence emission was observed, indicating the existence of strong CT interaction between TCNB and meso-TB[2] quenches the fluorescence of host-guest complex (Fig. 5a). X-ray crystallographic data indicated they crystallize in the monoclinic C2/c (BTN@meso-TB[2]), monoclinic P21/n (BTN@race-TB[2]) and triclinic P-1 (TCNB@meso-TB[2]) space group. In the two crystal structures of BTN@TB[2], BTN molecules are encapsulated inside the host cavity forming 1:1 ICCs through the π···π interactions [3.70–4.83 Å] between aromatic rings and BTN (Fig. 5b, c). Furthermore, some typical C–H···π/N/O interactions between BTN and TB[2] further stabilize the inclusion complex (Supplementary Figs. 65 and 68). We hypothesize that such multipoint constraints are beneficial for the reduction of vibrational dissipation in the photoexcited state. In addition, guest encapsulation induces the formation of a pair of enantiomers in BTN@meso-TB[2] complex (Supplementary Fig. 66), accompanied by significant structural changes due to the flip of dimethoxybenzenes compared to the solid-state structure of meso-TB[2] alone. For BTN@race-TB[2], guest binding endows structure regulation of race-TB[2] from its initially twisted conformation to rigidity (Fig. 5c and Supplementary Figs. 68 and 69). The results prove that the molecular box is adaptive after binding guest. A similar inclusion complex was observed for TCNB guest, and no enantiomers were found. In the structure of TCNB@meso-TB[2], only four weak π···π interactions [4.19–4.83 Å] and multiple C–H···N contacts (2.71–2.96 Å) were observed between meso-TB[2] and TCNB (Fig. 5d and Supplementary Fig. 71), which is in agreement with the complexation in solution. The propagation of these non-covalent interactions in the three ICCs in a 2D plane resulted in the formation of three different host-guest 2D tiling superstructures (Fig. 5b–d and Supplementary Figs. 67, 70, and 72). These noncovalent interactions greatly stabilize the host-guest spatial arrangement, facilitating intermolecular CT state that induces a dramatic alternation in both crystal color and fluorescence emission.Fig. 5Structural analysis of host-guest co-crystals.a Optical images of meso-TB[2], race-TB[2], BTN@meso-TB[2], BTN@race-TB[2] and TCNB@meso-TB[2] under sunlight (up) and UV light (down), which displays tunable fluorescence changes. b‒d Single crystal structures of BTN@meso-TB[2], BTN@race-TB[2] and TCNB@meso-TB[2]. The dashed line represents π⋯π distances (Å). Hydrogen atoms and solvents are omitted for clarity. All the above structure figures were prepared using PyMol (The PyMOL Molecular Graphics System, Version 1.8, Schrödinger, LLC). Structural analysis of host-guest co-crystals. a Optical images of meso-TB[2], race-TB[2], BTN@meso-TB[2], BTN@race-TB[2] and TCNB@meso-TB[2] under sunlight (up) and UV light (down), which displays tunable fluorescence changes. b‒d Single crystal structures of BTN@meso-TB[2], BTN@race-TB[2] and TCNB@meso-TB[2]. The dashed line represents π⋯π distances (Å). Hydrogen atoms and solvents are omitted for clarity. All the above structure figures were prepared using PyMol (The PyMOL Molecular Graphics System, Version 1.8, Schrödinger, LLC). Photophysical properties of ICCs To understand the present phenomena, solid-state ultraviolet–visible (UV–vis) absorption and emission spectra were measured to elucidate the photophysical properties of the molecules before and after the formation of ICCs. As shown in Fig. 6a and Supplementary Figs. 73‒75, compared to the peak corresponding to the individual components, the two co-crystals of BTN@meso-TB[2] and BTN@race-TB[2] displayed obviously red-shifted absorption appeared at 384 and 379 nm, respectively, while TCNB@meso-TB[2] showed larger red shift at 511 nm. The differences in the degrees of red-shift indicate that TCNB@meso-TB[2] has stronger intermolecular CT interactions between TB[2] and TCNB than BTN. The emission spectra revealed bathochromic shifts of the peaks appeared at 520 nm for BTN@meso-TB[2] and 481 nm for BTN@race-TB[2] compared to TB[2] with emission at around 360 nm (Fig. 6b and Supplementary Figs. 76 and 77). The TCNB@meso-TB[2] exhibited weak deep-red emission at 720 nm with 1.06 ns fluorescence lifetime due to the strong CT interaction quenching the fluorescence of host-guest exciplex (Supplementary Figs. 78 and 81). To confirm that the luminescence of BTN@TB[2] were contributed by TADF features, the transient photoluminescence (PL) decay curve of BTN@meso-TB[2] and BTN@race-TB[2] were measured under N2 atmosphere at ambient temperature. Both emitters exhibited double-exponential decay containing prompt and delayed fluorescence components in the absence of oxygen, with prompt fluorescence lifetime (τp) /delayed fluorescence lifetime (τd) at 678.37 ns/10.51 μs for emitter BTN@meso-TB[2] and 35.19 ns/18.71 μs for emitter BTN@race-TB[2] (Supplementary Figs. 79–83), indicating a potential TADF property. Additionally, the temperature-dependent PL spectra and time-resolved PL spectra of both emitters exhibited that the intensity of the delay component gradually increased with raising temperature from 77 to 300 K (Fig. 6c, d and Supplementary Figs. 84–87), which clearly confirmed the a thermal activation process responsible for TADF generation. Oxygen can quench the triplet excitons of TADF. Thus, we also measured the PL spectra of BTN@TB[2] under vacuum condition and in the air (Fig. 6e and Supplementary Fig. 88). The significant increase in emission intensity were found in vacuum than in the air, further confirming the TADF nature of BTN@TB[2].Fig. 6The photophysical investigations of ICCs.Solid-state UV–Vis absorption (a) and fluorescence (b) spectral profiles recorded for meso-TB[2], race-TB[2], BTN@meso-TB[2], BTN@race-TB[2] and BTN. c Temperature-dependent transient PL decay curves for BTN@meso-TB[2] under vacuum from 77 to 300 K. d Transient PL curves under vacuum of BTN@meso-TB[2] from 77 to 300 K. e Steady-state photoluminescence (PL) spectroscopies for BTN@meso-TB[2] in vacuum and air. f Simplified hRISC diagram showing the plausible excited-state formation upon host-guest complexation and the mechanism of delayed fluorescence emission. The photophysical investigations of ICCs. Solid-state UV–Vis absorption (a) and fluorescence (b) spectral profiles recorded for meso-TB[2], race-TB[2], BTN@meso-TB[2], BTN@race-TB[2] and BTN. c Temperature-dependent transient PL decay curves for BTN@meso-TB[2] under vacuum from 77 to 300 K. d Transient PL curves under vacuum of BTN@meso-TB[2] from 77 to 300 K. e Steady-state photoluminescence (PL) spectroscopies for BTN@meso-TB[2] in vacuum and air. f Simplified hRISC diagram showing the plausible excited-state formation upon host-guest complexation and the mechanism of delayed fluorescence emission. The solid-state PLQYs of BTN@meso-TB[2] and BTN@race-TB[2] were determined to be 21.3 and 21.6%, respectively, while TCNB@meso-TB[2] had very low solid-state PLQYs of 0.3% (Supplementary Figs. 89–91). The relatively low PLQY could be ascribed to the intrinsically weak radiative transition probability of spatial CT states, enhanced nonradiative internal conversion in small-bandgap complexes, and vibrational energy dissipation facilitated by flexible peripheral methoxy groups and exciton migration induced by π–π stacking83–85. The results implied that BTN@TB[2] reaches an opportune balance of the strength between host-guest complexation and CT interaction for TADF generation. In other words, strong inclusion complexation by the synergism of multiple non-covalent bonds and relatively weak CT interactions benefit from the radiative decay by constraining molecular motions and vibration, but strong CT promote non-radiative decay pathways86–88. Theoretical calculations To gain deeper insight into the photoluminescence mechanism, density functional theory (DFT) and time‑dependent DFT calculations were performed using the range‑separated CAM‑B3LYP functional. The initial structures were taken from single‑crystal X‑ray diffraction data and hydrogen atoms were optimized while all other atoms were fixed to preserve the experimental packing. As shown in Fig. 7a, the calculated results suggested that direct S1 ↔ T1 RISC is inefficient due to the relatively large singlet-triplet energy gap (ΔES1-T1 = 0.38 eV). Instead, higher-lying triplet states (T5‒T7) are energetically close to S1 with energy differences ΔES1-Tn = 0.02–0.06 eV (n = 5‒7), and dominate the RISC pathway via a “hot exciton” mechanism. Notably, T7 is nearly resonant with S1, providing an efficient up‑conversion channel assisted by moderate spin–orbit coupling (SOC) (Fig. 7a). Therefore, the efficient TADF emission arises from the hRISC pathway (Tn→S1, n = 5–7), consistent with recent reports on hot‑exciton TADF materials89–93.Fig. 7Theoretical calculations.a Singlet and triplet energy level and SOC values of BTN@meso-TB[2]. b The dominant natural transition orbitals of BTN@meso-TB[2] and TCNB@meso-TB[2] for electron transitions (isovalue = 0.02). Transition contribution and oscillator strength (f) are indicated for major transition. Figures showing structural features were prepared using the program VESTA3. Theoretical calculations. a Singlet and triplet energy level and SOC values of BTN@meso-TB[2]. b The dominant natural transition orbitals of BTN@meso-TB[2] and TCNB@meso-TB[2] for electron transitions (isovalue = 0.02). Transition contribution and oscillator strength (f) are indicated for major transition. Figures showing structural features were prepared using the program VESTA3. Natural transition orbital (NTO) analysis reveals that both inclusion co‑crystals exhibit strong intermolecular CT character: the hole is localized on the electron‑rich TB[2] host, while the electron is distributed over the electron‑deficient BTN or TCNB guest (Fig. 7b), indicating CT occurs from the host to the guests. However, the oscillator (f) strengths differ markedly: BTN@meso‑TB[2] give f ≈ 0.006, whereas TCNB@meso‑TB[2] yields f ≈ 0.001. This indicates that although both systems possess strong CT character, the CT transition in TCNB@meso‑TB[2] is nearly forbidden, leading to weak radiative decay and quenched luminescence. In contrast, BTN@meso‑TB[2] retains sufficient oscillator strength to enable emission. The CT nature facilitates SOC and enables effective coupling between the singlet and higher‑lying triplet states. These results demonstrate that efficient TADF in this system does not rely on a small ΔES1-T1 but rather on the synergistic effect of higher‑lying triplet states and CT‑induced SOC, which together provide multiple RISC pathways. This finding underscores the importance of excited‑state engineering in host-guest co‑crystal systems for achieving efficient TADF. Discussion In summary, we demonstrate an inclusion co-crystal strategy to achieve the efficient TADF emission by controlling the CT interaction and host-guest complexation. We designed and synthesized two electron-rich Tröger’s base-derived molecular boxes (meso-TB[2] and race-TB[2]), by two-step facile and efficient reactions, which could encapsulate electron-deficient BTN and TCNB molecules into its cavity driven by the synergistic effect of π···π interactions and C–H···π/N hydrogen bonds, forming 1:1 host-guest ICCs. BTN@meso-TB[2] and BTN@race-TB[2] co-crystals exhibit solid-state tunable TADF emission (PLQY, 21.3 and 21.6%) because the strong inclusion complexation and relatively weak CT interactions between TB[2] and BTN efficiently promote the radiative decay by constraining molecular motions and vibration. However, for TCNB@meso-TB[2], stronger CT interactions between TCNB and meso-TB[2] benefit from non-radiative decay pathways, leading to weak deep-red fluorescence emission. Theoretical calculations reveal that the TADF mechanism proceeds via a “hot exciton” process through hRISC from higher‑lying triplet states (T5‒T7) to the S1 state. This work provides a promising strategy to build efficient TADF materials by the combination of host-guest chemistry and organic co-crystal engineering. Methods Materials All reagents and solvents were commercially available and used without further purification. TBBr was prepared according to literature procedures72,94,95. The preparation of TBBr was provided in the Supplementary Information. NMR were recorded on Bruker Avance III 400 MHz. Chemical shifts are reported in ppm relative to the signals corresponding to the residual non-deuterated solvents (CD2Cl2: δH = 5.32 ppm and δC = 54.0 ppm, CDCl3: δH = 7.26 ppm and δC = 77.0 ppm). 2D 1H-1H COSY spectra were recorded on Bruker Avance III 600 MHz. High-resolution mass spectra (HRMS) were determined on Agilent Accurate-Mass Q-TOF 6520 instrument with ESI source and Bruker solariX XR with MALDI source. Chiral resolution of race‑TB[2] The enantiomers of race‑TB[2] were resolved by preparative chiral HPLC on a ThermoFisher Vanquish Access system equipped with a ChromRunner® C‑4 column (4.6 × 250 mm, 5 μm). The samples were dissolved in ethanol/DCM (1:2, v/v). The mobile phase consisted of hexane (A) and a mixture of ethanol/methanol (1:2, v/v) as component B, with an isocratic elution of A/B = 72/28 (v/v) at a flow rate of 1.0 mL min−1. The column temperature was maintained at 25 °C, and the elution was monitored at a detection wavelength of 268 nm. Specific rotations were measured on a Zuoguang GP70 automatic polarimeter at 25 °C using the sodium D-line (589 nm). The R2N,R2N-TB[2] and S2N,S2N-TB[2] enantiomers exhibited specific rotations of [α]25D = −220.1 and +193.8 (c = 1.0 g/L, CHCl3), respectively. Single crystal X-ray crystallography All single crystal X-ray diffraction data were collected on a Bruker D8-Venture or Bruker APEX-II CCD diffractometer using Cu-Kα radiation (λ = 1.54178 Å). The crystal structure was solved and refined against all F2 values using the SHELX and Olex 2 suite of programs96,97. All the structure figures were prepared using PyMol (The PyMOL Molecular Graphics System, Version 1.8, Schrödinger, LLC.) or the software package Mercury (Cambridge Crystallographic Data Centre). The detailed experimental parameters are summarized in Supplementary Tables 1–9. Photophysical characterization Solid UV–Vis absorption spectra were recorded on an Agilent Cary 100 UV–Vis. Solid fluorescent and phosphorescent spectra or lifetimes were performed with FLS1000. Solid fluorescence quantum yields were performed on HAMAMATSU C9920-02 by absolute method. Theoretical and computational method All DFT and TD-DFT calculations were performed using ORCA 6.1.0. The starting geometry was constructed from single-crystal X-ray structures, with hydrogen atoms optimized while other atoms were fixed. All the atomic coordinates dataset of optimized computational models are shown in Supplementary Data 1. CAM-B3LYP functional was employed in combination with the def2-TZVP basis set, with DFT-D3 (BJ damping) applied to account for dispersion interactions. Excited-state properties were computed at the same level, and SOC constants between relevant singlet and triplet states were calculated to evaluate the RISC processes. NTO analysis was performed to characterize the nature of the electronic excitations. Supplementary information Supplementary Information Description of Additional Supplementary Information Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Supplementary Data 4 Supplementary Data 5 Supplementary Data 6 Supplementary Data 7 Supplementary Data 8 Supplementary Data 9 Supplementary Data 10 Transparent Peer Review file Supplementary Information Description of Additional Supplementary Information Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Supplementary Data 4 Supplementary Data 5 Supplementary Data 6 Supplementary Data 7 Supplementary Data 8 Supplementary Data 9 Supplementary Data 10 Transparent Peer Review file Source data Source Data Source Data