Distinct properties of the triplet pair state from singlet fission

The triplet pair from singlet fission is characterized by distinct spectroscopic signature and can be difficult to break apart.

Abstract

Singlet fission, the conversion of a singlet exciton (S₁) to two triplets (2 × T₁), may increase the solar energy conversion efficiency beyond the Shockley-Queisser limit. This process is believed to involve the correlated triplet pair state ¹(TT). Despite extensive research, the nature of the ¹(TT) state and its spectroscopic signature remain actively debated. We use an end-connected pentacene dimer (BP0) as a model system and show evidence for a tightly bound ¹(TT) state. It is characterized in the near-infrared (IR) region (~1.0 eV) by a distinct excited-state absorption (ESA) spectral feature, which closely resembles that of the S₁ state; both show vibronic progressions of the aromatic ring breathing mode. We assign these near-IR spectra to ¹(TT)→S_n and S₁→S_n′ transitions; S_n and S_n′ likely come from the antisymmetric and symmetric linear combinations, respectively, of the S₂ state localized on each pentacene unit in the dimer molecule. The ¹(TT)→S_n transition is an indicator of the intertriplet electronic coupling strength, because inserting a phenylene spacer or twisting the dihedral angle between the two pentacene chromophores decreases the intertriplet electronic coupling and diminishes this ESA peak. In addition to spectroscopic signature, the tightly bound ¹(TT) state also shows chemical reactivity that is distinctively different from that of an individual T₁ state. Using an electron-accepting iron oxide molecular cluster [Fe₈O₄] linked to the pentacene or pentacene dimer (BP0), we show that electron transfer to the cluster occurs efficiently from an individual T₁ in pentacene but not from the tightly bound ¹(TT) state. Thus, reducing intertriplet electronic coupling in ¹(TT) via molecular design might be necessary for the efficient harvesting of triplets from intramolecular singlet fission.

INTRODUCTION

Singlet fission is a many-body photophysical process in molecules where the photoexcited singlet (S₁) splits into two triplets (2 × T₁) with spin conservation (1, 2). Since its discovery, efficient singlet fission has been reported mostly for solids and aggregates of conjugated molecules (1, 2), and a dominant mechanistic picture is the molecular dimer model (3, 4)

$${\mathrm{S}}_0+{\mathrm{S}}_1\leftrightarrows {}^1(\text{TT})\leftrightarrows 2{\mathrm{T}}_1$$ (1)

where S₀ is the ground state and the intermediate ¹(TT) is the correlated triplet pair with both singlet spin and double-excitation characters. Despite its prevalent use, Monahan et al. (5, 6) pointed out the inadequacy of the dimer model in describing inherently delocalized excitons in the solid state. Exciton delocalization has been cited as an important driving force for singlet fission (7–10). This problem is circumvented in recent demonstrations of efficient singlet fission in single molecules [particularly in dimers of acenes (11–16)] that allow for accurate application of the dimer model and for closely connecting experiment with theory (17). The isolation of the transient ¹(TT) state in a single molecule leads to a much longer lifetime than that in the condensed phase, thus allowing spectroscopic characterization of this ambiguous and poorly understood state. This is exemplified in the detection by electron spin resonance spectroscopy in pentacene dimers of the quintet state, ⁵(TT), which is mixed with the ¹(TT) state as predicted by the spin Hamiltonian (18).

The ¹(TT) state is a singlet excited state with double-excitation characters and differs from 2 × T₁ not only by the electronic and spin entangled nature of the former but also by the presence of orbital overlap, which changes its excitation energy from the sum of two triplet energies. Scholes (19) recently clarified some persistent confusion on the ¹(TT) state in theoretical descriptions. The energetic difference between the correlated triplet pair state and two individual triplets, that is, the triplet pair binding energy, can be as large as 1 eV, as is known for the excited states of oligoenes (20–22), including carotenoids (23), where the tightly bound triplet pair has been called the “dark” S₁ state serving as a sink for nonradiative recombination and a less tightly bound triplet pair (S^*) has been associated with singlet fission (15, 24–26). In contrast, in prototypical model systems of pentacene or tetracene dimers (both covalent and van der Waals), computational analysis predicted little, if any, triplet pair binding energy (17, 27–32). However, a recent finding of similar ¹(TT) lifetimes in polypentacene and bipentacene indicates that the triplet pair does not dissociate even in a long conjugated chain (33), suggesting that the correlated triplet pair state is more strongly bound than previously thought.

A major obstacle to a clear understanding of the ¹(TT) state is the lack of spectroscopic signatures from experiments. Zhu and co-workers applied time-resolved two-photon photoemission spectroscopy to quantitatively determine the energetic position of the ¹(TT) state from its ionization potential (IP) in crystalline pentacene (34), tetracene (10), and hexacene (6). This approach is unambiguous only for hexacene (6) where the ¹(TT) state is energetically well separated from S₁ but is difficult for other singlet fission systems where the ¹(TT) states are in close energetic resonance with S₁. The most widely used technique to probe singlet fission has been transient absorption (TA) spectroscopy, but most studies to date have only identified spectral features assigned to S₁ and T₁ states, and there has been little explanation as to why these TA peaks nearly always overlap (12–16, 35, 36). Exceptions to this prevalent practice can be found in the recent work of Sanders et al. (11) who found, in pentacene dimers, an excited-state absorption (ESA) peak at ~690 nm whose magnitude is strongly correlated with the strength of intertriplet electronic coupling and in the work of Pensack et al. (37) who observed near-infrared (IR) (1200 to 1400 nm) ESA in pentacene aggregates assigned to ¹(TT) but not to the triplet pair labeled ¹(T…T), which has lost electronic coherence but retained spin coherence. These two examples reveal the presence of spectroscopic signatures for the ¹(TT) state in TA, but the origins of these transitions and their relationships to the energetics of the ¹(TT) state remain unknown.

The distinct electronic structure of the ¹(TT) state should be reflected not only in its spectroscopic signature but also in its chemical and physical properties. The oft-cited motivation for nearly every recent paper on this subject has been the potential “usefulness” of singlet fission to solar energy conversion. The basic argument was put forward initially by Dexter (38) for the sensitization of conventional solar cells by singlet fission chromophores in 1979, but a more recent paper by Hanna and Nozik (39) on using singlet fission to increase the solar cell efficiency above the Shockley-Queisser limit really energized the field. A number of research groups have explored the harvesting of triplet pairs from intermolecular singlet fission using solid interfaces between a singlet fission material and an electron or triplet acceptor material (34, 40–43). These efforts have also led to the successful demonstration of singlet fission–based solar cells with quantum efficiencies exceeding 100% (44). The recent demonstration of efficient intramolecular singlet fission in single molecules (11–16, 36) opens the door to new opportunities for the realization of singlet fission–sensitized solar cells (45). A more exciting opportunity is the potential for the harvesting of two electron-hole pairs for photocatalysis, for example, by coupling a singlet fission molecule to a molecular or cluster-based catalytic center (46) to enable two-electron redox reactions. Unlike intermolecular singlet fission in the solid state in which electronic delocalization (5, 47) and entropy (10) are driving forces to split the ¹(TT) state to two electronically decoupled triplets (which can nonetheless retain spin coherence) on ultrafast time scales (19), the confinement in a molecular dimer or oligomer traps the two triplets in the ¹(TT) state in a single molecule (18, 33). Thus, instead of individual triplets at solid-state interfaces, the harvesting of triplets in intramolecular single fission would likely come from the ¹(TT) state. However, the two triplets in the ¹(TT) state from intramolecular singlet fission can be tightly bound, and charge or energy transfer from each triplet may be inhibited.

Here, we use triisopropylsilylethynyl-functionalized pentacene (TIPS-Pc) dimers, each coupled at the 2-position without or with a phenylene spacer, BP0 or BP1 (Fig. 1) (11), as well as pentacene dimers with different dihedral angles (17), as model systems to quantitatively probe the nature of the tightly bound ¹(TT) state from the ESA spectra. Molecules of this type allow for the systematic tuning of electronic coupling between the two pentacene units and between the S₁ and the ¹(TT) states, as reflected in the singlet fission time constants of τ_SF = 0.76, 20, and 220 ps and triplet recombination time constants of τ_AN = 0.45, 16.5, and 270 ns for dimers with zero, one, and two phenylene spacers (BP0, BP1, and BP2), respectively, obtained from an analysis of TA spectra in the visible region (11). Here, we focus on the distinct ESA peak in the near-IR region (hν ~ 1 eV), which is a signature of the ¹(TT) state from singlet fission in BP0. Its intensity diminishes as the intertriplet electronic coupling is lowered in BP1 or significantly decreases in bipentacene with different dihedral angles (17). The ESA peak of ¹(TT) in BP0 closely resembles that of the S₁ state in the near-IR region; both show vibronic progressions of the aromatic ring breathing mode and can be assigned to the ¹(TT)→S_n and S₁→S_n′ transitions, respectively. This finding unambiguously establishes that ¹(TT) is spectroscopically distinct from 2 × T₁, and such a spectroscopic signature enables one to quantitatively follow the dynamics of this critical intermediate in singlet fission.

Fig. 1. The model systems for intramolecular singlet fission and triplet harvesting.

(A) Schematics of BP0, BP1, [Fe₈O₄]-Pc, and [Fe₈O₄]-BP0. R = (triisopropylsilyl)ethynyl (TIPS) for [Fe₈O₄]-Pc and (n-octyldiisopropyl)silylethynyl (NODIPS) for [Fe₈O₄]-BP0 and [Fe₈O₄]-BP1. The inset shows estimated IP and EA (electron affinity) from electrochemical oxidation/reduction potentials of [Fe₈O₄] and TIPS-pentacene. (B and C) Optical absorption spectra of (B) TIPS-Pc, BP0, and BP1 in toluene and (C) [Fe₈O₄], [Fe₈O₄]-Pc, and [Fe₈O₄]-BP0 in dichloromethane solutions.

To establish the distinct chemical properties of the ¹(TT) state, we use the redox-active molecular cluster Fe₈O₄pz₁₂Cl₄ (pz, pyrazolate), which we label [Fe₈O₄], as an electron acceptor (48, 49) and tether BP0 to [Fe₈O₄] through a Fe-phenoxide bond (schematically illustrated in Fig. 1A). As a control, we replace the pentacene dimer BP0 with a pentacene monomer (Pc). Note that the formation of the Fe-phenoxide bond in both [Fe₈O₄]-Pc and [Fe₈O₄]-BP0 introduces a low-energy absorption tail (~1.3 to 1.8 eV; Fig. 1C). This spectral feature has been assigned to the phenolate-to-Fe(III) ligand-to-metal charge transfer (LMCT) transition (50) but may also have contributions from pentacene-to-[Fe₈O₄] charge transfer (CT) transitions. We show that electron transfer from pentacene to [Fe₈O₄] occurs efficiently from an individual T₁ state in pentacene ([Fe₈O₄]-Pc), but not from the tightly bound triplet pair state in [Fe₈O₄]-BP0. This finding establishes that the chemical property of the ¹(TT) state is distinctively different from that of an individual triplet and suggests that reducing intertriplet electronic coupling in ¹(TT) might be needed for the harvesting of triplets from intramolecular singlet fission.

RESULTS AND DISCUSSION

Spectroscopic signature of the ¹(TT) state

We use TA spectroscopy to probe singlet fission in BP0 and BP1 (11). We excite the S₁ state of each pentacene dimer at hν₁ = 2.1 eV and probe the subsequent dynamics from the TA of a white-light continuum (Fig. 2). Figure 2 (A and B) shows TA spectra at selected pump-probe delays, Δt = 0.1 ps (red), 10 ps (purple), and 100 ps (blue) for BP0. The visible parts of the TA spectra have been discussed extensively before, and the broad positive TA features at Δt < 1 ps and Δt > 2 ps are assigned to the ESA of S₁ and T₁, respectively (11). The latter is confirmed by the ESA spectrum of T₁ obtained from sensitization (black). On the basis of the calculated triplet energies in pentacene (T₃ and T₄ are close in energy and are not distinguished here) (51), we assign the ESA peak at 2.42 eV to the T₁→T₃ transition. For the triplet pair from singlet fission, this transition corresponds to ¹(T₁T₁)→¹(T₁T₃). In each case, the ESA transition also shows vibronic progression (hν_vib ~ 0.17 eV) similar to those in the ground-state absorption spectrum (11). The singlet decay and triplet formation can be clearly seen from kinetic profiles at probe photon energies of hν₂ = 2.13 eV (gray) and 2.42 eV (green), respectively (Fig. 2C), with τ_SF = 0.7 ps (11); note that there is an overlapping contribution to ESA signal at hν₂ = 2.42 eV from the singlet at short time scales. The two triplets confined to the pentacene dimer can be assigned to ¹(TT), which decays on the time scale of τ_TT1 = 450 ps (see the green curve at long pump-probe delays in Fig. 2C), much shorter than the 30-μs lifetime of an individual triplet (11). Here, we focus on the near-IR region, which provides key spectroscopic insight into the triplet pair state.

Fig. 2. TA in the near-IR and visible regions reveal singlet and triplet characters of ¹(TT).

TA spectra in (A) the near-IR and (B) the visible regions for BP0 at different pump-probe delays, Δt = 0.1 ps (red), 10 ps (purple), and 100 ps (blue), following excitation at time zero by hν₁ = 2.1 eV. The triplet TA spectrum from sensitization (black) is also shown in (A) and (B). (C) Kinetic profiles from TA spectra for BP0 at the indicated probe photon energies. (D) TA spectra at Δt = 1 ps (red) and 100 ps (blue) for BP1 following excitation at time zero by hν₁ = 2.1 eV. The corresponding triplet spectrum (black) from sensitization is also shown.

There is a distinct ESA peak at 0.922 ± 0.005 eV when the singlet dominates at Δt = 0.1 ps (red) (Fig. 2A); this peak is also accompanied by a vibronic feature on the higher energy side, with hν_vib ~ 0.17 eV, similar to the vibronic progressions of S₀→S₁, T₁→T_n, and ¹(T₁T₁)→¹(T₁T_n) discussed above. This ESA is assigned to an S₁→S_n′ transition, with transition energy close to the S₁→S₂ transition for a single pentacene molecule. In the absorption spectrum of TIPS-Pc in Fig. 1B (blue), there is a weakly allowed S₀→S₂ peak at 2.82 eV, in agreement with the two-photon absorption spectrum of the same molecule (52). Given the S₀→S₁ peak at 1.93 eV (blue spectrum in Fig. 1B), we obtain the S₁→S₂ transition energy at 0.89 eV. In conjugated bipentacene dimers, the singlet states are described by linear combinations of two localized states on each pentacene chromophore (17). Although both symmetric and antisymmetric linear combinations are possible, the optically bright S₁ state in BP0 is of odd parity (u). Therefore, excited state transitions must occur to S_n′ states of even parity (g). We assign the 0.92-eV peak to a transition from S₁ to the symmetric linear combination of the monomer S₂ states.

At longer pump-probe delays, for example, Δt = 10 ps (purple) or 100 ps (blue), when there is only the triplet pair state, the ESA spectrum blue-shifts to 1.012 ± 0.005 eV and the vibronic signature becomes better resolved. This ESA peak does not originate from a T₁→T_n transition as it is completely absent in the triplet absorption spectrum (black) from sensitization. On the basis of the similarity of this ESA peak to that of the S₁→S_n′ transition at early times, we assign the former to a ¹(TT)→S_n transition. The ¹(TT) state in BP0 is expected to correspond to the totally symmetric representation, as shown theoretically by Fuemmeler et al. (17); it will be of opposite parity to the S₁ state and will exhibit a distinct set of excited state transitions to states of odd parity (17). S_n is expected to be close in energy to S_n′, because the difference in the S₁→S_n′ and ¹(TT)→S_n transition energies, Δ_E = 90 meV, is close to the predicted exoergicity of ~100 to 150 meV for singlet fission in bipentacene (17, 27, 29–32). The small energy spacing implies that both S_n and S_n′ likely originate from different linear combinations of the S₂ monomer state of different parity. Note that, unlike the results shown here for the pentacene dimer, the near-IR ESA assigned to ¹(TT) in pentacene aggregates does not show vibronic features (37).

The ESA spectrum of the ¹(TT) state reveals its delocalized singlet and localized triplet characters in the near-IR and the visible regions, respectively. We use “delocalized singlet” or “delocalization” to emphasize ¹(TT) in a single electronic state, which can be approximately viewed as two T₁ states (on two pentacene units) that are electronically coupled and coherent. Likewise, the term “localized triplet” or “localization” refers to a T₁ state on an individual pentacene unit with physical properties that are insensitive to the presence or absence of electronic coupling and coherence with a neighboring T₁ state. Spectroscopically, delocalization and localization are reflected in the transitions ¹(TT)→S_n and ¹(T₁T₁)→¹(T₁T_n), respectively. Note that the two notations, ¹(TT) and ¹(T₁T₁), describe the same triplet pair state. The kinetic profiles (Fig. 2C) for the ¹(TT)→S_n (orange) and ¹(T₁T₁)→¹(T₁T_n) (green) transitions are similar; the difference at short time scales (<1 ps) can be attributed to the different overlapping contributions from ESA of S₁. Note that transitions to S_n are allowed from ¹(TT) but spin-forbidden from ³(TT) or ⁵(TT). The perfect agreement between the decays of ¹(TT)→S_n and the ¹(T₁T₁)→¹(T₁T_n) signals suggests that there are negligible transitions within the triplet pair manifold, for example, ¹(TT)→⁵(TT) or ¹(TT)→³(TT), during the lifetime (450 ps) of the triplet pair state. Transitions within the triplet pair manifold are expected to occur on much longer time scales (18, 53).

Supporting the conclusion that delocalization or intertriplet electronic coupling in the ¹(TT) state is reflected in the ¹(TT)→S_n transition strength, we find that, in BP1, the weakening of the inter-T₁ electronic coupling diminishes its delocalized character as reflected in the ¹(TT)→S_N transition strength (Fig. 2D) where the near-IR peak for ¹(TT) at long times, for example, Δt = 100 ps (blue), becomes nonresolvable from the broad background, in distinct contrast to the S₁→S_n peak at Δt = 1 ps. In contrast, the localized character represented by the ¹(T₁T₁)→¹(T₁T_n) transition in the visible region remains (11).

To more quantitatively isolate the S₁ spectrum from that of the ¹(TT), we carry out global analysis based on a sequential kinetic model, S₁→¹(TT)→S₀ (11). The resulting S₁ (red) and ¹(TT) (blue) spectra are shown in Fig. 3. The global analysis yields time constants of 0.75 ± 0.05 ps and 460 ± 10 ps for singlet fission and triplet pair annihilation, respectively, in agreement with the previous report (11). Similar to the S₁-S_n′ transition, the ¹(TT)-S_n transition is also characterized by vibronic peaks assigned to 0-0 and 0-1 transitions, with a vibrational energy spacing of 0.16 to 0.17 eV, which corresponds to the ring breathing mode of pentacene along the short molecular axis (54). In addition to the near-IR peak, the ¹(TT) state in BP0 also features a distinct peak at 1.810 ± 0.005 eV. Similar to the transition at 1.012 ± 0.003 eV, the peak at 1.810 ± 0.005 eV diminishes as the intertriplet coupling weakens in BP1 and BP2 (11). Thus, the peak at 1.810 ± 0.005 also reflects the singlet character of the ¹(TT) state and can be assigned to a ¹(TT)→S_n″ transition. Because of the overlapping bleaching feature (S₀→S₁), we are not able to resolve vibronic progression for this transition.

Fig. 3. TA spectra of BP0 for the S₁ and ¹(TT) states from global analysis.

Red: Singlet state. Blue: Triplet pair state. Inset: 2D pseudocolor (intensity) plot of TA spectra following excitation at time zero by hν₁ = 2.1 eV. The transitions, along with vibronic progressions, are shown on each spectrum.

Further supporting the conclusion that the near-IR ¹(TT)→S_n″ transition is a spectroscopic signature of the tightly bound triplet pair state, we turn to modified BP0 molecules with different dihedral angles. In this approach, we control the dihedral angle twist by steric hindrance from the phenyl group attached to the 1-position of one or both pentacene units in the bipentacene molecule, as shown schematically in the insets in Fig. 4 (17). Computational analysis gives dihedral angles between the two pentacene molecules of 42° and 57°, and these two molecules are therefore labeled as BP-42 and BP-57, respectively (17). For comparison, the dihedral angle in BP0 is 37°; thus, BP0 ≡ BP-37. Theoretical analysis showed that the intertriplet electronic coupling decreases with increasing dihedral angle (17). The singlet fission time constants are τ_SF = 0.76, 1.69, and 3.38 ps, and the corresponding triplet-triplet annihilation time constants are τ_SF = 0.45, 1.6, and 5.2 ns for BP-37, BP-42, and BP-57, respectively (17). Figure 4 shows the near-IR region of the S₁ (red) and ¹(TT) (blue) ESA spectra for BP-42 (top) and BP-57 (bottom). We multiply the ¹(TT) spectra by factors of 2.5 and 4.6 for BP-42 and BP-57, respectively, to normalize the ¹(TT)-S_n peak intensity to the S₁-S_n′ intensity in each case. For comparison, the normalization factor would be 1.25 for BP-37 in Fig. 3. Thus, relative to the S₁-S_n′ transition, the ¹(TT)-S_n transition strength is 80, 40, and 22% for BP-37, BP-42, and BP-57, respectively. This confirms the correlation between the ¹(TT)-S_n ESA transition strength and the intertriplet electronic coupling in the ¹(TT) state.

Fig. 4. TA of the ¹(TT) state in the near-IR region depends on electronic coupling.

Near-IR TA spectra of BP-42 (top) and BP-57 (bottom). The ¹(TT) spectra (blue) have been multiplied by factors of 2.5 and 4.6 for BP-42 and BP-57, respectively, to normalize the peak intensities of ¹(TT) to those of S₁ (red).

In all the pentacene dimers investigated here, the decay rate of the triplet pair state is also found to be strongly correlated with the extent of delocalization in the ¹(TT) state, which is reflected in the ¹(TT)-S_n transition strength. This is understood as the rate of T₁-T₁ annihilation is determined by the inter-pentacene electronic coupling strength, as addressed in detail elsewhere (11, 17).

The relative amplitudes of the 0-0 and 0-1 transitions allow us to estimate the Huang-Rhys factor (S) in each case and, thus, the relative shifts in the potential energy surfaces (PESs) involved. The Huang-Rhys factor is related to the offset (ΔQ_e) in the equilibrium positions of the two PESs in an optical transition: S = 0.5α (ΔQ_e)², where α = μω/ℏ; μ is the reduced mass, and ω is the angular frequency of the vibration (55). In the harmonic oscillator and low-temperature approximation appropriate for the pentacene ring breathing mode at room temperature, the ratio in the Franck-Condon factors (and the ratio in peak intensities) between the 0-1 and 0-0 transition is equal to the Huang-Rhys factor (55). Thus, we obtain S = 0.36 ± 0.05 and 0.45 ± 0.05 for the S₁→S₃ and ¹(TT)→S₃ transitions, respectively, from the near-IR ESA spectra for BP0 in Fig. 3. For comparison, we obtain from the optical absorption spectrum a value of S = 0.55 ± 0.05 for the S₀-S₁ transition (11). For the pentacene ring breathing mode, we neglect the difference in equilibrium geometries between S_n′ and S_n, because they both likely come from the linear combination of S₂ in each pentacene chromophore. The spectroscopic results obtained above allow us to construct PESs for singlet fission in BP0. Although there are four possible arrangements of the PES from experimental ΔQ values, Fig. 5 shows the scenario that is more consistent with the expectation of increasing nuclear displacement with excitation energy. The offset in equilibrium positions of the S₁ PES and the ¹(TT) PES (ΔQ_e ~ 0.081 Å) is also consistent with theoretical results on the covalent dimer (17). The barrierless nature of the crossing point between S₁ and ¹(TT) explains the fast singlet fission rate for BP0. Furthermore, the PES of ¹(TT) crosses that of S₀ with only two vibrational quanta on the former; this opens up an efficient nonradiative decay pathway. The nonradiative lifetime of ¹(TT) (450 ps in BP0) is shorter than that of the radiative lifetime (~13 ns) of S₁ in TIPS-pentacene (11, 56). Although Fig. 5 is only an approximation given the uncertainties in the spectroscopic determination of Huang-Rhys factors, it represents the first estimation of PES for singlet fission from experimental data.

Fig. 5. Estimated PES for BP0 molecule.

The barrier-less nature for the crossing from S₁ (red) to ¹(TT) (blue) facilitates the fast singlet fission for BP0. The near-IR transition for BP0 can be explained by the transition from ¹(TT) to S_n, which is similar to the transition from S₁ to S_n′.

Distinct chemical property of the ¹(TT) state

The above spectroscopic analysis of singlet fission in BP0 provides evidence for a strong coupled triplet pair state, ¹(TT), whose delocalized and localized characters are revealed in ESA in the near-IR and visible regions, respectively. Here, we show that the tightly bound triplet pair state exhibits chemical properties that are different from those of an individual triplet.

The inset in Fig. 1A shows the estimated values for the IP and EA for TIPS-Pc and [Fe₈O₄], respectively. These values are obtained from the electrochemical oxidation potentials for TIPS-Pc (57) and [Fe₈O₄] (58), respectively, based on the reference value of the Ag/AgCl electrochemical potential at 4.4 eV below vacuum energy (E_V) (59). Also shown are the estimated IPs of S₁ and T₁ states from the excitation energies of TIPS-Pc. The use of IPs and EAs of both ground and excited states allows us to accurately put all relevant energy levels on the same single-particle diagram, as discussed in detail by Zhu (60). Note that the energy levels obtained from electrochemistry are adiabatic single-particle energies and can be used to approximate the vertical single-particle energies, that is, highest occupied molecular orbital and lowest unoccupied molecular orbital, when the reorganization energies are negligible (60). Given this approximate energy level diagram, we expect efficient electron transfer from either the T₁ or the S₁ in pentacene to the [Fe₈O₄] cluster. Figure 1C compares the optical absorption spectra of the [Fe₈O₄] cluster (black) and those of compounds [Fe₈O₄]-Pc (red) and [Fe₈O₄]-BP0 (blue). The absorption spectra of both [Fe₈O₄]-Pc and [Fe₈O₄]-BP0 primarily arise from the sum of the absorption spectra of [Fe₈O₄] and that of pentacene or bipentacene. An additional broad feature below 1.75 could contain a CT state from the Pc-PhO-ligand or the BP0-PhO-ligand to the [Fe₈O₄] cluster, in addition to the more local LMCT transition (50).

The energy level alignment in Fig. 1A suggests that, in addition to direct photoexcitation of the CT state, electron transfer can occur from T₁ in pentacene to [Fe₈O₄] to indirectly form the CT state. We find that CT and T₁ are strongly coupled resonantly. When we directly excite the CT state in [Fe₈O₄]-Pc or [Fe₈O₄]-BP0 at hν = 1.65 eV (Fig. 6A), we observe in each case a TA spectrum characteristic of the T₁ state in pentacene, including an ESA peak at ~2.4 eV and a ground-state bleaching at 1.88 and 2.05 eV (see fig. S1 for complete TA data for [Fe₈O₄]-Pc). Although we observe small differences in the TA spectra for [Fe₈O₄]-Pc (green) and [Fe₈O₄]-BP0 (red), they all match their T₁ spectra obtained by sensitization very well. For clarity, here, we only present the T₁ spectrum of [Fe₈O₄]-BP0 (gray); see the Supplementary Materials for T₁ spectra of the other molecules. Note that neither the isolated [Fe₈O₄] nor the uncoupled pentacene molecules absorb light below ~1.75 eV. Excitation of isolated [Fe₈O₄] at higher photon energies results in completely different TA spectra (fig. S2). The ultrafast formation of T₁ within experimental time resolution (~100 fs) from the selective excitation of CT indicates that the cluster and pentacene ligands are strongly electronically coupled. Supporting this conclusion, we found in a triplet sensitization experiment that the observable T₁ signal from [Fe₈O₄]-Pc is an order of magnitude lower than that from TIPS-Pc (fig. S3).

Fig. 6. TA reveals the strong coupling of CT state to T1.

(A) TA spectra at 1 ps for [Fe₈O₄]-Pc (green) and [Fe₈O₄]-BP0 (red) upon CT excitation of 1.65 eV. The gray curve is the triplet spectrum of [Fe₈O₄]-BP0 from triplet sensitization. (B) Triplet decay dynamics for [Fe₈O₄]-Pc (green) and [Fe₈O₄]-BP0 (red and blue for ESA and ground-state bleaching, respectively). The solid curves are single-exponential fits with the indicated lifetimes (τ = 16 ± 2 ps for [Fe₈O₄]-Pc and 28 ± 3 ps for [Fe₈O₄]-BP0).

The coupled T₁-CT state features first-order decay kinetics well described by single-exponential decays (solid curves in Fig. 6B), with time constants of τ_CT-T1 = 28 ± 3 ps and 16 ± 2 ps for [Fe₈O₄]-BP0 and [Fe₈O₄]-Pc, respectively. The simple first-order kinetics is reflected in both the decay in T₁-like ESA signal (red dots) and the recovery in ground-state bleaching (blue dots) for [Fe₈O₄]-BP0 in Fig. 6B. The T₁-CT decay constant is five orders of magnitude shorter than that of an individual T₁ state in pentacene or bipentacene molecules (11). Because no fluorescence emission is observed for any of the cluster-pentacene complexes, we assign the fast decay in the T₁-CT state to nonradiative recombination. Both CT across the pentacene-cluster interface and the presence of paramagnetic Fe atoms can couple to electron spins, thus facilitating recombination (61).

Unlike the strong coupling of individual T₁ in pentacene or bipentacene to the CT state at their interfaces to [Fe₈O₄], we find that the triplet state in the tightly bound ¹(TT) in BP0 does not undergo CT to the electron accepting cluster. Figure 7A shows TA spectra for [Fe₈O₄]-BP0 as a function of pump-probe delay, following initial photoexcitation at hν₁ = 2.1 eV. Figure 7B shows horizontal cuts at selected pump-probe delays (Δt = 0, 10, and 100 ps), along with a T₁ spectrum obtained from sensitization of [Fe₈O₄]-BP0. At this excitation photon energy, BP0 is known to undergo efficient singlet fission (11), and the results for [Fe₈O₄]-BP0 are nearly identical to those in BP0. Initially (Δt = 0 ps; red spectrum in Fig. 7B), the TA spectrum is that of S₁ characterized by the broad ESA in the visible region and a vibronically resolved ESA in the near-IR region. The singlet exciton decay and triplet rise in [Fe₈O₄]-BP0 are both characterized by a single-exponential lifetime of τ_SF = 0.55 ± 0.02 ps, which is slightly shorter than the corresponding process in BP0 (τ_SF = 0.76 ps) (11). Figure 7D compares the ¹(TT) decay dynamics in [Fe₈O₄]-BP0, as monitored by the decays of ESA signals attributed to both triplet (2.36 eV, red) and singlet (0.97 eV, green) characters. For comparison, we also show in Fig. 7D the ¹(TT) decay dynamics in BP0 (2.36 eV, blue). The three decay traces are superimposable. The data for [Fe₈O₄]-BP0 are well described by a single-exponential decay with a time constant of τ_TT = 0.42 ± 0.03 ns, which is, within experimental uncertainty, identical to that of BP0. In stark contrast to the efficient CT from an individual T₁ state in [Fe₈O₄]-Pc, there is no measurable CT from the tightly bound ¹(TT) state in [Fe₈O₄]-BP0.

Fig. 7. TA spectra and dynamics of [Fe₈O₄]-BP0 under 2.1 eV excitation.

(A) 2D pseudocolor plot of TA (= −ΔT/T; T, transmission) as a function of pump-probe delay (Δt) and probe photon energy. (B) TA spectra at Δt = 0 ps (red), 10 ps (blue), and 100 ps (green), along with T₁ spectrum from sensitization (gray). (C) Singlet fission dynamics, as represented by S₁ decay at 2.07 eV (red) or ¹(TT) buildup at 2.36 eV (blue). (D) Comparison of ¹(TT) decay dynamics for [Fe₈O₄]-BP0 and BP0.

In summary, using covalently linked pentacene dimers as model systems, we show evidence for a tightly bound triplet pair state, which reveals its delocalized ¹(TT) and localized T₁ characters in the near-IR and visible ESA spectra, respectively. The near-IR ESA spectra can be assigned the ¹(TT)→S_n transition, which is similar to the S₁→S_n′ transition, with vibrational progression corresponding to the well-known aromatic ring breathing mode. The ¹(TT)→S_n transition is an indicator of the intertriplet coupling strength; when a phenylene spacer is inserted between the pentacene moieties (BP1) or varies the angle between the pentacene moieties (BP45, BP90, and 1,2-BP) to decrease this coupling, we find that the ¹(TT)→S_n ESA peak decreases. This is in contrast to the spectrum in the visible region, assigned to the ¹(T₁T₁)→¹(T₁T₃) transition present with similar intensities for all bipentacene molecules. Using an electron-accepting iron oxide molecular cluster [Fe₈O₄] linked to pentacene and bipentacene (BP0), we find that electron transfer to the cluster occurs efficiently from an individual T₁ but not from the ¹(TT) state. Thus, the tightly bound ¹(TT) state exhibits a distinctively different chemical reactivity from that of an individual T₁ state. A viable strategy to efficiently harvest triplets from intramolecular singlet fission is to control the intertriplet electronic coupling via molecular design.

MATERIALS AND METHODS

Synthesis

The synthesis of TIPS-Pc, BP0, and BP1 molecules (11); BP0 with different dihedral angles (17); and the [Fe₈O₄] cluster (49) has been previously described. To install the pentacene-based ligands on [Fe₈O₄], we first deprotonated the pendent phenol group with an excess of sodium hydride in tetrahydrofuran (THF). The reaction mixture was filtered through a 0.2-μm syringe filter and added dropwise to a solution of [Fe₈O₄] in THF. We used a 1:1 stoichiometric ratio of the ligand to [Fe₈O₄] to prepare the monosubstituted clusters, which were purified by reversed-phase chromatography. Additional synthetic details and characterization data can be found in the Supplementary Materials.

Optical absorption

The TIPS-Pc, BP0, and BP1 samples were dissolved in dry toluene (with a concentration of <100 μM) and kept free from oxygen and moisture for optical measurements on a Shimadzu UV 1800 UV-Vis Spectrophotometer. Ultaviolet-visible (UV-Vis) absorption spectra of BP0 and BP1 (Fig. 1B) showed a slight red shift from that of TIPS-Pc but contained otherwise nearly identical vibronic features near the absorption threshold (S₀→S₁) (11). Solutions of [Fe₈O₄], [Fe₈O₄]-Pc, or [Fe₈O₄]-BP0 in chloroform were used for absorption measurements. Optical absorption spectra of [Fe₈O₄], [Fe₈O₄]-Pc, and [Fe₈O₄]-BP0 in Fig. 1C will be discussed later.

Transient absorption

To investigate singlet fission and triplet transfer, we used femtosecond TA (fs-TA) spectroscopy. The samples were dissolved in dry toluene and kept free from oxygen and moisture. The pump pulse came from an optical parametric amplifier (tunable from UV to the near-IR, 100-fs pulse width, 1 kHz rep rate). The probe pulse was a white-light supercontinuum (from 450 to 850 nm and from 850 to 1600 nm for the visible and near-IR range, respectively). The delay between pump and probe pulses was controlled by a translational stage with a delay time up to 3 ns. The detection consisted of a pair of multichannel detector arrays coupled to a high-speed data acquisition system (HELIOS, Ultrafast System Inc.). The sample solution was at room temperature during measurement. The nanosecond-microsecond TA measurements were carried out on the same setup as fs-TA with the same pump pulse. The probe pulse was a white-light supercontinuum (from 400 to 1600 nm) generated by a supercontinuum laser (Leukos). The laser pulse width was ≤1 ns at 2 kHz. The pump-probe delay was controlled electrically.

The triplet-sensitizing experiment was carried out on the same setup except for the fact that the white-light probe beams were generated by a picosecond laser and the pump-probe delay was controlled electrically. A mixture of a (bi)pentacene compound and an excess of anthracene was dissolved in toluene with the concentration of anthracene ~100× that of (bi)pentacene. Photoexcitation at 3.35 eV created singlets in anthracene, which underwent intersystem crossing to form triplets. The triplets in anthracene subsequently transferred to (bi)pentacene molecules via diffusional collisions on a time scale of 1 to 2 μs (see the Supplementary Materials).

SUPPLEMENTARY MATERIALS

Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/3/7/e1700241/DC1

Transient absorption

Triplet-sensitizing experiments

Compound synthesis

fig. S1. Transient absorption (TA) spectra and dynamics of [Fe₈O₄]-Pc.

fig S2. Transient absorption for Fe₈O₄pz₁₂Cl₄ cluster (pumped at 2.58 and 2.07 eV) and [Fe₈O₄]-Pc (pumped at 1.65 eV).

fig. S3. Triplet-sensitizing experiments.

fig. S4. Synthetic route for compound 2.

fig. S5. Synthetic route for compound 3.

fig. S6. Synthetic route for compound Pc-Phenol.

fig. S7. Synthetic route for compound BP0-Phenol.

fig. S8. Synthetic route for compound BP1-Phenol.

fig. S9. Synthetic route for compound [Fe₈O₄]-Pc, [Fe₈O₄]-BP0, and [Fe₈O₄]-BP1.

fig. S10. Infrared spectra.

fig. S11. Absorption spectra.

fig. S12. Normalized absorption spectra.

fig. S13. Normalized absorption spectra.

fig. S14. NMR spectrum (0–9.5 ppm), compound 2.

fig. S15. NMR spectrum (0–145 ppm), compound 2.

fig. S16. NMR spectrum (0–9.5 ppm), compound 3.

fig. S17. NMR spectrum (0–145 ppm), compound 3.

fig. S18. NMR spectrum (0–9.5 ppm), Pc-Phenol.

fig. S19. NMR spectrum (0–155 ppm), Pc-Phenol.

fig. S20. NMR spectrum (0–9.5 ppm), BP0-Phenol.

fig. S21. NMR spectrum (0–155 ppm), BP0-Phenol.

fig. S22. NMR spectrum (0–9.5 ppm), BP1-Phenol.

fig. S23. NMR spectrum (0–160 ppm), BP1-Phenol.

fig. S24. NMR spectrum (−40 to 55 ppm), [Fe₈O₄]-Pc, before solvent addition.

fig. S25. NMR spectrum (−40 to 55 ppm), [Fe₈O₄]-Pc, after solvent addition.

fig. S26. Negative and positive mode NMR spectra (1000 to 5000 m/z), [Fe₈O₄]-BP0.

fig. S27. Negative and positive mode NMR spectra (1000 to 5000 m/z), [Fe₈O₄]-BP1.