Introduction

The electron (or charge) carrier dynamics in photoelectric and electrochemical devices are key to determining the performance of devices1,2,3,4,5,6. Chromophore–chromophore interaction as well as the electronic-state coupling of a chromophore can modulate such electron carrier dynamics7,8,9,10,11,12,13,14,15,16. In this regard, many multi-chromophore systems have been widely used for developing highly efficient photoelectric or electrochemical devices using chromophore–chromophore interaction7,12,13,14,15,16,17,18,19,20,21. Among the multi-chromophore systems, covalently linked dimers (CLDs) have attracted much attention as potential materials to provide high energy-conversion efficiencies in photovoltaic devices because their excited-state relaxation dynamics, such as the excimer formation, intramolecular charge transfer (ICT), and singlet fission (SF), can be modulated by strategic molecular design9,15,22,23,24,25,26,27. Especially the dynamics of SF, which is a conversion process from one singlet exciton into two triplet excitons, have been actively investigated with various time-resolved spectroscopies to overcome the limit of Shockley–Queisser power conversion efficiency17,28,29,30,31,32.

The results of extensive studies on SF dynamics and mechanisms showed that they cannot be easily explained by a single unifying mechanism of SF; instead, the intermolecular and intramolecular SF dynamics occur through various types of species such as charge transfer species, excimers, and higher excited vibrational and electronic states, depending on the interactions between chromophores21,24,25,26,29,31,33,34,35,36,37. For example, Zirzlmeier et al. suggested that the SF process occurring in ortho-, meta-, and para-linked pentacene dimers proceed through virtual CT states29. Margulies et al. reported that the covalently linked terylene-3,4:11,12-bis(dicarboximide) (TDI) dimer with a stacked structure forms an excimer within <200 fs, whereas the slip-stacked TDI dimer in a nonpolar solvent forms the correlated triplet pair ((T1T1)), which is an intermediate in the SF process30. They also showed that the slip-stacked TDI dimer forms a CT state in a few picoseconds in a polar solvent, suggesting that adjusting the CT state energy relative to exciton states via solute–solvent interaction can either promote or inhibit SF30. Ni et al. reported that, upon excitation at 250 nm, the cofacial perylene dimer undergoes SF from the upper excited vibrational and electronic states, whereas upon excitation at 450 nm, it fast forms an excimer, which relaxes to the ground state within nanoseconds38. Korovina et al. showed that the ortho- and para-bis(ethynyltetracenyl)benzene dimers with a relatively stronger through-bond coupling exhibit more efficient SF dynamics than the meta-bis(ethynyltetracenyl)-benzene dimer39. Especially, unlike the ortho-bis(ethynyltetracenyl)benzene dimer, which forms only (T1T1), the para-bis(ethynyltetracenyl)benzene dimer shows complete SF dynamics to form free triplets, suggesting that the rotational flexibility between the acenes in the dimers plays an important role in SF dynamics39. Shizu et al. demonstrated that the efficient SF dynamics of para-bis(ethynyltetracenyl)benzene dimer is due to large vibronic coupling and the small energy difference between the singlet excited state and the (T1T1) state40. In contrast, Nakamura et al. showed that in a series of ortho-, meta-, and para-bis(tri isopropyl silylethynyl)-tetracenyl)benzene dimers, the meta-linked tetracene dimer exhibits more efficient SF dynamics than ortho- and para-linked tetracene dimers15. They suggested that the large conformational flexibility and weak electronic coupling play a critical role in their SF dynamics15. In addition, the experimental and theoretical calculation results of various CLDs showed that compared to ortho- and para-linked dimers, meta-linked dimers exhibit more efficient SF dynamics due to the small binding energy (Eb) of (T1T1) (Eb = 2E | S0T1〉−E | (T1T1)〉)29,41,42,43.

Despite numerous experimental and theoretical approaches to determining the SF dynamics of CLDs, a full understanding of the parameters influencing the SF dynamics of CLDs is still lacking. Additionally, studies on the effects of a molecular configuration, which can affect its excited-state relaxation dynamics, may provide clues for the optimized spatial arrangements that ensure the high-energy conversion efficiency of a real device. In this regard, the role of conformational flexibility in the SF dynamics of CLDs has been studied, but still needs further clarification. Accordingly, in-depth studies are needed to understand their excited-state relaxation dynamics, including SF. From this perspective, we designed and synthesized two covalently ortho-linked pyrene (Py) dimers, anti- and syn-1,2-di(pyrenyl)benzene (Anti-DPyB and Syn-DPyB) (see Supplementary Methods and Fig. 1 and Supplementary Figs. S1S6), which we expected to have different configurations, to elucidate the effect of the molecular configuration for their excited-state relaxation dynamics using steady-state and time-resolved spectroscopies. The data show that both Anti-DPyB and Syn-DPyB form excimers, which rapidly relax to the (T1T1) state regardless of solvent polarity, indicating the occurrence of SF dynamics. Notably, the (T1T1)s of Anti-DPyB in both n-hexane and acetonitrile dissociate to free triplets as the end product, completing SF, whereas the dissociation of (T1T1)’s in Syn-DPyB is less favorable compared with that in Anti-DPyB, indicating that the (T1T1) of Syn-DPyB is more bound with respect to the separated triplets than that of Anti-DPyB. Our findings show that the SF dynamics of Anti- and Syn-DPyB differ due to the different molecular configurations of Anti-DPyB and Syn-DPyB.

Fig. 1: Structures and UV–visible absorption (solid line) and emission spectra (dotted line).
figure 1

A Geometric isomers of anti-1,2-di(phenyl)benzene (Anti-DPyB) and syn-1,2-di(phenyl)benzene (Syn-DPyB). B UV–visible absorption spectra of Anti-DPyB and Syn-DPyB in n-hexane (black) and acetonitrile (red). Emission spectra of Anti-DPyB and Syn-DPyB in n-hexane (black) and acetonitrile (red) (λex = 345 nm).

Results

Molecular structures

To characterize the structures of Anti-DPyB and Syn-DPyB, we calculated their minimum energy structures using density functional theory (DFT). The optimized structures are shown in Fig. 1A. Two Py moieties in Anti-DPyB are far from each other, whereas Syn-DPyB shows a partial overlap of two pre-stacked Py moieties. According to the calculations using B3LYP/6-31G(d,p), the distances between two Py moieties in Anti-DPyB and Syn-DPyB are 10.326 and 8.481 Å, respectively (see Supplementary Data 2). This result is consistent with the calculation results reported by Jo et al. 44. The optimized structure of Anti-DPyB obtained from our calculation is similar to its crystal structure (Supplementary Fig. S7, Supplementary Tables S1S6, and Supplementary Data 1).

Steady-state absorption and emission spectra

As shown in Fig. 1B, Anti-DPyB in n-hexane and acetonitrile shows two vibrationally resolved absorption bands at around 250–280 and 300–380 nm. This feature is similar to the absorption spectrum of 1-phenylpyrene (Ph-Py)9,45. In contrast, Syn-DPyB exhibits a broad and vibrationally resolved absorption band at 250–475 nm. The vibrationally resolved absorption bands observed from Anti-DPyB and Syn-DPyB indicate that both compounds have highly rigid structures in the ground state. Furthermore, the absorption bands of Anti-DPyB and Syn-DPyB do not show noticeable dependence on solvent polarity (Fig. 1B).

To investigate the excited-state behaviors of Anti-DPyB and Syn-DPyB, we measured their emission spectra in n-hexane and acetonitrile with excitation at 345 nm, which corresponds to the major absorption peak. Anti-DPyB and Syn-DPyB in both solvents show dual emission bands (380 and 480 nm for Anti-DPyB and 420 and 480 nm for Syn-DPyB), the detailed features of which are different. We checked the possibility that Py molecules are present as impurities in Anti-DPyB and Syn-DPyB solutions. It is known that the Py molecule shows a strong fluorescence in solutions. Therefore, if Py molecules are present as impurities in Anti-DPyB and Syn-DPyB solutions, they may contaminate the fluorescence spectra from the Anti-DPyB and Syn-DPyB samples. To check this possibility, we measured the fluorescence excitation spectra of Anti-DPyB and Syn-DPyB in acetonitrile at two emission peak positions (380 and 480 nm). As shown in Supplementary Fig. S8, the fluorescence excitation spectra from Anti-DPyB and Syn-DPyB are significantly different from the absorption spectrum of the Py molecule. This result indicates that Py molecules do not exist as impurities in Anti-DPyB and Syn-DPyB solutions. The 380 nm band of Anti-DPyB is structured, whereas the broad band centered at 480 nm is structureless (Fig. 1B). Unlike Anti-DPyB, both of the emission bands of Syn-DPyB are structured. Based on numerous previous studies on Py and Py derivatives46,47, the shorter-wavelength (380 and 420 nm) emissions from Anti-DPyB and Syn-DPyB can be assigned to Py monomer moieties, and the longer-wavelength (480 nm for both) emissions to the excimer formed via the association of excited and unexcited Py’s. In addition, the broad structureless bands centered at 480 nm observed for Anti-DPyB are highly similar to the typical excimer bands observed for Py and Py derivatives. In terms of solvent dependence, in Anti-DPyB, the relative intensities of the emissions from the monomeric Py moiety and excimer show a significant dependence on solvent polarity, whereas Syn-DPyB does not show definite solvent dependency.

To accurately determine the singlet energy (ES1) and triplet energy (ET1) values of Py, Ph-Py, Anti-DPyB, and Syn-DPyB, we also measured emission spectra of Py, Ph-Py, Anti-DPyB, and Syn-DPyB in MTHF containing iodomethane at 77 K. As shown in Supplementary Fig. S9, all four compounds show dual emission bands at around 370–550 and 580–800 nm, corresponding to fluorescence and phosphorescence, respectively. The ES1 values of Py, Ph-Py, Anti-DPyB, and Syn-DPyB are determined to be 3.3, 3.3, 3.3, and 3.1 eV, respectively. From the phosphorescence spectra, the ET1 values of Py, Ph-Py, Anti-DPyB, and Syn-DPyB are determined to be 2.10, 2.03, 2.04, and 1.87 eV, respectively.

Fluorescence lifetime

To further elucidate the excited-state relaxation dynamics, we measured the fluorescence decay profiles of Anti-DPyB and Syn-DPyB in n-hexane and acetonitrile. As depicted in Fig. 2, all decay profiles satisfactorily fit with bi-exponential functions. The determined fluorescence lifetimes are summarized in Table 1. The fast (τf1) and slow (τf2) time constants are predominantly observed in the shorter- and longer-wavelength emissions, respectively. Since the shorter- and longer-wavelength emissions arise from the Py monomeric unit and excimer, respectively, the fast (τf1) and slow (τf2) time constants correspond to the lifetimes of the Py monomeric unit and excimer, respectively. In addition, as depicted in Fig. 2A, the fluorescence decay profile of Anti-DPyB measured in the range of 500–600 nm in n-hexane shows an additional kinetic component with a rise time of 1.24 ns.

Fig. 2: Fluorescence decay profiles.
figure 2

A Fluorescence decay profiles of Anti-DPyB in n-hexane (black) and acetonitrile (blue), respectively. In A, the inset shows the fluorescence decay profile of Anti-DPyB measured at the long wavelength (500–600 nm) in n-hexane. The fluorescence decay profile of Anti-DPyB measured at the long-wavelength (500–600 nm) shows a rise time of 1.24 ns as well as a decay time of 11.1 ns. B Fluorescence decay profiles of Syn-DPyB in n-hexane (black) and acetonitrile (blue), respectively.

Table 1 Emission quantum yields (Φ), emission lifetimes (τ), average emission lifetimes (〈τ〉), radiative rate constants (kR), and nonradiative rate constants (kNR) of Anti-DPyB and Syn-DPyB in n-hexane and acetonitrile.

TA spectra

To elucidate the excited-state relaxation dynamics, we measured the femtosecond transient absorption (fs-TA) spectra for Anti-DPyB and Syn-DPyB in n-hexane and acetonitrile with 350 nm excitation. The TA spectra of Anti-DPyB in n-hexane and acetonitrile exhibit broad signals at 400–700 nm, corresponding to the excited-state absorption (ESA) (see Fig. 3 and Supplementary Fig. S10). With time, these broad positive signals transform into structured signals. On the other hand, the TA spectra of Syn-DPyB in n-hexane and acetonitrile exhibit intense ESA signals at 450–530 nm with a weak absorption tail (550–700 nm) (Fig. 3C and D). To extract further information, we analyzed the TA spectra of Anti-DPyB and Syn-DPyB using singular value decomposition (SVD) analysis (Supplementary Information). The SVD analysis for the TA spectra of Anti-DPyB and Syn-DPyB identified four and three significant singular components, respectively (Supplementary Figs. S11 and S12). As shown in Supplementary Fig. S13, the significant rSVs for Anti-DPyB in n-hexane and acetonitrile can be expressed by a tetra-exponential function with shared relaxation times (3.6 ± 0.3, 231 ± 19 ps, 1.75 ± 0.12 ns, and >10 ns in n-hexane; 2.8 ± 0.1, 24.3 ± 0.5, 495.7 ± 6.5 ps, and >10 ns in acetonitrile). The significant rSVs for Syn-DPyB in n-hexane and acetonitrile can be expressed by a tri-exponential function with shared relaxation times (2.3 ± 0.8, 9.7 ± 0.5 ps, and 6.4 ± 0.2 ns in n-hexane; 2.8, 8.0 ± 0.6 ps, and 4.8 ± 0.2 ns in acetonitrile). The time constants are summarized in Table 2.

Fig. 3: Transient absorption spectra.
figure 3

A, B Transient absorption spectra of Anti-DPyB in n-hexane and acetonitrile, respectively. C, D Transient absorption spectra of Syn-DPyB in n-hexane and acetonitrile, respectively.

Table 2 Time constants determined from TA measurements for Anti-DPyB and Syn-DPyB in n-hexane and acetonitrilea.

To observe the long-lived species, we measured the nanosecond TA spectra of Anti-DPyB and Syn-DPyB in n-hexane and acetonitrile with 355 nm excitation. Anti-DPyB in n-hexane and acetonitrile shows a weak and broad absorption band around 445 nm at a time delay of a few microseconds (Supplementary Fig. S14), suggesting the presence of a long-lived species. In contrast, Syn-DPyB does not exhibit any absorption band in both n-hexane, and acetonitrile, indicating that no long-lived species, such as a triplet species, exist at the μs–ms time scale.

Discussion

Intramolecular excimer formation and electronic coupling

The optimized structures of Anti-DPyB and Syn-DPyB show that the distance between two Py moieties in Anti-DPyB is longer than that in Syn-DPyB, suggesting that the interaction in Anti-DPyB between two Py moieties should be weaker than that in Syn-DPyB with the partial overlap of two Py moieties. Anti-DPyB has an absorption spectrum similar to that of Py or Ph-Py9,45, indicating that in the ground state, the two Py moieties in Anti-DPyB have a monomeric character. In contrast, the pre-stacked structure of Syn-DPyB is expected to show the characteristic feature of an excimer. Indeed, Syn-DPyB exhibits a single broad and vibrationally resolved absorption band at 250–475 nm owing to the strong ππ interaction between the two Py moieties. The steady-state spectroscopic results confirm this prediction: in the ground state, the interaction between the two Py moieties in Anti-DPyB is much weaker than that in Syn-DPyB. On the other hand, the cyclic voltammograms for the reduction of Anti-DPyB and Syn-DPyB in THF show two separate peaks, whereas the cyclic voltammogram of Py exhibits a single peak (Supplementary Fig. S15). From the cyclic voltammograms, the splitting energies (Ered2Ered1) of Anti-DPyB and Syn-DPyB are determined to be 0.09 and 0.12 V, respectively (Table 3). Compared to Syn-DPyB, the lower splitting energy of Anti-DPyB indicates that the electronic coupling between the two Py moieties in Anti-DPyB is relatively weaker than that in Syn-DPyB. This result is consistent with the steady-state spectroscopic results.

Table 3 Electrochemical parameters of Py, Anti-DPyB and Syn-DPyB evaluated by cyclic voltammograms (Eox1 and Eox2: first and second oxidation potentials, Ered1 and Ered2: first and second reduction potentials).

We note that the excimer formation in Anti-DPyB and Syn-DPyB is intramolecular. The emission spectra of Anti-DPyB and Syn-DPyB are not influenced by the solute concentration (Supplementary Fig. S16). This result confirms that the excimers in Anti-DPyB and Syn-DPyB form due to intramolecular rather than intermolecular interaction. Because, in Anti-DPyB, the interaction between the two Py moieties in the ground state is weak due to the long distance between the two chromophores with twisted alignment, the intramolecular excimer formation requires the rearrangement of two distant Py moieties. This scenario is consistent with the flexible structure of the excimer of Anti-DPyB, reflected in its broad structureless excimer emission band (480 nm). In contrast, Syn-DPyB is expected to rapidly form an excimer with no or less structural rearrangement because of the partial overlap of the two Py moieties. Based on the structured excimer emission band (480 nm) of Syn-DPyB, we suggest that the excimer structure is as rigid as the structure in the ground state.

The solvent dependence on the emission spectra of Anti-DPyB can be rationalized by considering the following scenario: As Py is a hydrophobic molecule, two Py moieties in a nonpolar solvent show monomeric behavior, whereas a high polarity solvent facilitates the hydrophobic interaction of two Py moieties, resulting in more efficient excimer formation. The absence of solvent dependence on the emission of Syn-DPyB is probably due to its rigid structure owing to the strong ππ interaction between two Py moieties. The solvent dependency on the emission of Anti-DPyB and Syn-DPyB can be also interpreted in terms of the change in the dipole moment (μ) induced by the structural change in the excited state. The solvent dependency on the emission of Anti-DPyB suggests that the Δμ (= μeμg) value for Anti-DPyB is likely larger than that for Syn-DPyB. The DFT calculation, which was performed using CAM-B3LYP-D3/6-31G**, shows that the dipole moments (μg) of Anti-DPyB and Syn-DPyB in the ground state are 0.021 and 0.1394 D, respectively (see Supplementary Data 2). According to the TDDFT calculation, the dipole moments (μe) of the optimized Anti-DPyB and Syn-DPyB in the excited state are 0.3052 and 0.2729 D, respectively. These TDDFT/DFT calculations demonstrate that the Δμ (0.284 D) of Anti-DPyB is larger than that of Syn-DPyB (0.134 D), suggesting that in the excited state, the structural change in Anti-DPyB to form the excimer may induce a relatively large Δμ compared to that of Syn-DPyB. Unlike Anti-DPyB and Syn-DPyB that form the excimer in the excited state, 1,4-di(1-pyrenyl)benzene (Py-Benz-Py), which is a covalently para-linked pyrene (Py) dimer, shows significant different excited-state relaxation dynamics9. Indeed, it was reported that Py-Benz-Py exhibits solvent-dependent ICT dynamics, followed by the twisting motion between Py and phenyl moieties, without intramolecular excimer formation9. The difference in the excited-state relaxation dynamics of Anti-DPyB, Syn-DPyB, and Py-Benz-Py indicates that the molecular structure and configuration play a vital role in their excited-state relaxation dynamics.

Excited-state dynamics dependent on molecular configuration

As shown in Table 2, the TA data for both Anti-DPyB and Syn-DPyB show similar τ1 time constants (2.3–3.7 ps) regardless of solvent polarity. This time scale falls into the well-known time scale for vibrational relaxation. Thus, the earliest kinetic component (τ1) of 3 ps can be interpreted as the intramolecular vibrational relaxation (IVR) from the initially populated local excited state (Franck–Condon state).

After IVR (τ1), the excited molecules in the S1 state have various potential fates, including relaxation to other excited states, such as excimer or triplet excited states, and returning to the ground state via fluorescence. The observation of excimer fluorescence for Anti-DPyB and Syn-DPyB leads to the interpretation that a part of the excited molecules in the S1 state relaxes to the excimer state. As discussed above, the excimer formation in Anti-DPyB should require the rearrangement of the two distant Py moieties to induce the interaction between them. In this regard, the τ2 time constants observed from Anti-DPyB can be interpreted as conformational change. The τ2 time constant (231 ps) of Anti-DPyB in n-hexane is similar to the 323 ps assigned to the twisting motion between the Py and phenyl moiety of Py-Benz-Py9 in a nonpolar solvent. Thus, we attribute the τ2 of Anti-DPyB to excimer formation via S1 → excimer transition accompanying the twisting motion between the Py and phenyl moieties. The faster excimer formation in acetonitrile probably occurs due to the strong hydrophobic interaction between the two Py moieties in a high-polarity solvent. The excimer of Anti-DPyB decays to a long-lived species with τ3 time constants. The TA spectra of the long-lived species are highly similar to the triplet–triplet absorption spectra of free Py derivatives corresponding to the T1 → Tn transition48,49. Specifically, the long-lived species of Anti-DPyB show structured TA spectra, similar to the T1-to-Tn absorption spectra for carbonylpyrenes reported by Rajagopal et al.49. Furthermore, the TA signals of Anti-DPyB observed at >5 ns resemble the T1-to-Tn absorption spectrum of 1-(2-bromophenyl)pyrene measured in dichloromethane (DCM) (Supplementary Fig. S17), although the peak positions are slightly different from each other. Therefore, the long-lived species observed in Anti-DPyB are T1 (free triplet state) or a similar state that produces an absorption spectrum similar to that of T1-to-Tn.

The second fastest time constant (9.7 and 8.0 ps in n-hexane and acetonitrile, respectively) for Syn-DPyB is faster than that for Anti-DPyB by two orders of magnitude. Nevertheless, these time scales are much longer than the typical subpicosecond time scale reported for the excimer formation of prestacked dimeric systems24,30. For example, Hong et al. showed that the excimer state of cofacial stacked perylene bisimide dimer is formed within 200 fs24. Unlike Anti-DPyB, the excimer in Syn-DPyB with a prestacked structure should be rapidly formed with no or less structural rearrangement. For this reason, we ruled out the possibility that the second fastest time constants in Syn-DPyB can be attributed to excimer formation dynamics. Instead, we consider two possibilities: the excited molecules in S1 formed by IVR (3 ps) relax to the excimer state (i) within a subpicosecond (≤200 fs) and (ii) with a time constant comparable to IVR (3 ps). Notably, the TA spectra of a long-lived species formed with the second fastest time constants in Syn-DPyB are highly similar to the triplet–triplet absorption spectra of free Py derivatives corresponding to the T1 → Tn transition, like those of a long-lived species formed with τ3 time constants in Anti-DPyB. For this reason, we denote the time constants of 9.7 and 8.0 ps for Syn-DPyB as τ3, not τ2. In other words, the long-lived species in Anti-DPyB and Syn-DPyB are formed with τ3 time constants and they are attributed to T1 or a similar state.

The free triplet states of a molecule can be generated through an ISC process or the dissociation of (T1T1). It has been accepted that the ISC process in organic molecules with a small spin–orbit coupling occurs with a timescale of 10 ns to 1 ms and the lifetime of the triplet state is longer than the timescale in the order of 1 μs. In contrast to ISC, (T1T1) rapidly forms within the range of 10 fs to 1 ns and has a significantly shorter lifetime than that of the free triplet formed via the ISC process, although the TA spectrum for the (T1T1) is similar to that of the free triplet formed through ISC. A previous study of Py in micelles reported an ISC time constant of 1.7 μs50, which is much longer than the τ3 time constants of 1.75 ns and 495.7 ps for the formation of the long-lived species of Anti-DPyB in n-hexane and acetonitrile, respectively. Meanwhile, the long-lived species for Syn-DPyB in n-hexane and acetonitrile are formed with τ3 time constants of 9.7 and 8.0 ps, respectively, and then relax to other states with τ4 time constants of 6.4 and 4.8 ns, respectively. These lifetimes of the long-lived species of Anti-DPyB and Syn-DPyB are significantly shorter than the triplet lifetimes of Py (9.4–11 ms)51. Thus, we attribute the long-lived species observed from Anti-DPyB and Syn-DPyB to the (T1T1) formed through the first step of SF. Similar examples were reported for the SF dynamics of bis(triisopropylsilylethynyl)-tetracene (Tips-tetracene) in solution, which occur via an excimer with a (T1T1) character34, and four 3,6-bis(thiophene-2-yl)diketopyrrolopyrrole derivatives with different substituents in film, which involve an intermediate excimer-like state52. However, it has been generally accepted that excimer formation inhibits an SF process30,38. Contrary to this generally accepted view, our results demonstrate that the excimers in both Anti-DPyB and Syn-DPyB are rapidly converted to (T1T1). 2ET1 values of Anti-DPyB (4.08 eV) and Syn-DPyB (3.74 eV) are higher than their ES1 values (3.3 and 3.1 eV, respectively), suggesting that the SF processes in Anti-DPyB and Syn-DPyB are endothermic reactions.

Singlet fission

The dissociation dynamics of (T1T1) in the SF process is key to determining the energy conversion efficiency in photoelectric or electrochemical devices, as the faster dissociation of (T1T1) to free triplets is preferable in terms of energy conversion efficiency. The nanosecond TA experiments for Anti-DPyB and Syn-DPyB provide a clue for the free triplet that can be generated by the dissociation dynamics of (T1T1). The nanosecond TA spectra for Anti-DpyB and Syn-DPyB suggest that the triplet species of Anti-DPyB exists at the μs–ms time scale, whereas Syn-DPyB does not exhibit any absorption band in both n- hexane and acetonitrile at this time scale. As shown in Supplementary Figs. S14 and S18, the absorption band around 445 nm of Anti-DPyB measured from the nanosecond TA experiment is similar to the T1-to-Tn absorption spectrum of 1-(2-bromophenyl)pyrene measured in DCM (Supplementary Fig. S17), suggesting that the chemical species of Anti-DPyB observed at the μs–ms time scale are attributed to T1. To further confirm our interpretation, we additionally performed the nanosecond TA experiment for Anti-DPyB in iodomethane to maximize the heavy atom effect. As shown in Supplementary Fig. S18A, Anti-DPyB in iodomethane exhibits an intense absorption band in the range of 350–600 nm at a 1 μs time delay, which is almost identical to those of Anti-DPyB measured in n-hexane and acetonitrile using femtosecond TA spectroscopy. Thus, we suggest that the chemical species of Anti-DPyB observed at the μs–ms time scale are the free triplets generated by the dissociation dynamics of (T1T1). Furthermore, we performed the nanosecond TA experiment for Py, Ph-Py, and Syn-DPyB in iodomethane. Their TA spectra show absorption bands in the range of 350–600 nm at a 1 μs time delay (Supplementary Fig. S18B). The absorption band of Syn-DPyB measured in n-hexane and acetonitrile using femtosecond TA spectroscopy is similar to the T1-to-Tn absorption band of Syn-PDyB measured in iodomethane. In this regard, the last relaxation times (τ4) observed from Anti-DPyB and Syn-DPyB by femtosecond TA experiments are attributed to the dissociation dynamics of (T1T1).

On the other hand, the (T1T1) has two fates: (i) dissociation to free triplets and (ii) decay to the ground state (Fig. 4). The femtosecond TA measurements showed that the time profile for transient absorption bands of Anti-DPyB around 440 nm, which well reflects the relaxation kinetics of (T1T1), shows slow but distinct rising features (Supplementary Fig. S19A), whereas Syn-DPyB shows a relatively fast relaxtion dynamics of (T1T1) to 2T1 and S0 in parallel with a few nanosecond time constants (Supplementary Fig. S19B and Table 2). This difference in the time profiles of Anti-DPyB and Syn-DPyB suggests that the (T1T1) of Anti-DPyB dissociates dominantly into 2T1, whereas the (T1T1) of Syn-DPyB decays mainly to S0 state. Using the data from nanosecond TA spectroscopy, the triplet quantum yield (ΦT) values of Anti-DPyB in n-hexane and acetonitrile are determined to be 44.1% and 17.5%, respectively (Supplementary Methods). Meanwhile, we could not determine the ΦT values of Syn-DPyB in n-hexane and acetonitrile, suggesting that the dissociation reaction of (T1T1) in Syn-DPyB is significantly suppressed compared to Anti-DPyB and the (T1T1) of Syn-DPyB decays mainly to S0 state. The suppressed dissociation of (T1T1) into free triplets in Syn-DPyB is probably due to triplet–triplet annihilation (TTA), which can be facilitated by the proximity of two Py moieties in the excimer state. Several studies of SF suggested that TTA is one of the decay processes of (T1T1)17,53,54,55. The TTA results in two fates: decay to the ground state and upconversion to a higher excited singlet state (excimer). The suppressed dissociation of (T1T1) into free triplets in Syn-DPyB is probably due to TTA. Since the upconversion to the excimer state proceeds on a time scale in the tens of picoseconds, we attribute the τ4 of Syn-DPyB to the (T1T1) → S0 relaxation dynamics.

Fig. 4: Kinetic schemes for photoinduced reactions of Anti-DPyB and Syn-DPyB.
figure 4

S0: ground state, FC: Franck–Condon state, S1: singlet excited state, (T1T1): correlated triplet pair, and T1: free triplet state.

To confirm the SF dynamics, we performed time-resolved electron paramagnetic resonance (TR-EPR) measurements for Anti-DPyB and Syn-DPyB. The X-band (9.728 GHz) perpendicular mode TR-EPR spectra of Anti-DPyB and Syn-DPyB in toluene were measured at 80 K. Supplementary Fig. S20 shows EPR spectra of Anti-DPyB and Syn-DPyB at 128 and 200 ns after photoirradiation. The EPR signals for Anti-DPyB and Syn-DPyB show the narrow peak splitting of 34 and 19 mT around 340 mT (g = 2.002), respectively. In addition to the narrow peak splitting, Anti-DPyB and Syn-DPyB exhibit a large peak splitting of 150 and 115 mT, respectively. The EPR signals for Anti-DPyB and Syn-DPyB are well reproduced by the simulated curve for its triplet (see Supplementary Information). We further confirmed the origins of TR-EPR signals via the nutation experiment for the Q-band (34 GHz) TR-EPR signal of Anti-DPyB (see Supplementary Fig. S21). Although the nutation measurement on the EPR signal of Syn-DPyB was not performed, we speculate that the X-band EPR signal measured from Syn-DPyB arises from triplet species as well. The EPR signals at a few hundred nanoseconds do not show evidence for (T1T1). The absence of EPR signals of (T1T1) for Anti-DPyB and Syn-DPyB at a few hundred nanoseconds is probably due to the shorter lifetimes of (T1T1)s than the temporal resolution (~120 ns) of our TR-EPR system. Syn-DPyB shows a relatively fast decay feature in the time profile for TA bands of 450 nm (Supplementary Fig. S19B). As shown in Table 2, the (T1T1)s of Syn-DPyB in n-hexane and acetonitrile relax to 2T1 and S0 in parallel with time constants of 6.4 and 4.8 ns, respectively, indicating that the lifetime of (T1T1) for Syn-DPyB should be significantly shorter than the temporal resolution (~120 ns) of our TR-EPR system. Meanwhile, we could not precisely determine the lifetime of (T1T1) for Anti-DPyB because of the limited range of investigated delay times in the femtosecond TA measurement. Overall, the EPR data lead us to conclude that the lifetime of (T1T1) for Anti-DPyB should be shorter than the temporal resolution (~120 ns) of our TR-EPR system.

The difference in the SF dynamics of Anti-DPyB and Syn-DPyB may be interpreted in terms of the Eb of (T1T1). According to the kinetic model for SF proposed by Kolomeisly et al., the rates of the formation and dissociation of (T1T1) depend on Eb56. They defined that if Eb > 0, then the (T1T1) state is bound, and if Eb < 0, then it is unbound. The (T1T1) destabilization (Eb < 0) simultaneously results in the fast dissociation of (T1T1) and the slow formation of (T1T1). Based on a spin-lattice model, Abraham and Mayhall predicted that in various covalently linked tetracene or pentacene dimers42, the (T1T1) state of the meta-linked dimer would be unbound with respect to the separated triplets due to the smaller or negative Eb compared to those of the ortho- and para-linked dimers, whereas the (T1T1) states of the ortho- and para-linked dimers would be bound. Similarly, the analysis of wave functions by Chesler et al. showed that the slow formation of (T1T1) in meta-bianthracene may be due to the small or negative Eb, whereas the fast formation of (T1T1) of para-bianthracene results from the larger or positive Eb41. In other words, these theoretical studies predicted that compared to ortho- and para-linked dimers, meta-linked dimers with a smaller Eb would exhibit a relatively slower formation of (T1T1) state and a relatively fast dissociation of (T1T1) into free triplets with a large yield. Indeed, the experimental results for several CLDs agree with the theoretical predictions29,41,42,43. The theoretical calculation results for many ortho-linked CLDs predict that Anti-DPyB and Syn-DPyB, which are also ortho-linked CLDs, form the bound (T1T1) state due to the large or positive Eb. The TA data demonstrate that Syn-DPyB shows a low dissociation reaction of the (T1T1) into free triplets, suggesting that the (T1T1) state of Syn-DPyB is relatively bound compared to that of Anti-DPyB. In contrast to Syn-DPyB, Anti-DPyB shows a significantly slower formation of (T1T1) followed by the dissociation into free triplets, suggesting that the (T1T1) state of Anti-DPyB is unbound. This result indicates that as with meta-linked dimers that show favorable SF dynamics, Anti-DPyB, even if it is an ortho-linked dimer, forms the unbound (T1T1) state due to a small or negative Eb, leading to the efficient dissociation of (T1T1) into free triplets. The theoretical calculation results by Nakano and coworkers demonstrated that compared to ortho- and para-linked pentacene dimers, the electronic coupling between chromophores for the meta-linked pentacene dimer is very low, resulting in the relatively slow formation of (T1T1) and an efficient SF57,58. In this regard, the efficient SF in Anti-DPyB is due to the low electronic coupling owing to the twisted alignment of the two chromophores. This result indicates that the SF dynamics in ortho-linked dimers, which show a significant π-orbital overlap between two chromophores, can be modulated by controlling the molecular configuration. Consequently, our results for Anti-DPyB and Syn-DPyB suggest that the molecular geometry of a CLD plays a critical role in their SF dynamics as well as excimer formation and ICT.

TA spectra analysis with kinetic models

We performed the kinetic analysis of the TA spectra of Anti-DPyB and Syn-DPyB considering various plausible kinetic models. For Anti-DPyB, considering five principal components from SVD analysis (Supplementary Fig. S11) and four-time constants obtained from the fitting of rSVs (Supplementary Fig. S13), we set up the simplest kinetic model with five intermediates assigned to FC, S1, excimer, (T1T1), and 2T1, and the four-time constants connecting them. Considering the relaxation process of the (T1T1) state into 2T1 and S0, the decay process from (T1T1) to S0 was added to the kinetic model. The resulting kinetic model is Kinetic Model (1) in Fig. 4. Details regarding Kinetic Model (1) are provided in Supplementary Information.

Whereas Kinetic Model (1) could explain the TA data well, it could not explain the emission behavior. The fluorescence decay profiles showed two-time constants assigned to the fluorescence lifetimes of the Py monomeric unit and excimer. By adding these two fluorescence decay times to Kinetic Model (1), we set up a different kinetic model (Kinetic Model (2) in Fig. 4). In this kinetic model, we also included the backreaction from (T1T1) to the excimer for the following reason: As shown in the inset of Fig. 2A, the rising time of 1.24 ns in the fluorescence decay profile for Anti-DPyB in n-hexane is approximately five times larger than the time constant (0.23 ns) corresponding to the S1 → excimer transition determined from femtosecond TA experiments. This difference indicates that the observed excimer fluorescence is not prompt emission but delayed emission, suggesting equilibrium between the excimer state and the (T1T1) state. Details regarding Kinetic Model (2) are provided in Supplementary Information. Figure 5 shows the five SADS curves and population changes for five intermediates (FC, S1, excimer, (T1T1), and 2T1) obtained from the kinetic analysis for Kinetic Model (2), and Supplementary Fig. S24 shows the experimental TA spectra, the best-fit spectra, and the residuals between them for Anti-DPyB in n-hexane and acetonitrile obtained from the kinetic analysis using Kinetic Model (2). The residuals between the experimental and the best-fit spectra are small, suggesting that the measured TA spectra for Anti-DPyB are well constructed as a linear combination of the five SADS curves according to the employed kinetic model. As the fit qualities of both Kinetic Models (1) and (2) are comparable, fit qualities alone could not be used to determine which kinetic model is more accurate (Supplementary Figs. S22S24). As discussed above, however, Kinetic Model (2) is preferred because it is more consistent with the emission data.

Fig. 5: Species-associated difference spectra and population changes of intermediates for Anti-DPyB obtained from the kinetic analysis for Kinetic Model (2) in n-hexane and acetonitrile.
figure 5

A, B Species-associated difference spectra in A n-hexane and B acetonitrile. C, D Population changes of intermediates in C n-hexane and D acetonitrile. The solid lines are the concentrations obtained from the kinetics analysis. The open circles represent the measure time delays.

For Syn-DPyB, considering the three exponential time constants obtained from the exponentials fitting of rSVs (Supplementary Fig. S13), the two-time constants and emission quantum yields from the emission experiments (Table 1), and the four principal components from SVD analysis results (Supplementary Fig. S12), we can preferentially set up a kinetic model with five-time constants and four intermediates. Simultaneously, the TR-EPR signal for Syn-DpyB indicates that the (T1T1) of Syn-DPyB also dissociates to free triplets. Based on this result, the (T1T1) → 2T1 transition was included in the kinetic model. We also included the backreaction from (T1T1) to the excimer as in Anti-DPyB. Consequently, as in Anti-DPyB, we used the kinetic model with five intermediates assigned to FC, S1, excimer, (T1T1), and 2T1 (see Fig. 4). In the case of Syn-DPyB, it is noteworthy that the τ2 time constant corresponding to the S1 → excimer transition was not observed in Syn-DPyB (Table 2). The excimer of Syn-DPyB with a pre-stacked structure likely forms quickly within a subpicosecond (≤200 fs) or with a time constant comparable to the IVR (3 ps). In this regard, we considered two kinetic models (Kinetic Models (3) and (4) in Fig. 4). In the former kinetic model, the S1 → excimer transition occurs in a subpicosecond (≤200 fs), and in the other kinetic model, the S1 → excimer transition occurs with a time constant comparable to the IVR (3 ps). As shown in Supplementary Figs. S25 and S26, both Kinetic Models (3) and (4) show small residuals between the experimental and the best-fit spectra, suggesting that the measured TA spectra for Syn-DPyB are well constructed as linear combinations of the five SADS curves according to both kinetic models, and that fit qualities alone cannot be used to determine which kinetic model is better. Nevertheless, the SADS curves from the two kinetic models are different and provide clues regarding which kinetic model is more accurate. Whereas the SADS for the S1 state from Kinetic Model (4) is positive (Fig. 6), that from Kinetic Model (3) is strongly negative (Supplementary Fig. S25), which is not possible for excited state absorption (ESA) from the S1 state, so Kinetic Model (3) could be ruled out. In other words, the kinetic analysis suggests that the S1 → excimer transition in Syn-DPyB occurs with a time constant comparable to the IVR (3 ps). Figure 6 shows the five SADS curves and population changes for five intermediates (FC, S1, excimer, (T1T1), and 2T1) for Syn-DPyB obtained from the kinetic analysis for Kinetic Model (4) in n-hexane and acetonitrile. The small differences among SADS curves of Syn-DPyB are probably due to the rigid structure of Syn-DPyB in the ground and excited states and the small energy differences between the states.

Fig. 6: Species-associated difference spectra and population changes of intermediates for Syn-DPyB obtained from the kinetic analysis for Kinetic Model (4) in n-hexane and acetonitrile.
figure 6

A, B Species-associated difference spectra in A n-hexane and B acetonitrile. C, D Population changes of intermediates in C n-hexane and D acetonitrile. The solid lines are the concentrations obtained from the kinetics analysis. The open circles represent the measure time delays.

The kinetic analyses demonstrate that in Anti-DPyB, the (T1T1) formed through the excimer slowly dissociates into free triplets. On the other hand, it was also proposed that the intermolecular and intramolecular SF dynamics can rapidly occur with a direct process from the S1 state to the free state due to strong coupling between the S1 state and the free triplet state59,60,61. For example, Dover et al. suggested that the SF channel is dominated by a direct mechanism from the S1 state and the formation of the excimer state inhibits efficient SF dynamics61. Thus, we also explored the possibility that our data from Anti-DPyB and Syn-DPyB can be explained using the same direct mechanism by applying the kinetic analysis with the reaction schemes compatible with the direct SF mechanism (see Supplementary Fig. S27). Those reaction schemes involving a direct SF process did not satisfactorily reproduce the measured TA spectra for Anti-DPyB in n-hexane and acetonitrile or yielded an unphysical SADS curve (Supplementary Information and Supplementary Fig. S28). This result indicates that the coupling between the S1 state and the free triplet state in CLDs such as Anti-DPyB and Syn-DPyB is weaker than in the molecules that showed such direct SF processes, although further systematic studies are needed to confirm this hypothesis.

Conclusions

To understand the ultrafast excited-state relaxation dynamics of intramolecular SF materials such as a CLD, we elucidated the ultrafast excited-state relaxation dynamics of covalently linked pyrene dimers Anti-DPyB and Syn-DPyB. In the excited state, Anti-DPyB, in which two Py moieties are oriented in a twisted configuration, forms the excimer through a conformational change with time constants of 231 and 24.3 ps in n-hexane and acetonitrile, respectively (Fig. 7). Syn-DPyB, with a pre-stacked configuration, rapidly forms the excimer without any conformational change with a time constant of 3 ps. Our results also demonstrated that the excimer emissions observed from Anti-DPyB and Syn-DPyB are not prompt but delayed emissions. The time-resolved spectroscopic results showed that the resulting excimers rapidly relax to the (T1T1) state, suggesting that the (T1T1)’s of Anti-DPyB and Syn-DPyB are formed through the excimer state. The (T1T1) of Anti-DPyB dissociates to free triplets as the end product, completing SF, whereas the dissociation reaction of (T1T1) in Syn-DPyB is significantly suppressed compared to Anti-DPyB. This means that Anti-DPyB forms unbound (T1T1), resulting in efficient SF dynamics, whereas Syn-DPyB forms bound (T1T1). The suppressed dissociation of (T1T1) into free triplets in Syn-DPyB is probably due to the triplet–triplet annihilation. This result differs from the prediction based on theoretical studies proposing that meta-linked dimers with a smaller Eb, compared with ortho- and para-linked dimers, exhibit efficient SF dynamics. This finding suggests that the efficiency of SF dynamics in CLDs cannot be predicted solely by the substitution position of the chromophore in a CLD. As Anti-DPyB and Syn-DPyB have relatively more distorted structures than previously studied CLDs (Fig. 1A), the orbital interaction likely has a much greater effect on their SF dynamics than the substitution position. Indeed, our data show that the relatively more efficient SF in Anti-DPyB compared to Syn-DPyB is caused by the relatively low electronic coupling between two chromophores owing to their twisted alignment. This result indicates that the SF dynamics in ortho-linked dimers, which generally show a significant π-orbital overlap between two chromophores, can be modulated by the control of the molecular configuration, consequently suggesting that the molecular geometry of a CLD plays a critical role in its SF dynamics, excimer formation, and ICT.

Fig. 7: Schematics for proposed relaxation dynamics of Anti-DPyB and Syn-DPyB.
figure 7

Excited-state relaxation dynamics of A Anti-DPyB and B Syn-DPyB. IVR stands for intramolecular vibrational relaxation.

Methods

Synthetic procedures for Anti-DPyB and Syn-DPyB

See Supplementary Methods and Supplementary Fig. S1 in the Supplementary Information.

Characterization for Anti-DPyB and Syn-DPyB

See Supplementary Figs. S2S5 for 1H- and 13C-NMR spectra and Supplementary Fig. S6 for GC–MS data.

Preparation of single crystal Anti-DPyB

See Supplementary Methods.

X-ray crystal structure analysis

See Supplementary Methods, Supplementary Tables S1S6, and Supplementary Fig. S7 for the thermal ellipsoid plot of Anti-DPyB.

Steady-state and time-resolved spectroscopic measurements

See Supplementary Methods, Supplementary Fig. S8 for fluorescence excitation spectra, Supplementary Fig. S9 for emission spectra of Py, Ph-Py, Anti-DPyB, and Syn-DPyB at 77 K, Supplementary Fig. S10 for femtosecond TA spectra of Anti-DPyB and Syn-DPyB, Supplementary Fig. S14 for nanosecond TA spectra of Anti-DPyB, Supplementary Fig. S16 for the concentration dependence of the emission spectra of Anti-DPyB and Syn-DPyB, Supplementary Fig. S17 for nanosecond TA spectra of 1-(2-bromophenyl)pyrene, Supplementary Fig. S18 for nanosecond TA spectra of Py, Ph-Py, Anti-DPyB, and Syn-DPyB, and Supplementary Fig. S19 for time profiles for transient absorption bands of Anti-DPyB and Syn-DPyB.

Cyclic voltammograms of Py, Anti-DPyB, and Syn-DPyB

See Supplementary Fig. S15.

Triplet quantum yield (Φ T) of Anti-DPyB

See Supplementary Methods.

Time-resolved EPR spectroscopy

See Supplementary Methods and Supplementary Figs. S20 and S21.

Singular value decomposition (SVD) analysis

Details regarding SVD analysis are provided in the Supplementary Methods. The SVD analysis results (Supplementary Figs. S11 and S12) and the fits of rSVs (Supplementary Fig. S13) are provided in the Supplementary Methods.

Kinetic analysis

See Supplementary Methods. The results for the kinetic analysis of the TA spectra of Anti-DPyB and Syn-DPyB are given in Supplementary Figs. S22S26.

Direct SF mechanisms from the S1 state

See Supplementary Methods and Supplementary Figs. S27 and S28.

Cartesian coordinates from computational studies

See Supplementary Data 2.