Two-photon excited circularly polarized luminescence of hetero[9]helicene with multiresonant thermally activated delayed fluorescence

Abstract

Circularly polarized luminescence (CPL) imaging of biomolecules using designed chiral probes holds significant promise for biochemical and analytical applications. In this work, we report two chiral hetero[9]helicene multi-resonant thermally activated delayed fluorescent (MR-TADF) organic molecules DDCzBPh and DDCzBPy, consisting of two annulated MR-TADF units to a central naphthyl or quinoxanyl group, respectively, which exhibit simultaneously strong CPL, two-photon absorption, high photoluminescence quantum yields, and long-lived delayed fluorescence. We demonstrate two-photon-excited CPL of TADF emitters, with their circularly polarized brightness of 0.2 GM, demonstrating their potential for CPL biological imaging.

The design of chiral probes for circularly polarized luminescence imaging of biomolecules is relevant for different applications. Here, the authors report chiral hetero[9]helicene multi-resonant thermally activated delayed fluorescent organic molecules showing strong two-photon circularly polarized luminescence with potential application in biological imagining.

Introduction

Chiral materials that emit circularly polarized luminescence (CPL) have attracted significant attention due to their applications in sensing¹, security², and optical communications³. Recent advance in CPL spectroscopy and microscopy have expanded their utility in biological imaging, encoding vital chiral molecular fingerprints within photoluminescence (PL) spectra^4–6. Traditionally, emitters are excited into a low-energy absorption band, relying on single-photon excitation. Most emitters have absorption onsets that require high-energy photons for excitation, which limits their biological applications, as they often cause cellular damage and suffer from limited penetration depth according to the modified Beer-Lambert law⁷. This limitation can be overcome by employing two-photon excitation (2PE), where molecules are excited by two sequential photons of half the energy of the absorption band. In addition, long-lived emission is strongly desirable in bioimaging, as this allows for time-gated detection methods to be used that increase the signal-to-noise and permit the exclusion of the short-lived autofluorescence background (with lifetimes on the nanosecond scale)⁸. Consequently, developing imaging agents that are chiral, have long-lived PL, and that can be excited by 2PE are desirable. To date, reports on 2PE-CPL emitters remain scarce. Recently, several Eu(III) complexes exhibiting 2PE-CPL properties have been reported, demonstrating exceptional chiroptical properties, narrow emission bands, and millisecond-long PL lifetimes^4–6. However, the development of organic compounds possessing both 2PE-CPL properties and comparable photophysical characteristics remains largely unexplored.

Multi-resonant thermally activated delayed fluorescence (MR-TADF) compounds are one class of emitters that show high photoluminescence quantum yield (Φ_PL) and long PL lifetimes, providing an excellent scaffold for developing 2PE-CPL materials⁹. Compared to conventional donor-acceptor TADF materials, MR-TADF compounds exhibit narrowband emission and longer delayed luminescence lifetimes, primarily due to their rigid polycyclic aromatic hydrocarbon (PAH) frameworks and moderate singlet and triplet energy gap (ΔE_ST)^10,11. To date, dozens of circularly polarized MR-TADF (CP-MR-TADF) emitters have been reported. However, like most small molecules, these compounds have relatively low PL dissymmetry factors (g_PL), typically smaller than 2 × 10^–3 (Supplementary Fig. 1). The g_PL factor describes the chiroptical strength of the CPL and is expressed as: g_PL = 2(I_L – I_R)/(I_L + I_R), where I_L and I_R are the PL intensities of left- and right-CP emission, respectively. Thus, g_PL values range from −2 to +2 for perfectly right- or left-CP emission, respectively, and 0 for the unpolarized or linearly polarized light. Recently, a more comprehensive metric, circularly polarized brightness (CPB), has been introduced to quantify the total number of circularly polarized photons emitted. CPB integrates three key parameters: the molar absorption coefficient (ε_abs), Φ_PL, and g_PL values, and is formulated by the equation: CPB = ε_abs$\cdot$Φ_PL$\cdot$(g_PL/2)¹². Among CP-MR-TADF emitters, those with a helical skeleton generally show higher g_PL values (>10⁻³) compared to other types of CP-MR-TADF emitters (Supplementary Fig. 1). Amongst these, BN[9]H (denoted as DDCzBPh in the current study) has one of the strongest CPL signals in solution^13,14, exhibiting |g_PL| up to 5.8×10⁻³ and CPB of 220.75 M⁻¹ cm⁻¹ in toluene, and showing a delayed fluorescence lifetime, τ_d, of 51.5 μs in 2 wt% doped films in DMIC-TRZ. Its superior chiroptical properties were attributed by the authors to an intrinsic extended helical skeleton. It is worth noting that increasing the helicity of hetero-helicenes has become a go-to strategy to improve g_PL values¹⁵; however, this strategy poses significant synthetic challenges and risks compromising the Φ_PL. Thus, careful molecular design is required to balance both.

For a 2PE process, the two-photon absorption (2PA) cross-section (σ², with the unit of GM, 1 GM = 10⁻⁵⁰ cm⁴ s photon⁻¹), quantifying the probability of the 2PA process, is the key parameter to evaluate the performance of emitters exhibiting 2PA characteristics. The 2PA cross-section is typically several orders of magnitude smaller than the one-photon absorption cross-section¹⁶. A parameter that can be used to quantify the two photon excited CPL is the two photon excited-CPB (2PE-CPB), which can be defined as: 2PE-CPB = σ²$\cdot$Φ_PL$\cdot$(g_PL/2)⁴. In principle, an extended π-conjugation framework can lead to compounds having large transition dipole moments, and therefore high Φ_PL and large 2PE-CPB¹⁷. To date, most reported 2PA-TADF emitters are donor-acceptor (D-A) TADF compounds (Supplementary Fig. 2), which typically require a highly twisted conformations to achieve the small ΔE_ST and efficient reverse intersystem crossing (RISC) required for TADF. However, the twisted conformation leads to modest transition dipole moments and generally lower Φ_PL, and consequently low 2PA σ². MR-TADF emitters, on the other hand, typically have very high Φ_PL as they are based on rigid planar structure and have emissive excited states of short range charge transfer (SRCT) character¹⁸. This combination of features leads to MR-TADF compounds having moderately small ΔE_ST and attractive, potentially enhanced 2PA characteristics. However, to date the 2PA properties of only three MR-TADF molecules, DABNA-2, CzBN, and CzBSe, have been reported, showing small 2PA σ² of 63, 35, and 53 GM at 740 nm¹⁹.

Reports of organic molecules showing 2PA and CPL are likewise scarce, likely due to the requirement for bespoke instrumentation^20–22. Since the breakthrough in 2PE CPL spectroscopy and CPL Laser Scanning Confocal Microscope achieved by Pal and co-workers⁴, one recent report has focused on organic 2PE CPL probes²³. There, a series of functionalized cationic chiral diaza[4]helicenes having high Φ_PL of around 70% in acetonitrile showed one photon exciton (1PE) g_PL values of +5.8/–6.7 × 10^–4 and 1PE-CPB of 3.4 M⁻¹cm⁻¹. 2PA properties were confirmed by the intensity-excitation power dependence, and 2PA σ², when excited at 810 nm (σ₈₁₀²), reached 137 GM. The corresponding 2PE g_PL values are +9.4/–9.2 × 10⁻⁴, which were comparable to those 1PE g_PL values, and the 2PE-CPB values are small at 0.05 GM. These fluorescent probes exhibit lifetimes of approximately 17 ns, which is relatively short compared to long-lifetime lanthanoid probes, such as a recent chiral Eu complex (Φ_PL of 1.6 ms, 2PE-CPB values of 0.16/0.17 GM and high g_PL values of 10⁻²)⁶. Such long lifetimes (μs-ms scale) are advantageous for time-gated luminescence (TGL) by enabling clear temporal separation from short-lived biological background signals like autofluorescence (~ns), significantly improving contrast and sensitivity in bioimaging applications²⁴. Here, we report two chiral 2PE-CP-MR-TADF emitters DDCzBPh and DDCzBPy that have extended hetero-helical skeletons and contain two MR-TADF moieties based on the structure of DtBuCzB that are annulated through a naphthyl or a quinoxanyl group (Fig. 1). The extended helicity and π-conjugated framework contribute to their relatively large |g_PL| and σ² values. Furthermore, their moderate ΔEST and rigid structure are responsible for microsecond-long delayed PL lifetime and high Φ_PL in toluene. As a result, these two emitters have high Φ_PL of 65 and 80%, high 1PE g_PL values of −5.2 × 10⁻³/ + 5.2 × 10⁻³ for P-/M-DDCzBPh and −2.9/ + 2.7 × 10⁻³ for P-/M-DDCzBPy, long delayed lifetimes of 65 and 109.5 μs, and large 2PE σ² of 118 and 142 GM, respectively. Given their attractive 2PA and CPL properties, 2PE CPL measurements were conducted, and comparable 2PE g_PL values were obtained as −4.6 × 10⁻³/ + 3.3 × 10⁻³ for P-/M-DDCzBPh and −2.3 × 10⁻³/ + 2.0 × 10⁻³ for P-/M-DDCzBPy, leading to 2PE-CPB values of 0.20/0.20 GM for P-/M-DDCzBPh and 0.16/0.15 GM for P-/M-DDCzBPy. This is the first study of chiral organic MR-TADF compounds showing 2PE CPL.

Fig. 1. Molecular structures.

Previously reported 2PE CPL active molecules and hetero[9]helicenes DDCzBPh and DDCzBPy in this work.

Results

Molecular design and synthesis

The molecular design strategy and synthetic routes are illustrated in Fig. 1 and Supplementary Fig. 3, respectively. Both DDCzBPh and DDCzBPy incorporate a [9]helicene fragment, constructed from two MR-TADF DtBuCzB fragments bridged by either a naphthalene or quinoxaline unit, respectively. The DtBuCzB motif was selected due to its near-unity photoluminescence quantum yield and long-lived photoluminescence lifetime of 68.8 μs in 1 wt% doped films in mCBP²⁵. Moreover, the parent core CzBN that does not contain peripheral tert-butyl groups has previously been shown to have 2PA character with a relatively modest σ² of 35 GM at 740 nm¹⁹. The extended π-conjugation of DDCzBPh and DDCzBPy is expected to produce a larger 2PA cross section. The naphthalene or quinoxaline moieties were chosen as the bridging unit to modulated the strength of the intramolecular charge transfer between the bridge and the MR-TADF moiety, thereby modulating both the 2PA cross-section and g_PL factor.

Both DDCzBPh and DDCzBPy were synthesized following a common procedure to that previously reported for the former¹³, starting from dibromobenzene or dibromopyrazine, respectively. The final step relies on an oxidative Scholl reaction using DDQ²⁶, which gave DDCzBPh and DDCzBPy as orange and red solids in 80 and 74% yield, respectively. The structures of these compounds were verified by ¹H and ¹³C nuclear magnetic resonance (NMR) spectroscopy, single-crystal X-ray diffraction and high-resolution mass spectrometry (HRMS), and their purity was assessed by melting point determination, elemental analysis (EA), and high-performance liquid chromatography (HPLC) (Supplementary Figs. 4–18). The enantiomers of both compounds were isolated by chiral HPLC with ees of 100% (Supplementary Figs. 10 and 17). The absolute configurations were inferred by comparing the experimental circular dichroism (CD) spectra to the predicted spectra calculated at the TD-DFT-B3LYP/def2-SVP level, including an implicit toluene solvent model (Supplementary Fig. 28). The predicted CD spectrum for P-DDCzBPh is also in line with that reported in the literature¹³.

Single crystals suitable for X-ray diffraction crystallographic analysis were grown by slow diffusion of ethanol into a toluene solution of racemic DDCzBPh, and methanol into a dichloroethane solution of racemic DDCzBPy. DDCzBPh and DDCzBPy crystalize as toluene and dichloroethane solvates, respectively, with both enantiomers present in the asymmetric unit in both structures. The structure of DDCzBPh is a toluene solvate of the known structure of the compound BN[9]H¹³, and has a very similar molecular geometry. The fused rings bridging the DtBuCzB groups show significant distortion (mean deviation from plane 0.073 − 0.087 Å and 0.064 − 0.082 Å, for DDCzBPh and DDCzBPy, respectively) and with DDCzBPy adopting a slightly tighter helical twist (C51-C53-C72-C70 21.4(5)° and C155-C157-C176-C174 22.3(5)°) than that in DDCzBPh (C51-C53-C72-C70 24.7(5)° and C155-C157-C176-C174 24.0(4)°) (Fig. 2a and b, Supplementary Table 1). The observed helical structures are reinforced by strong intramolecular π···π stacking (centroid···centroid 3.269(19) − 3.4283(17) Å and 3.2155(19) − 3.3947(18) Å, for DDCzBPh and DDCzBPy, respectively) between aromatic rings in the two DtBuCzB units (Supplementary Fig. 19) and supported by weaker C-H···π interactions (H···centroid 2.6207(14) − 3.316(2) Å and 2.9904(14) − 3.5666(15) Å, for DDCzBPh and DDCzBPy, respectively) from the tert-butyl groups. Their helical geometry makes these compounds promising candidates for chiroptical applications¹⁵. In both structures, additional weaker intermolecular π···π interactions (centroid···centroid 3.5648(17) − 3.7832(18) Å and 3.6372(18) − 3.8037(18) Å, for DDCzBPh and DDCzBPy, respectively) between adjacent molecules of alternating helicity (Fig. 2c and d) form chains along [010] in DDCzBPh and [100] in DDCzBPy, in a similar manner to that seen in the known structure of DDCzBPh¹³.

Fig. 2. Single-crystal structures.

a DDCzBPh and b DDCzBPy, each with a single independent molecule shown, and minor components of disorder, tert-butyl groups and solvents omitted for clarity in both structures. One representative π···π stacking interaction is indicated. Views of the packing of alternating helicity (orange and light blue) of c DDCzBPh and d DDCzBPy, minor components of disorder, tert-butyl groups, and solvents omitted for clarity.

Theoretical calculations

Theoretical calculations were performed to predict the optoelectronic, chiroptical, and 2PA properties for both compounds. For both compounds, the HOMO/LUMO distribution is reminiscent of the pattern of the MR-TADF emitter DtBuCzB²⁵, showing LUMOs that are localized on the boron atoms and the carbon atoms ortho/para to them, whereas the HOMOs are distributed on the nitrogen atoms and the carbon atoms ortho/para to them (Supplementary Fig. 20). The LUMOs of both compounds extend onto the phenyl or pyrazine ring, leading to a narrower HOMO-LUMO energy gap (2.82 eV for DDCzBPh and 2.79 eV for DDCzBPy) than the parent DtBuCzB (3.37 eV). For DDCzBPy, the HOMO/LUMO energy levels are −5.03/−2.24 eV, both of which are slightly more stabilized than those of DDCzBPh (−4.98/−2.15 eV), attributed to the presence of the π-accepting pyrazine motif.

The excited-state energies and ΔE_ST were calculated at the SCS-ADC2/cc-pVDZ level, as we have previously demonstrated that this level of theory is sufficient to accurately predict ΔE_ST^27,28. The difference density plots for the S₁ and T₁ excited states are shown in Supplementary Fig. 21. The alternating pattern of increasing and decreasing electron density of the DtBuCzB units in the excited S₁ and T₁ states for both compounds reveals that these excited states possess SRCT character. The electron density on the naphthalene or quinoxaline fragments is increased upon the transition between ground state to excited S₁ and T₁ states, indicating that these excited states in both emitters also possess a certain degree of long-range charge transfer (LRCT) character. Both DDCzBPh and DDCzBPy exhibit identical S₁ energies (2.22 eV) and similar T₁ levels (2.13 and 2.11 eV), resulting in moderate ΔE_ST values of 0.09 and 0.11 eV, respectively. In contrast, DtBuCzB has much higher S₁ and T₁ levels at 2.87 and 2.75 eV, respectively, while maintaining a similar magnitude ΔE_ST of 0.13 eV²⁹. The comparable ΔE_ST values and conserved SRCT character in both the S₁ and T₁ states indicate that the designed molecules retain the desired MR-TADF characteristics of the parent DtBuCzB. The spin-orbit coupling matrix elements (SOCME) between low-lying excited states were calculated based on the optimized T₁ geometry at the PBE0/6-31 G(d,p) level (Supplementary Fig. 23). The S₁ and T₁ of both compounds have similar orbital character, resulting in small SOCME values $H_{SOC}^{S1T1}$, of 0.04 cm⁻¹ for DDCzBPh and 0.10 cm⁻¹ for DDCzBPy. The closely low-lying T₂ state to S₁ has distinct orbital character to S₁, leading to larger $H_{SOC}^{S1T2}$ of 0.67 and 0.60 cm⁻¹ for DCzBPh and DCzBPy, respectively. Thus, it is likely that the RISC process involves participation from this state in both compounds.

The g_PL value were calculated following the simplified formula g = 4cosθ×|$\mathbf{m}$ | /|$\mathit{\mu }$ | , where $\mathbf{m}$ and $\mathit{\mu }$ are the magnetic and electronic transition dipole moments, and θ is the angle between these two, obtained from the time-dependent DFT (TD-DFT) calculations based on the optimized S₁ geometry at the B3LYP/def2-SVP level with an implicit toluene solvent field (IEFPCM). The calculated |g_PL| values are −7.2 × 10⁻³ and −7.1 × 10⁻³ for DDCzBPh and DDCzBPy, respectively.

Optoelectronic properties

We began our photophysical study by measuring the absorption and photoluminescence (PL) spectra of the racemates in toluene. The UV-vis absorption spectrum of DDCzBPh (Fig. 3a) shows an intense absorption band peaking at λ_abs of 501 nm (molar absorptivity coefficient, ε, of 44.09 × 10³M^-1 cm^-1) and a lower intensity shoulder band at λ_abs of 546 nm (ε of 10.99 × 10³M^-1 cm^-1). The absorption spectrum of DDCzBPy is slightly bathochromically shifted, with corresponding bands peaking at λ_abs of 511 nm (ε of 31.99 × 10³M^-1 cm^-1) and λ_abs of 546 nm (ε of 29.92 × 10³M^-1 cm^-1). The lowest-energy absorption bands of both emitters are ascribed to SRCT transitions to their S₁ states, whereas the higher-energy bands are attributed to the transitions to the S₂ states with SRCT and slightly LRCT between DtBuCzB fragments and naphthalene or quinoxaline for DDCzBPh and DDCzBPy, respectively. The simulated absorption spectra match well with the experimental ones (Supplementary Fig. 22). In contrast, the absorption spectrum of DtBuCzB shows only one sharp absorption band at around 470 nm (ε of around 30 × 10³M^-1 cm^-1), which corresponds to the SRCT transition. The HOMO/LUMO energy levels are estimated as −5.04/−2.69 eV and −5.11/-2.83 eV for DDCzBPh and DDCzBPy, respectively, calculated from their reversible oxidation and reduction potentials by cyclic voltammetry (CV) and differential pulse voltammetry (DPV) in deaerated DCM (Supplementary Fig. 24, Supplementary Table 3).

Fig. 3. Photophysical properties of DDCzBPh and DDCzBPy.

a UV-vis absorption and steady-state (SS) PL spectra recorded in aerated toluene at 298 K (λ_exc = 340 nm). b SS PL spectra at 77 K (λ_exc = 340 nm), and phosphorescence spectra (1–10 ms, λ_exc = 340 nm) of DDCzBPh and DDCzBPy in methylcyclohexane at 77 K. Time-resolved PL decays (λ_exc = 375 nm) of c DDCzBPh and d DDCzBPy in toluene in air and under vacuum (inset figures are the time-resolved PL decays of the prompt component, λ_exc = 375 nm, time range is 0–100 ns) at room temperature. Concentration of 1 × 10^-5M.

The unstructured PL spectra in toluene for both compounds are almost identical. The PL spectrum of DDCzBPh peaks at λ_PL at 572 nm (Table 1) and is moderately broad, with a FWHM of 0.17 eV (47 nm), while the PL spectrum of DDCzBPy is slightly bathochromically shifted and broadened at λ_PL at 583 nm (FWHM of 0.22 eV, 62 nm, Table 1). Solvatochromic PL studies for DDCzBPh reveal a small degree of positive solvatochromism that is consistent with its excited state being of SRCT character (Supplementary Fig. 25). There is relatively stronger positive solvatochromic behavior for DDCzBPy, which can be attributed to the relatively greater LRCT character of the emissive excited state promoted by the presence of the more strongly electron-withdrawing quinoxaline bridge. The S₁ and T₁ energies were determined from the onsets of the steady-state PL and phosphorescence spectra, respectively, in methylcyclohexane glass at 77 K (Fig. 3b). DDCzBPy possesses more stabilized S₁/T₁ states (2.19/2.04 eV) than those of DDCzBPh (2.23/2.08 eV); however, this nonetheless translates into both compounds having the same ΔE_ST of 0.15 eV. The trend in excited-state energies aligns well with the SCS-ADC2 calculations. This measured ΔE_ST is the same value as that previously reported for DDCzBPh¹³. The moderate ΔE_ST is sufficiently small that these two compounds should emit TADF.

Table 1.

Photophysical properties of DDCzBPh and DDCzBPy

Compound	λabsa / nm	λPLa / nm	FWHMa / nm	ΦPLa / %	τpa,b / ns	τda,b /μs	S1/T1a, b / eV	ΔESTc, d / eV
DDCzBPh	414, 501, 546	572	47	65 (33)	11.2	65.0	2.23/2.08	0.15
DDCzBPy	415, 511, 546	583	62	80 (40)	11.8	109.5	2.19/2.04	0.15

^aIn toluene at 298 K, values in parentheses are Φ_PL measured in aerated toluene.

^bPrompt and delayed emission were measured by TCSPC (λ_exc = 375 nm) and MCS (λ_exc = 375 nm), respectively.

^cMeasured in glassy methylcyclohexane at 77 K.

^dDetermined from the difference in the onsets of the steady-state PL and phosphorescence spectra at 77 K.

The Φ_PL of DDCzBPh and DDCzBPy measured in degassed toluene solutions are 65 and 80% (Table 1), respectively. The PL is sensitive to O₂ and the Φ_PL values decrease dramatically to 33 and 40% in air. The time-resolved PL behavior of both molecules in degassed toluene show biexponential decay kinetics, with prompt and delayed PL lifetimes, τ_p and τ_d, of 11.2 ns and 65.0 μs for DDCzBPh, 11.8 ns and 109.5 μs for DDCzBPy (Fig. 3c, d, Table 1). The longer delayed lifetime for DDCzBPy is correlated with the slightly smaller $H_{SOC}^{S1T2}$ of 0.60 cm^-1 than that calculated for DDCzBPh (0.67 cm^-1) (Supplementary Fig. 23). The values for DDCzBPh are nearly identical to those previously reported (τ_p of 8.4 ns and τ_d of 63 μs)¹³, while for DtBuCzB, only prompt emission was detected in degassed toluene, having a τ_p of 5.0 ns^13,29. Based on the τ_d, the estimated k_RISC in toluene is 2.51×10⁴s^-1 for DDCzBPh and 1.45×10⁴s^-1 for DDCzBPy (see Supplementary Table 4 for full kinetics parameters for both compounds in toluene).

Chiroptical properties

The CD spectra of both enantiomers of DDCzBPh and DDCzBPy recorded in toluene are shown in Fig. 4a and c, respectively. The spectra of the enantiomers are near mirror images. The signal between 520 to 600 nm is associated mainly with the SRCT S₀-S₁ transition. The maximum absorption dissymmetry factors g_abs,max values were observed at around 558 nm and calculated to be −6.14 × 10^-3 and +6.91 × 10^-3 for P-DDCzBPh and M-DDCzBPh, slightly higher than the previously reported g_abs values of −5.6 × 10³ and +5.5 × 10³¹³. The g_abs,max values are −6.32 × 10^-3 and +6.25 × 10^-3 for P-DDCzBPy and M-DDCzBPy at λ_abs of 573 nm. Upon excitation at 416 nm, the enantiomers of both molecules emit mirror-like one-photon excitation (1PE) CPL (ΔI = I_L - I_R, Fig. 4b, d). The g_PL values are −5.2 × 10^-3 and +5.2 × 10^-3 for P-DDCzBPh and M-DDCzBPh (Supplementary Fig. 26), which are similar to those previously reported of −5.5 × 10³ and +5.8 × 10³, respectively¹³. The g_PL values for P-DDCzBPy and M-DDCzBPy are again lower at −2.9 × 10^-3 and +2.7 × 10^-3. These g_PL values are not dissimilar in magnitude to those of other CP-MR-TADF emitters^9,18, and are of the same magnitude as the DFT calculated results (calculated |g_PL| values of −7.2 × 10^-3 and −7.1 × 10^-3 for DDCzBPh and DDCzBPy, respectively). The NTO analysis indicates that the S₁ states of both emitters are predominantly SRCT in nature, with only limited intramolecular long-range charge-transfer (LRCT) character between the naphthyl/quinoxanyl unit and the DtBuCzB backbone. Furthermore, both emitters exhibit similar, approximately C₂-symmetric transition patterns rather than pronounced symmetry breaking (Fig. 5). For DDCzBPy, although the S₁ state is still predominantly SRCT in nature, its larger charge transfer distance (D_CT = 0.679 Å for DDCzBPy and D_CT = 0.270 Å for DDCzBPh) may account for the noticeably lower experimental g_PL than the calculated value³⁰. The 1PE CPL brightness values are calculated to be 74.51 and 74.51 M^-1 cm^-1 for P/M-DDCzBPh and 34.71 and 32.31 M^-1 cm^-1 for P/M-DDCzBPy, respectively. The measured 1PE CPB for P/M-DDCzBPh are smaller than the reported values of 209.7/220.75 M^-1 cm^-113. The detailed calculated parameters are summarized in Supplementary Table 2 and key chiroptical properties are listed in Supplementary Table 5.

Fig. 4. Experimental CD and 1PE-CPL spectra.

CD spectra of (a) DDCzBPh and (c) DDCzBPy in toluene. 1PE-CPL spectra of (b) DDCzBPh and (d) DDCzBPy in toluene at 298 K (λ_exc = 416 nm, dashed line is the Gaussian fit).

Fig. 5. Theoretical calculated chiroptical properties.

NTOs (particle and hole are represented by red and blue colors, respectively) and theoretical predicted electronic (yellow arrow), magnetic (blue arrow) transition dipole moment and their angles during the S₁ to S₀ transition processes, and predicted g_PL values based on the optimized S₁ geometries at the B3LYP/def2-SVP level in the toluene for a DDCzBPh and b DDCzBPy.

Both emitters show strong potential for two-photon absorption due to their extended π-conjugation, which is known to enhance the 2PA cross-section¹⁷. These two designed emitters are expected to show larger 2PA σ² than the weak reported 2PA activity of their parent MR-TADF compound, CzBN¹⁹. To validate these predictions, the 2PA properties of DDCzBPh and DDCzBPy were investigated (Fig. 6a, b). The maximum PL intensity was obtained at a 2PE of 785 nm (1.58 eV) for DDCzBPh and 780 nm (1.59 eV) for DDCzBPy, which is lower than the double of the one-photon absorption maximum wavelength (501 nm, 2.48 eV for DDCzBPh and 500 nm, 2.43 eV for DDCzBPy). This is due to the nonlinear effect of non-degenerate two-photon absorption⁶. Two-photon excitation processes were confirmed from the logarithmic dependence of the 2PE intensity with laser excitation power (Fig. 6c, d). The 2PA cross-sections are 118 GM at 780 nm and 142 GM at 780 nm for DDCzBPh and DDCzBPy, respectively. These values are slightly higher than those reported for the small number of other 2PA-TADF emitters (Supplementary Fig. 2).

Fig. 6. One-photon absorption and two-photon excitation properties.

One-photon absorption (solid line) and two-photon excitation (dotted line) spectra of a DDCzBPh (λ_em = 550-650 nm) and b DDCzBPy (λ_em = 550–650 nm). Log-log plot of the excitation-power dependence (blue dots) of the 2PE-induced photoluminescence intensity for c DDCzBPh, slope is 2.0 and d DDCzBPy, slope is 2.0.

With both compounds showing promising chiroptical and 2PA properties, we next explored their 2PE-CPL properties to assess their potential as 2PE-CPL probes in bioimaging applications. Upon excitation of each emitter by a pulsed laser at 780 nm, 2PE-CPL spectra were recorded using the same CPL spectrometer as the 1PE-CPL spectra (ΔI = I_L - I_R, Fig. 7). The CPL spectra of the enantiomers of each of the emitters show similar mirror-like behavior as their 1PE-CPL spectra (Fig. 4b, d). The g_PL values of -4.6×10^-3 and +3.3×10^-3 for P-/M-DDCzBPh, and −2.3 × 10^-3 and +2.0 × 10^-3 for P-/M-DDCzBPy were measured by two-photon excitation, and were of similar magnitude to their respective values measured by one-photon excitation (−5.2 × 10^-3 and +5.2 × 10^-3 for P-/M-DDCzBPh, −2.9 × 10^-3 and +2.7 × 10^-3 for P-/M-DDCzBPy) (Supplementary Fig. 27). The calculated 2PE-CPB values are 0.20 and 0.20 GM for P-/M-DDCzBPh and 0.16 and 0.15 GM for P-/M-DDCzBPy from their most intense CPL transitions at around 578 nm and 587 nm, respectively (Supplementary Table 6). These 2PE-CPB values are larger than the known example of diaza[4]helicenes (2PE-CPB value of 0.05 GM)²³, and comparable to recently reported chiral Eu complexes (0.16/0.17 GM)⁶. Notably, the extended π-conjugative framework within the helicene boosts 2PA independently of chiroptical effects, raising σ² from 35 GM at 740 nm in the parent compound CzBN to more than 100 GM at 780 nm in the current emitters. Furthermore, the presence of the embedded nitrogen atoms increases the π-accepting character of the quinoxanyl group, strengthening the intramolecular charge transfer (ICT) character of the emissive excited state of DDCzBPy, which was confirmed by a combination of a more pronounced positive PL solvatochromism and larger charge transfer distance (Supplementary Figs. 20 and 25). The stronger ICT character contributes to increasing σ² to 142 GM for DDCzBPy, compared with 118 GM for DDCzBPh. However, by introducing ICT character and decreasing the symmetry of the molecule, the parallel alignment between the electric and magnetic transition dipoles is disrupted, yielding smaller g_PL values, consistent with both experimental measurements and theoretical calculations. Although these g_PL values are smaller than chiral Eu(III) complexes (|g_PL | > 0.01)⁶, the present emitters remain promising for CPL bioimaging. This is supported by a prior report in which (S)-BOD1, with a small |g_PL| of 6 × 10^-4 and CPB of 12.3 M^–1 cm^–1, was used as a CPL probe in living cells using CPL-laser scanning confocal microscopy (CPL-LSCM)³¹.

Fig. 7. 2PE-CPL spectra.

a DDCzBPh (λ_exc = 780 nm) and b DDCzBPy (λ_exc = 780 nm) measured in toluene at room temperature.

Discussion

Our study reports chiral MR-TADF emitters that generate substantial 2PE-CPL. The materials design relies on bridging two MR-TADF DtBuCzB motifs via either a naphthalene or quinoxaline unit to form a helical structure. The CPL is relatively strong, with 1PE g_PL values of -5.2/ + 5.2 × 10^-3 for P-/M-DDCzBPh and −2.9/ + 2.7 × 10^-3 for P-/M-DDCzBPy, high Φ_PL of 65 and 80%, long delayed lifetimes of 65 and 109.5 μs, and large 2PA σ² of 118 and 142 GM for DDCzBPh and DDCzBPy, respectively. The 2PE-CPL spectra show comparable g_PL values of −4.6/ + 3.3 × 10⁻³ for P-/M-DDCzBPh and −2.3/ + 2.0 × 10⁻³ for P-/M-DDCzBPy, associated with 2PE-CPB values of 0.20/0.20 GM and 0.16/0.15 GM, respectively. The outstanding chiroptical and photophysical properties originate from the extended helicity and rigid π-conjugation of these structures, as well as the short-range charge transfer excited states of these MR-TADF emitters. This work marks a significant advancement toward chiral systems with enhanced 2PA properties and delayed luminescence.

Methods

General synthetic procedures

DtBuCzB-Bpin³² was synthesized according to a literature procedure (Supplementary Fig. 3). All other reagents and solvents were obtained from commercial sources and used as received. Air-sensitive reactions were performed under a nitrogen atmosphere using Schlenk techniques. Dry solvents used in the reactions were obtained from an MBRAUN SPS5 solvent purification system. Flash column chromatography was carried out using silica gel (Silia-P from Silicycle, 60 Å, 40-63 µm). Analytical thin-layer chromatography (TLC) was performed with silica plates with aluminum backings (250 µm with F-254 indicator). TLC visualization was accomplished by a 254/365 nm UV lamp. Analysis and separation of enantiomers were conducted on a Shimadzu LC-40 HPLC and FRC system. HPLC were performed using a Daicel 86335 CHIRALPAK IF (250 × 10 mm, 5 µm particle size) column with HPLC-grade dichloromethane/hexane as the mobile phase. GCMS analysis was conducted using a Shimadzu QP2010SE GC-MS equipped with a Shimadzu SH-Rtx-1 column (30 m × 0.25 mm). ¹H and ¹³C NMR spectra were recorded on a Bruker Advance spectrometer (500 MHz for ¹H, 126 MHz for ¹³C) in CDCl₃ or CD₂Cl₂. The following abbreviations have been used for multiplicity assignments: “s” for singlet, “d” for doublet, “dd” for doublet of doublets, “m” for multiplet. ¹H and ¹³C NMR spectra were referenced to residual solvent peaks. Melting points were measured using open-ended capillaries on an Electrothermal 1101D Mel-Temp apparatus and are uncorrected. High-resolution mass spectrometry (HRMS) was performed at the University of Edinburgh Mass Spectrometry Facility on a micrOTOF, and elemental analyses (EA) was performed at the London Metropolitan University by O. McCullough.

Electrochemistry measurements

Cyclic Voltammetry (CV) analyses were performed on an Electrochemical Analyzer potentiostat model 620E from CH Instruments at a sweep rate of 100 mV/s. Differential Pulse Voltammetry (DPV) analyses were carried out with an increment potential of 0.004 V and a pulse amplitude, width, and period of 50 mV, 0.06 s, and 0.5 s, respectively. Samples were prepared in DCM and degassed by sparging with DCM-saturated nitrogen gas for 5 minutes before measurements. All measurements were performed in 0.1 M DCM solution of tetra-n-butylammonium hexafluorophosphate, [ⁿBu₄N]PF₆, which was used as the supporting electrolyte. An Ag/Ag⁺ electrode was used as the reference electrode, while a carbon electrode and a platinum wire were used as the working electrode and counter electrode, respectively. The redox potentials are reported relative to a saturated calomel electrode (SCE) with a ferrocene/ferrocenium (Fc/Fc⁺) redox couple as the internal standard (0.46 V vs SCE for DCM)³³. The HOMO and LUMO energies were determined using the relation E_HOMO/LUMO = −(E_ox / E_red + 4.8) eV, where E_ox and E_red are the onset of anodic and cathodic peak potentials, respectively calculated from DPV relative to Fc/Fc⁺³⁴.

X-ray crystallography

X-ray diffraction data for compounds DDCzBPh and DDCzBPy were collected using a Rigaku MM-007HF High Brilliance RA generator/confocal optics with either XtaLAB P100 or P200 diffractometers [Cu Kα radiation (λ = 1.54187 Å)]. Data for both compounds were collected (using a calculated strategy) and processed (including correction for Lorentz, polarization, and absorption) using CrysAlisPro³⁵. Structures were solved by dual-space methods (SHELXT³⁶) and refined by full-matrix least-squares against F² (SHELXL-2019/3³⁷). Non-hydrogen atoms were refined anisotropically, and hydrogen atoms were refined using a riding model. Multiple tert-butyl groups in both structures showed rotational disorder (with one showing additional rocking out of the plane of the adjacent ring structure) and were each modelled in two parts with appropriate thermal and geometric restraints. Some disordered tert-butyl groups were not split due to the low occupancy of the minor part and instead were treated with thermal restraints. In the structure of DDCzBPh, two of the toluene solvates showed positional disorder and were each modelled in two parts with the minor part placed using FragmentDB^38,39 and refined with appropriate thermal and geometric restraints. The structure of DDCzBPy showed high proportions of void space (1268 ų, 14%) and the SQUEEZE⁴⁰ routine implemented in PLATON⁴¹ was used to remove the contribution to the diffraction pattern of the unordered electron density in the void spaces. All calculations except SQUEEZE were performed using the Olex2⁴² interface. Selected crystallographic data are presented in Supplementary Table 1. CCDC 2474402 and 2474403 contain the supplementary crystallographic data for this paper. These data can be obtained free of charge from The Cambridge Crystallographic Data Centre via www.ccdc.cam.ac.uk/structures.

Fitting of the time-resolved luminescence measurements

Time-resolved PL measurements were fitted to a sum of exponential decay model, with chi-squared (χ²) values between 1 and 1.5, using the Fluoracle software (Edinburgh Instruments, UK). Each component of the decay is assigned a weight, (w_i), which is the contribution of the emission from each component to the total emission. The average lifetime was then calculated using the following:

For two exponential decay model:

$${\tau }_{AVG}={\tau }_1{w}_1+{\tau }_2{w}_2$$ 1

with weights defined as $w_1$=$\frac{A_1{\tau }_1}{A_1{\tau }_1+A_2{\tau }_2}$ and $w_2$=$\frac{A_2{\tau }_2}{A_1{\tau }_1+A_2{\tau }_2}$, where A₁ and A₂ are the preexponential-factors of each component.

For three exponential decay model:

$${\tau }_{AVG}={\tau }_1{w}_1+{\tau }_2{w}_2+{\tau }_3{w}_3$$ 2

with weights defined as $w_1$=$\frac{A_1{\tau }_1}{A_1{\tau }_1+A_2{\tau }_2+A_3{\tau }_3}$, $w_2$=$\frac{A_2{\tau }_2}{A_1{\tau }_1+A_2{\tau }_2+A_3{\tau }_3}$ and $w_3$=$\frac{A_3{\tau }_3}{A_1{\tau }_1+A_2{\tau }_2+A_3{\tau }_3}$, where A₁, A₂ and A₃ are the preexponential-factors of each component.

Theoretical calculations

Density functional theory (DFT) and TD-DFT calculations were performed using the Gaussian 16 revision C.01 package⁴³. The ground-state optimizations and vibrational frequency calculations were carried out using DFT using the CAM-B3LYP⁴⁴ functional and the 6-31 G(2 d,p) basis set⁴⁵ in toluene with the PCM model, starting from a crystal structure. The vertical excited-state calculations were performed using time-dependent DFT (TD-DFT)⁴⁶ using the B3LYP functional and the def2-SVP basis set in toluene. Spin-orbit coupling matrix elements (SOCME) were calculated based on the optimized triplet excited-state geometry at the PBE0/6-31 G(d,p) level in the gas phase using PySOC⁴⁷. The molecular orbital distributions were visualized with Gaussview 6.0⁴⁸. Wavefunction calculations were performed using the Turbomole 7.5 package^49,50 using the resolution of the identity approximation⁵¹. The spin-component scaling (SCS) modification was applied to all calculations⁵². Vertical excited state calculations using SCS-ADC(2)^53,54 with the cc-pVDZ basis set⁵⁵ were performed on the DFT optimized ground state geometry, with the first singlet state⁵⁶ and first triplet state^57,58 calculated. Calculations were submitted and processed using the Digichem software package (version 6)^59,60, which incorporates a number of publicly available software libraries, including: cclib⁶¹ for parsing of result files, VMD⁶²/Tachyon⁶³ for 3D rendering, Matplotlib⁶⁴ for drawing of graphs, Open Babel⁶⁵/Pybe⁶⁶ file interconversion.

Photophysical measurements

Optically dilute solutions of concentrations on the order of 10⁻⁵ or 10⁻⁶ M were prepared in spectroscopic or HPLC-grade solvents for absorption and emission analysis. Absorption spectra were recorded at room temperature on a Shimadzu UV-2600 double-beam spectrophotometer with a 1 cm quartz cuvette. Molar absorptivity determination was verified by a linear regression analysis of values obtained from five independent solutions at varying concentrations with absorbance ranging from 0.237 to 0.842 for DDCzBPh and from 0.279 to 0.768 for DDCzBPy.

For emission studies, aerated solutions were sparged with compressed air for 5 min, degassed solutions were prepared via three freeze-pump-thaw cycles, and spectra were measured using a homemade Schlenk quartz cuvette. Steady-state emission, excitation spectra, and time-resolved emission spectra were recorded at 298 K using an Edinburgh Instruments FS5 fluorimeter. Samples were excited at 340 nm for steady-state PL measurements in toluene and the solvatochromism study using a 150 W xenon lamp. Phosphorescence spectra were collected using a 5 W pulsed xenon microsecond flash lamp by multi-channel scaling (MCS) mode, with a gate time of 1-10 ms. Time-resolved PL decays were carried out using time-correlated single-photon counting (TCSPC) mode for short-time range (1 ns-100 ns) and multi-channel scaling (MCS) mode for longer time ranges (1ns-400 μs, 1 ns-800 μs). The samples were excited at 375 nm by a 5 mW EPL-375 picosecond pulsed diode laser.

Photoluminescence quantum yields for solutions were determined using the optically dilute method⁶⁷ in which four sample solutions with absorbances of ca. 0.090, 0.070, 0.045, 0.024 at 350 nm were used for DDCzBPh and 0.092, 0.057, 0.031, 0.018 at 350 nm were used for DDCzBPy. The Beer-Lambert law was found to be linear over the solution concentrations. For each sample, linearity between absorption and emission intensity was verified through linear regression analysis and additional measurements were acquired until the Pearson regression factor (R²) for the linear fit of the data set surpassed 0.9. Individual relative quantum yield values were calculated for each solution, and the values reported represent the gradient value. The equation Φ_s = Φ_r(A_r/A_s)(I_s/I_r)(n_s/n_r)² was used to calculate the relative quantum yield of each of the sample, where Φ_r is the absolute quantum yield of the reference, n is the refractive index of the solvent, A is the absorbance at the excitation wavelength, and I is the integrated area under the corrected emission curve. The subscripts s and r refer to the sample and reference, respectively. A solution of quinine sulfate (Φ_r = 54.6% in 1 N H₂SO₄) was used as the external reference⁶⁸.

The singlet-triplet energy gaps (ΔE_ST) in methylcyclohexane and films were estimated from the onsets of the steady-state PL and phosphorescence spectra at 77 K. Phosphorescence spectra were measured using the time-gated method with a gate time of 1-10 ms, where the samples were excited at 340 nm by a 5 W microsecond flash lamp.

Electronic circular dichromism and one-photon excited circularly polarized luminescence

Electronic circular dichromism (ECD) spectra were measured in toluene at 10^-5M using a MOS-500 circular dichroism spectrometer.

One photon excited circularly polarized luminescence (CPL) was measured in toluene at 10^-5M with a home-built (modular) spectrometer⁶⁹. The excitation source was a EQ 99 broad band (200 - 1000 nm) laser-driven light source (Elliot Scientific). The excitation wavelength was selected by feeding the broadband light into an Acton SP-2155 monochromator (Princeton Instruments); the collimated light was focused into the sample cell (1 cm quartz cuvette). Sample PL emission was collected perpendicular to the excitation direction with a lens (f = 150 mm). The emission was fed through a photoelastic modulator (PEM) (Hinds Series II/FS42AA) and through a linear sheet polariser (Comar). The light was then focused into a second scanning monochromator (Acton SP2155) and subsequently on to a photomultiplier tube (PMT) (Hamamatsu H10723 series). The detection of the CPL signal was achieved using the field modulation lock-in technique. The electronic signal from the PMT was fed into a lock-in amplifier (Hinds Instruments Signaloc Model 2100). The reference signal for the lock-in detection was provided by the PEM control unit. The monochromators, PEM control unit and lock-in amplifier were interfaced to a desktop PC and controlled by a custom-written LabVIEW graphic user interface. The lock-in amplifier provided two signals, an AC signal corresponding to (I_L- I_R) and a DC signal corresponding to (I_L + I_R) after background subtraction. The emission dissymmetry factor was therefore readily obtained from the experimental data, as 2 AC/DC. Spectral calibration of the scanning monochromator was performed using a Hg-Ar calibration lamp (Ocean Optics, HG-2). A correction factor for the wavelength dependence of the detection system was constructed using a calibrated lamp (Ocean Optics, HL-2000). The measured raw data was subsequently corrected using this correction factor. The validation of the CPL detection systems was achieved using light emitting diodes (LEDs) at various emission wavelengths. The LED was mounted in the sample holder and the light from the LED was fed through a broad band polarising filter and π/4 plate (ThorLabs, LPVISE100-A and AQWP05M-600) to generate circularly polarised light. Prior to all measurements, the π /4 plate and a LED were used to set the phase of the lock-in amplifier correctly. The emission spectra were recorded with 2 nm step size and the slits of the detection monochromator were set to a slit width corresponding to a spectral resolution of 2 nm. CPL spectra (as well as total emission spectra) were obtained through an averaging procedure of 10 scans.

Two-photon spectroscopy and two-photon excited circularly polarized luminescence⁴

Two-photon excitation CPL spectroscopy has been achieved by coupling (beam routing using mirrors, Thor Labs BB1-E03) a tunable femtosecond pulsed laser (680–1300nm, Coherent Discovery TPC, 100fs, 80MHz) to a pre-existing CPL spectrometer, detailed above. Initial proof of concept two-photon spectroscopy has been achieved by perpendicularly mounting an Ocean Optics HR2000Pro (2048-pixel linear CCD Sony ILX5 chip, H3 grating, 350 - 850nm spectral region, coupled via a 74-DA collimating lens) spectrometer as a ‘third arm’ to the Discovery TPC laser. The laser beam was focused onto the centre of the 1 cm path sample holder (Thor labs CVH100) by a dedicated ultrafast laser lens (Edmund Optics 11711, 50 mm focal length). The spectrometer has also been equipped with a perpendicularly mounted 365nm LED (Nichia, 1W) and has been operated using a modified version of the above-mentioned custom time-resolved detection and accumulation algorithm written in LabVIEW 2013 software. To eliminate unwanted artefacts associated with stray light from multiphoton (MP) excitation, each spectrometer has been equipped with a rotating filter wheel (Thor Labs, CFW6) housing an LP420 (Comar Optics, for 365 nm UVLED excitation) and SP650 - SP800, 25 nm spectral intervals (Edmund Optics, for MP excitation) filters. MP(2PE)-PL spectra were obtained using the following parameters: integration time - 10 ms, resolution - 0.25 nm/pixel, accumulation - 3000.

MP-CPL spectra were obtained using the following parameters: slits - 9 nm max, integration time - 1 ms, resolution step - 2 nm, accumulations - 20.

Cross-section determination

The cross-sections (σ²) of the compounds discussed herein are calculated according to established procedures⁷⁰:

$${\sigma }_s^2={\sigma }_r^2\cdot {\Phi }_r\cdot \frac{C_r{n}_s{F}_s\left(\lambda \right)}{{\Phi }_s{C}_s{n}_r{F}_r\left(\lambda \right)}$$ 3

Where s is sample, r is reference, $\varphi$ is the total emission quantum yield of the compound, C is the concentration, n the refractive index and $F_s\left(\lambda \right)$ and $F_r\left(\lambda \right)$ are the integrated PL spectrum for the sample and reference, respectively. Additionally, we have demonstrated that the excitation process is definitely a two-photon event by recording an excitation power dependence; the resulting line has a slope of two on a logarithmic scale. The cross sections were calculated with reference to rhodamine B in methanol. The MP spectra were recorded in 1 cm path length quartz cuvettes, in air-equilibrated analytical grade acetonitrile at concentrations ca. 10^{- 5} M.

In the reaction conditions, RT indicates an ambient temperature of 20–30 °C.

Author contributions

S.W.: Conceptualization, Synthesis and Investigation, Review & Editing. Y.X.: Investigation, Writing-Original Draft Preparation, Review & Editing. D.J.B.: Investigation, Review & Editing. A.P.M. and D.B.C.: Crystal structure analysis, Review & Editing. R.P.: Resources, Review & Editing. E.Z.-C.: Project management, Supervision, Resources, Writing-Review & Editing.

Peer review

Peer review information

Nature Communications thanks Gennaro Pescitelli and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

S.W. acknowledges the China Scholarship Council (201906250199). Y.X. acknowledges the China Scholarship Council (202106310038) for support. E.Z.-C. acknowledges support from the Engineering and Physical Sciences Research Council (EPSRC; EP/Z535291/1, EP/W015137/1, and EP/W007517/1). R.P. and D.J.B. acknowledge support from BBSRC (BB/S017615/1 and BB/X001172/1) and the EPSRC (EP/X040259/1) for funding.

Data availability

The research data generated in this study have been deposited in the University of St Andrews datasets under accession code 10.17630/3b69fc62-063d-4411-8995-43d9cecd75e1 https://doi.org/10.17630/3b69fc62-063d-4411-8995-43d9cecd75e1. The X-ray structural data can be obtained from The Cambridge Crystallographic Data Centre as deposition numbers 2474402 https://doi.org/10.5517/ccdc.csd.cc2p1tf0 and 2474403 https://doi.org/10.5517/ccdc.csd.cc2p1tg1. Source Data are provided with this manuscript. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77554-5.