Excited-state triplet−doublet spin dynamics in compact donor–acceptor–radical triads

Abstract

Compact naphthalimide-based donor–acceptor–radical triads, together with their corresponding radical-free donor–acceptor dyads, were prepared as a tunable platform for manipulating long-lived excited-state spin dynamics in closely coupled organic multispin systems. Structural analysis and DFT calculations support a compact, orthogonal geometry, which preserves the distinct functions of the donor, acceptor, and radical fragments while keeping them close enough for efficient excited-state interactions. By combining optical spectroscopy with time-resolved electron paramagnetic resonance (TREPR), we observed a consistent progression across the series. The sequence starts from the localized triplet state of the naphthalimide chromophore and extends to the donor-containing dyads, where triplet locally excited (³LE) and triplet charge-separated (³CS) states coexist. Changing the donor from phenothiazine to phenoxazine increases the charge-transfer character, and increasing solvent polarity further stabilizes the charge-separated state. When TEMPO is attached, the proximal radical shortens the excited-state lifetimes, modifies triplet-state formation, and produces a distinct spin-polarization pattern, consistent with electronic coupling between the TEMPO moiety and the photoexcited naphthalimide chromophore. TREPR measurements show that the radical-containing dyads and triads are best described as triplet–doublet (TD) systems in the D/J-dominant regime and that the donor affects both the sign and the magnitude of the triplet–radical exchange interaction. Comparison of the matched radical-free and radical-containing series shows that donor identity sets the balance between localized and charge-transfer excited states, solvent polarity controls the stabilization of the charge-separated state, and the proximal radical governs spin-population redistribution and relaxation. These findings establish a unified picture of intersystem crossing pathways and excited-state dynamics in compact organic multispin assemblies.

The charge separation and intersystem crossing of electron donor–acceptor–radical triads were studied with time-resolved transient absorption and electron paramagnetic resonance spectroscopy.

Introduction

In organic donor–acceptor (D–A) systems, charge separation (CS), charge recombination (CR), and intersystem crossing (ISC) play a central role in determining the electron and spin dynamics of photoexcited states.^1–3 Control over the charge-separated states and locally excited triplet (³LE) states is particularly important for the design of organic photofunctional materials, including compact D–A dyads capable of formation of long-lived CS states and thermally activated delayed fluorescence (TADF) emitters.^4–7 More broadly, the ability to tune the lifetime and spin configuration of CS states is relevant to artificial photosynthesis,^8,9 organic electronics,^10,11 and molecular spintronics and quantum information science.^12–14

Electron donor–acceptor–radical (D–A–R˙) triads provide an attractive platform for studying multispin photophysics because photoexcitation generates a spin-correlated radical pair that can interact with a stable radical. Depending on the exchange and dipolar interactions, different spin-dependent processes can operate in such systems.¹⁵ Previous work on weakly coupled triads with long spacers showed that radical attachment can enhance singlet–triplet mixing and modulate charge-recombination kinetics, whereas in more compact systems, direct interaction of the radical with excited triplet states can lead to high-spin state.^16–21 However, the interaction of a stable radical with a long-lived ³CS state remains much less explored than its interaction with a locally excited triplet, largely because long-lived ³CS states are difficult to access in simple molecular architectures.

In our previous studies, we showed that long-lived ³CS states can be generated in compact dyads by increasing the donor–acceptor exchange interaction and using the ³LE state as a precursor.^22,23 These studies further established that a ³LE state close in energy to the CS states is essential for TADF, whereas a larger ³LE/CS energy splitting allows access to significantly longer-lived ³CS states.^6,24,25

The chromophores naphthalimide (NI), phenothiazine (PTZ), and phenoxazine (PXZ) have been widely used as compact D–A building blocks for photoinduced electron transfer and spin-dynamics studies.^24,26–30 On this basis, we designed compact NI-based triads by incorporating a stable TEMPO radical into the previously studied PTZ-NI and PXZ-NI dyads, for which close-lying ³CS and ³LE states had already been identified.^22,24 This molecular design differs from the more common D–A–R˙ architectures with long donor–acceptor separation and weakly coupled spin-correlated radical pairs because here the donor, acceptor, and radical are brought into close proximity.

As a reference, we also examine NI-TEMPO, whose basic photophysical properties have been reported previously.^31,32NI-TEMPO provides the simplest radical-containing member of the series and thus links the intrinsic NI-centered triplet photophysics to the more complex donor-containing triads. During the preparation of this manuscript, the identical PXZ-NI-TEMPO triad and the corresponding radical-free PXZ-NI dyad were reported by Gorgon et al., who discussed their excited-state behavior in relation to the energetic balance between locally excited and charge-separated triplet states,³³ highlighting the growing interest in compact radical-containing TADF-type architectures. That study analyzed the transient EPR (TREPR) spectra within a relatively strongly coupled spin framework (exchange-dominated), whereas the present work addresses the moderately coupled regime in which zero-field splitting remains dominant over exchange coupling (D/J-dominant regime).^34,35 The spectral features discussed here are consistent with the behavior expected for chromophore–radical conjugates in the D/J-dominant regime, where the TREPR spectra differ qualitatively from both the weak- and strong-coupling limits. Additional near-center lines separated by approximately 4J may appear, and their intensity can serve as characteristic signatures of radical-enhanced ISC.³⁵

Steady-state spectroscopy, femtosecond and nanosecond transient absorption, time-resolved EPR, and DFT calculations are combined to develop a unified physical picture for the entire NI-based series, including NI, PTZ/PXZ-NI, and their corresponding TEMPO-substituted analogues. The optical data show that replacing PTZ with PXZ shifts the balance between the locally excited and charge-transfer triplet states, while increasing solvent polarity stabilizes the charge-separated state and shifts the system from mixed ³LE/³CS behavior in nonpolar media toward a more distinct ³CS state in polar solvents. TREPR provides a complementary spin-level description: upon TEMPO incorporation, the photoexcited states are more appropriately described as triplet–doublet spin systems operating in the D/J-dominant regime. DFT supports the structural and electronic basis of this picture by reproducing the compact, nearly orthogonal geometry of the triads, showing that the ground-state unpaired spin is localized on TEMPO. In the quartet state, the spin density redistributes over the NI and donor fragments, with its extent and distribution varying in different environments. Taken together, these results show that donor identity controls the relative contributions of local and charge-transfer triplet states, solvent polarity stabilizes the charge-separated state, and the proximal radical introduces spin mixing that changes the population distribution, decay, and spin polarization of the long-lived excited states.

Synthesis and structural characterization

The target compounds NI-TEMPO, PTZ-NI-TEMPO, and PXZ-NI-TEMPO compounds were prepared as a chemically uniform series to enable direct comparison of donor and radical effects within the same compact NI-based architecture. Their synthesis is summarized in Scheme 1. NI-TEMPO was obtained by condensation of 4-amino-TEMPO with 1,8-naphthalic anhydride in N,N-dimethylformamide (DMF)/acetic acid, whereas PTZ-NI-TEMPO and PXZ-NI-TEMPO were synthesized from a common brominated naphthalimide precursor by Pd-catalyzed Buchwald–Hartwig C–N coupling with phenothiazine or phenoxazine, respectively. The molecular structures were verified by IR and HRMS analysis (Fig. S7–S12), and PTZ-NI-TEMPO was additionally characterized by single-crystal X-ray diffraction.

Scheme 1. Synthesis of NI-TEMPO, PTZ-NI-TEMPO, and PXZ-NI-TEMPO and the reference compounds. (a) DMF, acetic acid, 100 °C, 4 h, yield: 30%; (b) phenothiazine, Pd(OAc)₂, [(t-Bu)₃PtH]BF₄, NaOtBu, dry toluene, reflux 8 h, yield: 46%; (c) same conditions as (b) using phenoxazine, yield: 52%.

Single crystals of PTZ-NI-TEMPO suitable for X-ray diffraction were obtained by slow evaporation of a saturated tetrahydrofuran/n-hexane (THF/n-HEX) solution at 5 °C. The ORTEP structure confirms that the molecule adopts the intended compact and largely orthogonal geometry (Fig. 1). The PTZ and NI fragments form a dihedral angle of 87.9°, while the TEMPO and NI moieties are also strongly twisted, with a dihedral angle of 83.5°, consistent with steric constraints imposed by the imide carbonyl groups and the TEMPO fragment. At the same time, the distances between the NI acceptor and the adjacent units remain short, amounting to 5.0 Å for NI–PTZ and 5.7 Å for NI-TEMPO. The centroid-to-centroid distance between the PTZ and TEMPO units is 10.6 Å, and the PTZ fragment exhibits the expected butterfly distortion with a tilt angle of 26.4°. This compact spatial arrangement is achieved without strong ground-state planarization of the π-system. The representative structure provides the structural basis for the subsequent photophysical comparison across the series.

Fig. 1. ORTEP view of the molecular structure of PTZ-NI-TEMPO, determined by single-crystal X-ray diffraction analysis. Thermal ellipsoids are set at 50% probability. CCDC number 2354090 contains detailed information.

Optical spectroscopy: local and CS states

UV-vis absorption and luminescence properties

The UV-vis absorption spectra show that TEMPO attachment only weakly perturbs the ground-state electronic structure. For PTZ-NI-TEMPO, the spectral profile remains close to that of PTZ-NI, and the weak CT absorption band near 450 nm is only slightly enhanced (Fig. 2). Hereafter, CT is used to describe spectral features or partial charge-transfer character, whereas CS denotes a charge-separated state in which the electron is predominantly localized on NI and the hole on the PTZ or PXZ donor, as evidenced by the corresponding radical-ion absorption bands. In the PXZ series, the effect is more pronounced: the CT band centered at ca. 500 nm increases in intensity after TEMPO incorporation, indicating somewhat stronger donor–acceptor electronic coupling in the ground state.

Fig. 2. UV-vis absorption spectra of (a) PTZ-NI, NI-TEMPO and PTZ-NI-TEMPO and (b) PXZ-NI, NI-TEMPO and PXZ-NI-TEMPO in n-HEX. The inset in (a) is a magnified view of the CT absorption band. c = 1.0 × 10⁻⁵ M, 20 °C.

The steady-state emission data further show that the emissive CT state is only weakly perturbed by radical attachment (Fig. 3 and S15). PTZ-NI displays a structureless CT emission band centered at 622 nm, whereas PTZ-NI-TEMPO shows only a slight red shift to 627 nm, and the fluorescence quantum yields remain low in both cases, 3.2% and 1.8%, respectively. The radiative decay rate constants (k_r) for PTZ-NI and PTZ-NI-TEMPO are 2.5 × 10⁵ s⁻¹ and 3.1 × 10⁵ s⁻¹, respectively, while the non-radiative rates (k_nr) are 7.6 × 10⁶ s⁻¹ and 1.7 × 10⁷ s⁻¹. The close k_r values indicate that TEMPO attachment does not significantly alter the intrinsic radiative transition probability, whereas the marked increase in k_nr shows that the reduced quantum yield arises primarily from enhanced non-radiative deactivation, consistent with quenching of the fluorescence by the TEMPO radical.

Fig. 3. (a) Fluorescence emission spectra of PTZ-NI-TEMPO under different atmospheres, (b) corresponding decay traces at 615 nm, (c) fluorescence emission spectra of PTZ-NI under different atmospheres and (d) corresponding decay traces at 600 nm. In aerated and deaerated n-HEX, λ_ex = 340 nm, c = 1.0 × 10⁻⁵ M, and 20 °C. In (a) and (c), the emission peak marked by the asterisks indicates the Raman scattering peak.

Solvent-dependent fluorescence spectra of PTZ-NI-TEMPO also retain the same general solvatochromic behavior as PTZ-NI, consistent with the CT nature of the emissive state and indicating that TEMPO does not substantially change the polarity-sensitive CT fluorescence. These steady-state fluorescence data are collected in Fig. S15.

A much stronger radical effect appears in the luminescence kinetics. Under aerated and deaerated conditions, the PTZ samples show that the prompt fluorescence lifetime changes only slightly upon TEMPO attachment, from 0.021 µs to 0.018 µs, whereas the delayed fluorescence lifetime shortens substantially, from 1.92 µs to 0.42 µs. Thus, Fig. 3 directly shows that the delayed component is oxygen-sensitive and that TEMPO perturbs mainly the delayed, triplet-related channel, while the prompt emissive channel is only weakly affected. The same pattern is evident for the PXZ samples from Table 1 and Fig. S18, where the prompt lifetime remains nearly unchanged, 0015 µs versus 0.016 µs, while the delayed lifetime decreases from 15.8 µs to 2.0 µs. In the PXZ samples, this is accompanied by a strong increase in both the delayed fraction, from 4.6% to 75.2%, and the singlet oxygen yield, from 6.8% to 34.4%. The solvent-dependent singlet oxygen photosensitizing data are compiled in Table S1. Considering that virtually the same cyclohexyl moieties are attached at the imide N atom, these lifetime differences are unlikely to arise simply from the alkyl substituents themselves.³⁶ Unlike weakly coupled three-spin systems, the present compact triads are more reasonably considered in the strong-coupling regime with a large J_DA. In these systems, the donor–acceptor radical pair possesses substantial exchange coupling, such that the dominant spin interactions involve coupling of the TEMPO radical with the triplet radical-pair manifold, rather than direct magnetic-field-controlled ¹CS/³CS interconversion, as proposed for weakly coupled systems.³⁷

Table 1. Photophysical parameters of the compounds^a.

	λ abs [nm]	ε b	λ em [nm]	τ F c [µs]	k r d	k nr e	Φ F f [%]	Φ Δ g [%]	τ T h [µs]
PTZ-NI	330/452	1.90/0.06	622	0.021 (94.4%)/1.92 (5.6%)	25.1	7.6	3.2	10.2	2.44
PTZ-NI-TEMPO	333/450	1.98/0.07	627	0.018 (90%)/0.42 (10%)	30.9	16.9	1.8	2.3	0.43
PXZ-NI	330/499	2.22/0.09	582	0.015 (95.4%)/15.8 (4.6%)	4.0	1.3	3.0	6.8	13
PXZ-NI-TEMPO	330/501	2.09/0.13	596	0.016 (24.8%)/2.0 (75.2%)	1.5	0.65	2.3	34.4	1.8
NI-TEMPO	329	1.63	360	0.004 (42.5%)/0.028 (57.5%)	3.4	56.1	0.06	5.4	0.23

^aIn n-hexane (c = 1.0 × 10⁻⁵ M).

^bMolar absorption coefficient at the absorption maxima, ε: 10⁴ M⁻¹ cm⁻¹.

^cFluorescence lifetime under a N₂ atmosphere in n-hexane.

^dRadiative decay rate constant k_r = Φ/τ, in [10⁴ s⁻¹].

^eNon-radiative decay rate constant k_nr = (1 − Φ)/τ, in [10⁶ s⁻¹].

^fAbsolute fluorescence quantum yield, error ± 0.1.

^gSinglet oxygen quantum yield (¹O₂) with Ru(bpy)₃[PF₆]₂ as standard (Φ_Δ = 57% in DCM) in n-hexane (λ_ex = 330 nm, A = 0.23).

^hTriplet state lifetime measured by nanosecond transient absorption in deaerated n-hexane.

Femtosecond transient absorption spectroscopy

The fs-TA data show that both solvent polarity and TEMPO attachment strongly influence the excited-state relaxation pathways and the nature of the final triplet states. For PTZ-NI (Fig. 4a–c), photoexcitation first populates the ¹NI* state, as also established for related NI-based dyads.^24,38 In cyclohexane (CHX) (Fig. 4a), the ¹NI* state decays within 0.78 ps to form a hot ¹CS state, followed by vibrational and solvent relaxation to a relaxed ¹CS state in 166.8 ps. The final long-lived species, appearing on the nanosecond timescale, exhibits a weak broad absorption around 470 nm together with strongly diminished CS features, indicating that the terminal state is dominated by a ³LE state with only minor residual CS character.

Fig. 4. EADS obtained from global fitting of fs-TA spectral data for PTZ-NI (a–c) and PTZ-NI-TEMPO (d–f) in different polarity solvents: (a and d) CHX (E_T (30) = 30.9 kcal mol⁻¹), (b and e) THF (E_T (30) = 37.4 kcal mol⁻¹), and (c and f) ACN (E_T (30) = 46.0 kcal mol⁻¹). λ_ex = 340 nm, c = 1.0 × 10⁻⁵ M, 20 °C.

In THF and acetonitrile (ACN) (Fig. 4b and c), increasing solvent polarity stabilizes the CS state. The initially formed ¹CS state evolves into a long-lived ³CS state within 110.3 ps in THF and 12.7 ps in ACN. Simultaneously, the spectral contribution associated with the LE triplet feature near 470 nm becomes much weaker or essentially absent, whereas the NI radical-anion absorption remains prominent. These observations indicate that polar solvents favor formation and stabilization of the ³CS state, progressively shifting the system from mixed ³LE/³CS behavior in CHX toward a predominantly ³CS state in THF and ACN. Thus, Fig. 4a–c collectively show both the formation dynamics of the ¹CS state and the solvent-dependent evolution of the final triplet manifold from a ³LE state to ³CS character.

For PTZ-NI-TEMPO (Fig. 4d–f), the early-time fs-TA behavior remains qualitatively similar, indicating that photoexcitation again first generates the ¹NI* state followed by rapid CS. In CHX (Fig. 4d), the ¹NI* state decays within 0.52 ps to produce a hot ¹CS state, which evolves further on the nanosecond timescale. Compared with PTZ-NI, however, the long-lived spectrum retains stronger NI radical-anion absorption while the PTZ radical-cation band decreases substantially, accompanied by development of a weak feature around 470 nm. These spectral characteristics indicate that the terminal state is no longer predominantly localized ³LE, but instead corresponds to a mixed triplet manifold with dominant ³CS character and a smaller ³LE contribution.

In THF and ACN (Fig. 4e and f), incorporation of TEMPO further accelerates the ¹CS → ³CS conversion, which occurs within 64 ps and 10.6 ps, respectively. The resulting long-lived spectra are dominated by persistent CS absorption with minimal evidence of ³LE features, consistent with formation of a predominantly ³CS state. Overall, Fig. 4d–f show that TEMPO incorporation facilitates ³CS state formation and shifts the triplet-state equilibrium toward stronger CS character, most notably in the nonpolar solvent, where PTZ-NI alone preferentially forms a localized triplet state.

The PXZ series follows the same general trend but is more strongly biased toward the triplet channel with stronger CT character. From Table 1, the prompt fluorescence lifetimes remain nearly unchanged after TEMPO incorporation, whereas the delayed component shortens strongly and the delayed fraction rises from 4.6% to 75.2%. The complete fs-TA data for PXZ-NI and PXZ-NI-TEMPO are given in Fig. S22–S27. These data show that in CHX, the characteristic ISC time decreases from 5.8 ns for PXZ-NI to 4.2 ns for PXZ-NI-TEMPO, with considerably faster conversion in polar solvent. To isolate the direct role of TEMPO in ISC enhancement, the reference NI-TEMPO was also examined. As shown in Fig. S28–S30, the ¹NI* → ³NI* transition occurs in 168.9 ps in CHX, 24.7 ps in THF, and 8.28 ps in ACN, demonstrating a pronounced polarity dependence. Overall, replacing PTZ by PXZ shifts the balance further toward CT-dominated triplet behavior, while TEMPO provides an additional pathway that accelerates triplet-state formation. Previous fs-TA studies on D–A–R˙ systems have shown that stable radicals can either weakly perturb or substantially accelerate spin conversion, depending on the coupling regime and state energetics.^18,20,39 In the present compact NI-based triads, the dominant effect is acceleration of ¹CS → ³CS conversion rather than prolongation of the long-lived CS manifold.

Nanosecond transient absorption spectroscopy

The ns-TA data extend this picture to the longer-lived states. Reference NI-TEMPO supports the direct radical effect on the NI-localized triplet: as shown in Fig. 5a and b, it exhibits a characteristic ³LE ESA centered near 460 nm with a triplet lifetime of 0.40 µs, which decreases to 0.11 µs under aerated conditions; no ns-TA signal is observed in ACN. Under otherwise similar conditions, pristine NI shows a comparable triplet ESA but a much longer lifetime of 34.3 µs (Fig. S31).

Fig. 5. Nanosecond transient absorption spectra of (a) NI-TEMPO, (c) PTZ-NI and (e) PTZ-NI-TEMPO in deaerated n-HEX (polarity E_T (30) = 31.0 kcal mol⁻¹). Corresponding decay traces are shown for (b) NI-TEMPO at 465 nm, (d) PTZ-NI at 430 nm and (f) PTZ-NI-TEMPO at 430 nm. λ_ex = 355 nm, c = 1.0 × 10⁻⁵ M, 20 °C.

For PTZ-NI (Fig. 5c and d), the spectrum in n-HEX contains three characteristic bands near 430, 460, and 530 nm, assigned to NI⁻˙, ³NI*, and PTZ⁺˙, respectively. This combination indicates that the lowest triplet manifold in nonpolar media contains both ³CS and ³LE contributions. In more polar solvents (Fig. S32), a distinct ³CS state is observed instead; the lifetimes decrease with solvent polarity and are 1.70 µs in CHX, 2.6 µs in n-HEX, 0.32 µs in THF, and 0.16 µs in ACN. Thus, these measurements show directly that the long-lived state changes from mixed ³LE/³CS behavior in nonpolar media to a more distinct ³CS state in polar solvents. This interpretation is consistent with earlier work on NI-based dyads, where close-lying local and CS triplet states and spin-vibronic coupling were identified as key features of the TADF-active manifold.^{22,24,25,40,41}

For PTZ-NI-TEMPO (Fig. 5e and f), the same qualitative solvent dependence is observed. In CHX and n-HEX, the triad shows a mixed ³CS/³LE state, whereas in polar solvents a distinct ³CS state is detected (Fig. S33), with prominent bands at 420 and 510 nm assigned to NI⁻˙ and PTZ⁺˙, respectively. The main difference relative to PTZ-NI is kinetic: the triplet-state lifetime is shorter throughout the series after TEMPO attachment, with reported values of 0.43 µs in CHX, 0.42 µs in n-HEX, 0.24 µs in THF, and 0.10 µs in ACN. Fig. 5e and f therefore provide the nanosecond evidence that TEMPO does not merely accelerate triplet formation, but also shortens the lifetime of the long-lived triplet/charge-transfer states. The complete ns-TA data for the PXZ-NI dyad and PXZ-NI-TEMPO triad in different polarity solvents are collected in Fig. S34–S37. These show the same qualitative trend, but the quenching is even stronger, especially in nonpolar media: PXZ-NI-TEMPO gives 2.4, 1.8, 0.16, and 0.11 µs in CHX, n-HEX, THF, and ACN, respectively, compared with 9, 13, 0.23, and 0.14 µs for PXZ-NI. Previous radical-containing D–B–A systems showed that radical attachment can prolong the CT lifetime by enhancing radical pair ISC.¹⁸ In contrast, the present compact triads show the opposite behavior: after radical attachment, the long-lived manifold is quenched rather than prolonged.

To evaluate the quenching effect of the TEMPO radical on the ³LE lifetime, ns-TA spectra of PTZ-NI and PXZ-NI dyads were recorded in n-HEX upon incremental addition of TEMPO-NH₂ (Fig. S38–S40). In both systems, the triplet–triplet absorption bands decrease systematically with increasing TEMPO concentration, indicating efficient quenching of the ³LE state by the radical. The derived quenching rate constant is ∼9.4 × 10⁹ M⁻¹ s⁻¹ with an efficiency of ∼39% (Table S2). Linear Stern–Volmer behavior confirms dynamic bimolecular quenching of the ³LE state by TEMPO-NH₂ in solution. This experiment demonstrates that interaction with a nitroxide radical can quench the ³LE state; however, the fixed chromophore–radical geometry of the covalent triads may result in a different, intramolecular triplet–radical quenching mechanism.

Time-resolved EPR spectroscopy

Radical-free series

Time-resolved EPR spectroscopy allows studying the low-temperature (80 K) spin dynamics of the excited states. The TREPR spectrum of NI (Fig. 6a) provides the reference pattern for the radical-free part of the series (Fig. 6). Its spectral shape is characteristic of a localized triplet state of the naphthalimide chromophore and can be reproduced with zero-field splitting (ZFS) parameters D = 2.48 GHz and E = −0.13 GHz. The parameters p_Tz and p_Txy describe the polarization formed by the triplet mechanism and are defined in the basis of the ZFS tensor (molecular frame). The polarization pattern of the NI spectrum is consistent with preferential population of the in-plane triplet components and the out-of-plane orientation of the principal ZFS axis. The simulation parameters of the TREPR spectra are summarized in Table 2. The donor-containing compounds PTZ-NI and PXZ-NI display substantially more complex TREPR spectra than NI (Fig. 6b and c). Unlike NI, they cannot be described satisfactorily by a single localized triplet state. This is most evident for PTZ-NI, where the spectrum changes markedly with delay after the laser flash, including inversion of the sign of several lines. The spectra can be reproduced only when two metastable triplet states are included, assigned to ³LE and ³CS, respectively. The ³LE component has ZFS parameters close to those of the NI reference, whereas the ³CS component shows a much smaller splitting and an opposite sign of D value. For the localized NI triplet, the sign of D was assigned based on the dominant electron–electron dipolar contribution to the ZFS. For a planar chromophore with the principal ZFS axis normal to the molecular plane, this contribution is expected to yield D > 0. By contrast, for a charge-transfer triplet in which the unpaired electrons are localized on spatially separated molecular fragments, the dominant dipolar contribution is expected to yield D < 0, as discussed, for example, in ref. 42.

Fig. 6. TREPR spectra of (a) NI, (b) PTZ-NI and (c) PXZ-NI recorded at several intervals after the laser flash. Experimental data are shown in black and the model spectra are shown in red (shifted down for clarity). The positive and negative signals represent microwave absorption (A) and emission (E), respectively. The details of the simulations are described in the text. The simulation parameters are summarized in Table 2.

Table 2. Simulation parameters of the TREPR spectra^a.

Compound	State	Energy structure (GHz)				DAF µs	Populations
		D	E	J	d		p Tz	p Txy	p dx	p dy	p dz
NI	LE	2.48	−0.13	—	—		1.00	−0.606	—	—	—
PTZ-NI	LE	2.30	−0.12	—	—	0.7	0.047	−0.548	—	—	—
						6.0	−0.259	0.317	—	—	—
	CS	−0.92	0.02	—	—	0.7	−0.047	−0.178	—	—	—
						6.0	0.070	−0.028	—	—	—
PXZ-NI	LE	2.02	−0.06	—	—	0.7	0.036	−0.163	—	—	—
	CS	−1.12	0.03	—	—	0.7	1.000	0.324	—	—	—
NI-TEMPO	D/J	2.11	−0.31	0.19	−0.46	0.7	−0.654	−0.118	−0.547	−0.934	0.288
						19.2	0.057	0.009	0.045	0.078	−0.022
PTZ-NI-TEMPO	D/J	2.35	−0.32	−0.10	−0.45	0.7	0.274	0.087	0.245	−1.000	−0.420
						4.8	0.012	0.012	−0.013	−0.144	0.130
PXZ-NI-TEMPO	D/J	2.27	−0.21	−0.12	−0.40	0.7	0.453	0.110	−0.485	−0.598	−1.000

^aNotations: LE – locally excited triplet state, CS – charge separated triplet state, D/J – triplet-doublet spin system in a D/J dominant regime, DAF – delay after laser flash.

The polarization parameters further support the separation of these two components. At early delay times, the ³LE component of PTZ-NI retains a triplet-like polarization pattern, although it differs from NI by enhanced polarization along the linker direction. This is consistent with population transfer through a spin–orbit charge-transfer intersystem crossing (SOCT-ISC) channel.^43–45 The ³CS component exhibits a distinct set of sublevel populations and is shorter-lived than the ³LE component, most clearly in PTZ-NI. With increasing delay time, the polarization parameters of both components change, showing that the spectral evolution is governed not only by the spin-Hamiltonian parameters but also by redistribution of spin populations among the triplet sublevels.

The TREPR spectra of PXZ-NI are broader and less structured than those of PTZ-NI. Still, they also cannot be reduced to a single localized triplet contribution. Weak additional outer-wing features around ca. 300 and 385 mT indicate the presence of a second component, although it is less clearly resolved because of line broadening and spectral overlap. The simulations therefore use the same minimal model as for PTZ-NI, with a localized ³LE contribution and a second component assigned to ³CS.

The polarization parameters collected in Table 2 show markedly different ³LE and ³CS population patterns for PXZ-NI. This indicates that the spin polarization in PXZ-NI originates from a mechanism different from that of a perturbed NI-like triplet. Thus, the radical-free series progresses from the single localized triplet state of NI to donor-containing dyads in which localized and charge-separated triplet channels coexist. This coexistence is directly resolved for PTZ-NI and remains detectable, though less cleanly, for PXZ-NI.

TEMPO-containing series

A qualitatively different TREPR response is observed after incorporation of the TEMPO radical (Fig. 7). In NI-TEMPO, intense features appear in the near-center region in addition to the broader triplet-like wings (Fig. 7a). Thus, the spectrum does not correspond to the characteristic pattern of a strongly coupled triplet–doublet system that can be described using an effective ZFS tensor of the quartet state. The triplet chromophore–radical interactions must be included explicitly to describe such a complex spectrum. These interactions are described by the exchange coupling J and the dipolar coupling d, with the latter approximated as being directed along the linker. For the TEMPO-containing compounds, additional triplet-doublet (TD) polarization terms p_d are required, because the conventional triplet mechanism alone is insufficient to reproduce the near-center part of the spectra. The NI-TEMPO spectrum is reproduced with D = 2.11 GHz, E = −0.31 GHz, J = 0.19 GHz, and d = −0.46 GHz, which places the system in the triplet-doublet regime where the exchange interaction is significant but still smaller than the chromophore ZFS. In this regime, the coupling is not strong enough to split the excited state into distinct doublet and quartet manifolds; instead, the experimental pattern is more appropriately discussed as that of a TD system in the D/J-dominant regime.³⁵

Fig. 7. TREPR spectra of (a) NI-TEMPO, (b) PTZ-NI-TEMPO, and (c) PXZ-NI-TEMPO recorded at several intervals after the laser flash. Experimental data are shown in black and the model spectra are shown in red (shifted down for clarity). The positive and negative signals represent microwave absorption (A) and emission (E), respectively. The details of the simulations are described in the text. The simulation parameters are summarized in Table 2.

The time evolution of the NI-TEMPO spectrum is especially informative. The intense lines near the center of the spectrum reverse sign on the microsecond timescale. Related time-dependent polarization inversions have also been reported for coupled chromophore–radical systems,^46–48 although the underlying mechanisms depend on the molecular structure and coupling strength.⁴⁹ In the simulations, this inversion is reproduced by changing the polarization terms associated with the coupled TD manifold. The side wings retain the contribution from the conventional triplet mechanism of the NI chromophore, whereas the near-center part requires additional polarization terms induced by coupling to the radical. At early delay times, these TD-related population terms are large and strongly non-equilibrium; at later times, they become much smaller and partly change sign. Thus, the observed inversion of the central spectral pattern can be attributed to the time-dependent redistribution of populations between the spin sublevels.

The triads PTZ-NI-TEMPO and PXZ-NI-TEMPO retain the overall characteristics of a coupled triplet–doublet system, but their spectra differ clearly from that of NI-TEMPO in the central region. The most obvious difference is the inverted near-center polarization pattern: compared with NI-TEMPO, the donor-containing triads show an A/E sequence instead of E/A and a smaller splitting between the central features. These changes indicate that donor incorporation modifies both the magnitude and the sign of the triplet–radical exchange interaction. The fitted parameters are consistent with this conclusion, giving negative J values for both donor-containing triads, in contrast to the positive exchange found for NI-TEMPO.

At the same time, the two donor-containing radical triads are not equivalent. PTZ-NI-TEMPO exhibits a more pronounced spectral evolution and a more structured spectrum, whereas PXZ-NI-TEMPO shows slower decay and a different balance between the central and wing intensities. This parallels the distinction already seen in the radical-free dyads, where PTZ gives the more clearly resolved two-state behavior and PXZ gives broader, more overlapped spectra. For the radical-free compounds, a single triplet component cannot reproduce all spectral features, and separate ³LE and ³CS contributions are therefore required. In contrast, the spectra of the radical-containing compounds can be reproduced using a minimal model containing a single effective coupled TD system.

The polarization parameters in Table 2 describe the main spectral changes of the TEMPO-containing triads. For PTZ-NI-TEMPO, the early-time spectrum has strongly nonequilibrium TD polarization terms p_d. At later times, the magnitudes of the polarization terms decrease, and the central part of the spectrum changes accordingly. For PXZ-NI-TEMPO, the representative early-time parameters already reproduce the main spectral features, consistent with the less pronounced inversion behavior within the analyzed time window. In both systems, the central region is the most sensitive part of the spectrum. This agrees with the theoretical description of the TREPR spectra in the D/J-dominant regime, where the additional near-center features and their polarization are mainly controlled by exchange-sensitive TD terms and radical-enhanced ISC.³⁵

Overall, the TREPR data reveal two distinct but connected levels of spin-state complexity across the NI-based series. In the radical-free compounds, donor substitution converts the localized NI triplet into a system where localized and charge-separated triplet states coexist and show different polarization patterns. Upon TEMPO incorporation, the excited-state manifold becomes a coupled triplet–doublet system. Its most diagnostic signatures appear in the near-center region, and its time evolution is governed by redistribution of the populations of the TD system. Comparison of the radical-free and radical-containing NI derivatives shows that changing the donor from PTZ to PXZ changes the balance between localized and CT triplet character, whereas TEMPO introduces an additional exchange-mediated contribution to the spin dynamics that affects both the spectral shape and its temporal evolution.

Structure-dependent properties of the NI series

Taken together, the structural, optical, and spin-resolved data provide a consistent excited-state picture across the NI-based series. Single-crystal X-ray analysis of PTZ-NI-TEMPO shows that the donor, acceptor, and radical fragments are held in close spatial proximity while remaining largely orthogonal, and DFT reproduces this compact geometry for the triads. This structural motif is important for the present series: it preserves the electronically distinct roles of the donor, NI acceptor, and TEMPO radical, while still enabling efficient interfragment interactions in the excited state. In the DFT calculations (Fig. S41 and S42), the ground-state unpaired spin remains localized predominantly on TEMPO, whereas the excited-state spin density is redistributed over the NI and donor fragments in a donor- and environment-dependent manner. Thus, DFT provides the structural and electronic framework for understanding why donor substitution and solvent polarity strongly affect the balance between localized and CT triplet channels.

Within this framework, NI serves as a reference case of a localized triplet populated through intrinsic ISC. Incorporation of the donor (PTZ or PXZ) introduces an additional triplet configuration with charge-transfer character, so that localized and CS triplet states coexist in the donor-containing dyads PTZ-NI and PXZ-NI. This two-state character is seen most clearly for PTZ-NI, whereas in PXZ-NI it remains detectable but is less spectrally resolved because of stronger line broadening and overlap. The optical data show the same trend from an independent perspective: changing the donor from PTZ to PXZ shifts the excited-state manifold toward stronger charge-transfer character, while increasing solvent polarity stabilizes the charge-separated state and converts mixed ³LE/³CS behavior in nonpolar media into a more distinct ³CS regime in polar solvents. Thus, TREPR, optical data, and DFT all indicate donor-dependent changes in the excited-state electronic structure.

Incorporation of TEMPO introduces an additional unpaired electron into the system. In NI-TEMPO, triplet-doublet coupling mixes the spin wavefunctions and redistributes the TD sublevel populations, initially favoring the sublevels associated with the m_S = ±1/2 manifold. Faster depopulation of these sublevels relative to the m_S = ±3/2 states leads to inversion of the central TREPR polarization pattern on the microsecond timescale. In the radical-containing triads, the spectra are described most consistently within a single coupled TD system in the D/J-dominant regime rather than as separate doublet and quartet multiplets. The PTZ- and PXZ-containing triads show similar donor-dependent initial polarization patterns. However, the PTZ-NI-TEMPO spectrum evolves more rapidly with delay time. This behavior is consistent with faster spin relaxation and depopulation of the metastable TD state.

The results from the different methods are mutually consistent and can be summarized in terms of how each molecular or environmental factor perturbs the excited-state manifold. The NI core provides the localized triplet reference, donor substitution introduces charge-transfer triplet character, solvent polarity shifts the LE/CS balance toward charge-separated states, and TEMPO introduces an additional unpaired electron that couples to the triplet excitation, forming a coupled triplet–doublet manifold. Together, these relationships provide the basis for assigning the spin states, dominant processes, and TREPR signatures across the series, as summarized in Table 3.

Table 3. Comparative summary of the metastable state across the NI-based compounds^a.

Compound	Spin-state	Main processes	Dominant population/TREPR signature	Comment
NI	Localized triplet (LE)	ISC	Preferential population of the in-plane LE components	NI-centered triplet; out-of-plane principal ZFS axis
PTZ-NI	Coexisting LE and CS triplet states	ISC + SOCT-ISC	Two triplet contributions with distinct ZFS and LE and CS polarizations	Well resolved TREPR patterns
PXZ-NI	Coexisting LE and CS triplet states	ISC + SOCT-ISC	LE/CS contributions remain detectable but are more overlapped	Stronger CT character than PTZ-NI and broader TREPR spectra
NI-TEMPO	TD	ISC + radical-enchanced ISC	Early overpopulation of mS = ±1/2 followed by faster depopulations of these components	TD population redistribution governs the time evolution of the near-center lines
PTZ-NI-TEMPO	Expanded TD	ISC/SOCT-ISC + exchange-mediated polarization redistribution	Donor-modified central polarization pattern with faster evolution	PTZ accelerates relaxation and depopulation of the metastable manifold
PXZ-NI-TEMPO	Expanded TD	ISC/SOCT-ISC + exchange-mediated polarization redistribution	Donor-modified central polarization pattern with more overlapped response	PXZ remains more CT-biased and less rapidly evolving within the analyzed window

^aNotations: LE, locally excited triplet; CS, charge separated triplet; TD, TD system in the D/J-dominant regime; expanded TD, donor-modified TD system; SOCT-ISC, spin–orbit charge-transfer intersystem crossing.

Conclusion

Compact NI-based donor–acceptor–radical architectures provide a platform for tuning long-lived excited-state spin dynamics in closely interacting organic multispin systems. Structural analysis and DFT calculations support a compact, largely orthogonal geometry that preserves the donor, acceptor, and radical functions of the PTZ/PXZ, NI, and TEMPO fragments while maintaining the proximity required for efficient excited-state interactions. Across the series, the combined optical and magnetic-resonance data define a consistent progression from the localized NI triplet to donor-containing dyads in which localized and charge-separated triplet channels coexist. Donor substitution modulates this balance, with PXZ promoting stronger charge-transfer character, whereas increasing solvent polarity stabilizes the charge-separated state and shifts the system from mixed ³LE/³CS behavior toward a more distinct ³CS regime. TEMPO incorporation introduces exchange-mediated triplet–radical coupling, modifies triplet-state formation, shortens the lifetime of the metastable triplet/charge-transfer manifold, and reshapes the spin polarization of the excited states. TREPR shows that the radical-containing compounds are described most consistently as TD systems in the D/J-dominant regime, with donor substitution modifying the sign and magnitude of the triplet–radical exchange interaction. Overall, donor identity determines the localized-triplet/charge-transfer balance, solvent polarity controls charge-separated-state stabilization, and the proximal radical governs population redistribution, relaxation, and spin polarization. This combination modulates the ISC pathways and long-lived excited-state dynamics in compact donor–acceptor–radical molecular systems.

Experimental section

General method

All the chemical reagents used in the synthesis are analytically pure and were used as received. The major starting materials, 4-bromo-1,8-naphthalic anhydride (CAS: 81-86-7, purity ≥96%) and 4-amino-TEMPO (CAS: 14691-88-4, purity ≥97%) were purchased from Aladdin Chemicals. Phenothiazine (CAS: 92-84-2) and phenoxazine (CAS: 135-67-1) were obtained with purity of ≥98%. The solvents used in the synthesis were dried and distilled before use; the solvents used in the spectral measurements were analytical grade.

The UV-vis absorption spectra were recorded with a UV2550 spectrophotometer (Shimadzu Ltd, Japan). The fluorescence emission spectra were recorded with an FS5 spectrofluorometer (Edinburgh Instruments Ltd, UK). The prompt and delayed fluorescence lifetimes were measured with an OB920 luminescence lifetime spectrometer (Edinburgh Instruments Ltd, UK).

PTZ-NI-TEMPO

Under N₂, compound 1 (134 mg, 0.32 mmol), phenothiazine (75 mg, 0.38 mmol), Pd(OAc)₂ (14.0 mg, 0.06 mmol), [(t-Bu)₃PtH]BF₄ (42 mg, 0.14 mmol) were mixed in dry toluene (10 mL). After purging for 15 minutes with N₂, sodium tert-butoxide (240.0 mg, 2.5 mmol) was added and the reaction mixture was stirred under a N₂ atmosphere for 8 h at 120 °C. Upon cooling to room temperature, deionized water (10 mL) was added, and the mixture was extracted with dichloromethane (DCM) (3 × 20 mL). The organic layer was separated and washed with brine solution (2 × 20 mL). The organic layer was dried over anhydrous Na₂SO₄ and the solvent was evaporated under reduced pressure. The crude product was purified by column chromatography (silica gel, dichloromethane/ethyl acetate = 100/1, v/v). Compound PTZ-NI-TEMPO was obtained as a red solid (40 mg, yield: 46%). IR (KBr): ṽ cm⁻¹ = 2972.30, 2927.94, 2855.14, 1702.38, 1664.77, 1620.42, 1585.22, 1462.76, 1444.92, 1354.77, 1336.93, 1309.93, 1259.79, 1238.09, 923.75, 787.30, 750.66. MALDI−TOF−HRMS [C₃₃H₃₀N₃O₃S]˙: calcd m/z = 548.2002; found m/z = 548.1998.

PXZ-NI-TEMPO

It was synthesized with a method similar to PTZ-NI-TEMPO; phenoxazine (37.3 mg, 0.19 mmol) was used as a reactant and the product was obtained as a dark purple solid (62 mg, yield: 52%). IR (KBr): ṽ cm⁻¹ = 2973.26, 2924.08, 2851.76, 1703.34, 1665.25, 1622.83, 1588.60, 1488.31, 1463.73, 1407.32, 1356.21, 1338.37, 1291.61, 1274.25, 1239.06, 784.89, 756.45, 743.91. MALDI−TOF−HRMS [C₃₃H₃₀N₃O₄]˙: calcd m/z = 532.2231; found m/z = 532.2216.

Femtosecond transient absorption spectroscopy

A femtosecond Ti: Sapphire regenerative amplified Ti: Sapphire laser system (Coherent, Astrella-Tunable-F-1k) and femtosecond transient absorption spectrometer system (Ultrafast Systems, Helios Fire) were used for fs-TA measurements. The amplifier pulse width at 800 nm is 84 fs and the repetition frequency is 1 kHz; the output power of 800 nm from the amplifier is 7.4 W. The probe pulse was obtained by using approximately 4% of the amplified 800 nm output from the Astrella to generate a white-light continuum (320–700 nm) using a CaF₂ plate. Before passing through the sample, the probe beam is divided into two beams. One beam of detecting light passes through the sample, and the other is sent directly to a reference spectrometer, which monitors fluctuations in detecting light intensity. The fiber optics was coupled to a multichannel spectrometer with a 1.5 nm intrinsic resolution CMOS sensor. The maximum time delay was 8 ns and the instrument response function (IRF) was about 120 fs. At each time delay, data were averaged over 2 seconds and were collected by the acquisition system. In this study, the sample solution was excited by a 400 nm pump beam. The power irradiated to the sample during the test was 0.25 mW. The sample solutions were excited in a 2 mm path-length cuvette. The data are stored in a three-dimensional wave-time-absorbance matrix and exported for use. Chirp correction was performed for all data. The data were analyzed by means of singular value decomposition and global analysis, employing the software Glotaran.⁵⁰

Nanosecond transient absorption spectroscopy

Nanosecond transient absorption spectra were measured on an LP980 laser flash-photolysis spectrometer (Edinburgh Instruments, UK) and the transient signal was digitized with a Tektronix TDS 3012B oscilloscope. Samples were purged with N₂ for 20 minutes before measurement and excited with a nanosecond pulsed laser (Surelite I-10, USA; the wavelength is tunable in the range of 400–2400 nm). The decay traces were analyzed using the L900 software, in which the kinetic profiles were fitted by nonlinear least-squares analysis using appropriate exponential decay functions to extract the triplet-state lifetimes. The fitting quality was evaluated based on the goodness of fit and randomly distributed residuals.

Time-resolved electron paramagnetic resonance spectroscopy

TREPR magnetic measurements were performed on a TREPR spectrometer setup based on a continuous wave X-band Bruker ELEXSYS E680 spectrometer. The concentration of the samples was 1 × 10⁻⁴ M, and the solvent system used was a mixture of toluene and 2-methyltetrahydrofuran (2-MeTHF) in a 3 : 1 (v/v) ratio. The solutions in quartz tubes were deoxygenated by repeated freeze–pump–thaw procedures and sealed. Before the measurement, each sample tube was shock-frozen in liquid nitrogen. Next, sample tubes were quickly transferred into the spectrometer probe precooled chamber at 80 K after reaching equilibrium. The samples were photoexcited using a SOLAR LQ629 Nd:YAG laser at 355 nm (with 100 Hz repetition rate and 1 mJ energy).

Numerical simulation details

TREPR spectra of the radical-free compounds were analyzed within effective triplet spin Hamiltonian formalism. For NI, the excited state was described by the conventional triplet Hamiltonian containing the Zeeman and ZFS terms. For PTZ-NI and PXZ-NI, the localized NI-centered triplet was treated in the same way, whereas the CS contribution was represented by an effective triplet Hamiltonian with its principal axis taken along the linker.

For the TEMPO-containing compounds, the excited states were described within a coupled triplet–doublet spin system including the chromophore ZFS together with isotropic exchange J and effective dipolar coupling d between the chromophore triplet and the radical spin. The main axis of the chromophore ZFS tensor is the out-of-plane z-axis. The dipolar axis was taken along the linker direction (molecular frame x-axis). Since the chromophore ZFS remains larger than the exchange interaction, the spectra were analyzed within the D/J-dominant coupling regime.³⁵ The spin–spin interaction parameters are determined by fitting the experimental TREPR spectra, with the signs of D fixed according to the assignments discussed above. The exchange interaction is described by the scalar product of the vector operators of the triplet (T) and the radical (F) spins as follows: H_ex = −2J S_T·S_F. The signs of J reported in Table 2 provide better overall agreement with the experimental TREPR spectra, particularly in the exchange-sensitive near-center region. Although the effective model used to describe the radical-containing series does not reproduce all spectral details perfectly, simulations with the opposite signs of J yield a poorer overall description of the experimental spectra.

The spin polarization of the radical-free systems was described using the conventional triplet population terms. The corresponding traceless density matrix equals Δρ_T = p_Tz(S_Tz² − S_T²/3) + p_Txy(S_Tx² − S_Ty²). For the TEMPO-containing compounds, these terms were supplemented by additional triplet–doublet polarization terms that account for redistribution of sublevel populations induced by chromophore–radical coupling. To minimize the influence of time-dependent spectral reshaping, the spin-Hamiltonian parameters of the radical-containing systems were extracted from the earliest spectra after the laser flash, whereas the later spectral evolution was interpreted through changes in the polarization terms. The narrow central line near g ≈ 2 was treated separately as an effective spin−1/2 contribution.

TREPR spectra recorded at several delay times were fitted simultaneously by minimizing the total sum of squared residuals between the experimental and simulated spectra. The fitting variables included the spin-Hamiltonian parameters and polarization weights.

Author contributions

Y. E. K. and M. I. contributed equally to this work. J. Z. conceived the concept, acquired the funding, directed the data analysis and manuscript writing; Y. E. K. performed the TREPR spectral analysis and contributed to the writing of the manuscript; M. I. synthesized the compounds, measured and analyzed steady-state optical spectra, electrochemistry, nanosecond transient absorption spectra, analyzed the femtosecond transient absorption spectra, and modified the manuscript; A. A. S. and V. K. V. measured the TREPR spectra and analyzed part of the data; J. L. and M.-D. L. performed the femtosecond transient absorption studies; C. W. and Y. H. X. performed the single-crystal analysis; all authors contributed to the writing of the manuscript.

Conflicts of interest

There are no conflicts to declare.

Data availability

All relevant data are presented in the main text and/or supplementary information (SI). Supplementary information: the molecular structural characterization data, UV-vis absorption and fluorescence spectra, nanosecond and femtosecond transient absorption spectra, TREPR spectra and theoretical computation information. See DOI: https://doi.org/10.1039/d6sc04928a.