Evidence for conical intersection dynamics mediating ultrafast singlet exciton fission

Transcription

1 Evidence for conical intersection dynamics mediating ultrafast singlet exciton fission Andrew J Musser, Matz Liebel, Christoph Schnedermann, Torsten Wende Tom B Kehoe, Akshay Rao, Philipp Kukura Methods Materials 6,13-bis(triisopropylsiylethynyl)pentacene [CAS ] >99%, (TIPS-pentacene) was purchased from Sigma Aldrich and used as received. The material was dissolved in toluene (5 wt%) and spin-cast onto quartz substrates at 300 rpm. 1,8-diphenyl-1,3,5,7-octatetraene (DPO) was purchased from Sigma Aldrich and used as received. The material was dissolved in CHCl 3 to give an optical density of 1 at 390 nm in a 200 m flowcell used for spectroscopic measurements. Time-Resolved Spectroscopy Laser pulses were delivered by a Pharos-6W amplifier (1030 nm, 180 fs, 5.5 W at 10 khz). Focusing 1.5 μj of the output onto a 3 mm sapphire plate generates the chirped white-light (WL) probe (300 fs, 3 nj) after removal of the residual fundamental. The remainder of the amplifier output is frequencydoubled, followed by sum frequency generation to generate 250 μj at 515 nm and 35 μj at 343 nm. The latter pumps a WL-seeded non-collinear optical parametric amplifier (NOPA) that generates the broadband impulsive Raman pump ( nm, 12 nj) for the pump-dump-probe experiment 1 on TIPS-pentacene. The pulse was compressed to 10 fs using chirped mirrors (Layertec). A four-stage WL-seeded NOPA pumped at 515 nm (250 μj) generated narrow bandwidth pulses (ca. 10 nm) centred at 860 nm that are chirped by means of a 4f grating stretcher to make the dump (650 fs) tuned to the TIPS-pentacene T 1 T 2 absorption band. The time delay between pump and dump pulses was fixed at 300 fs. In the pump-raman-probe ( triplet reference ) experiment, the actinic pump was generated by frequency doubling the fundamental of the Pharos output (pulses at 515 nm, 200 fs). The impulsive Raman pump ( nm, ~10 fs, 24 nj), resonant with the T 1 T 2 transition, was generated by a 515 nm pumped NOPA 1. The time delay between actinic and impulsive Raman pump pulses was fixed at 20 ps, to allow for complete vibrational cooling of the system. For pump-dumpprobe measurements on DPO, the pump ( nm, 14 fs, 30 nj) was obtained by frequencydoubling the output of a 515 nm pumped NOPA 1 ( nm, ~10 fs) in a 25 m BBO crystal. The dump pulse (200 fs, 80 nj) was centred on the prominent 2A g PIA at 660 nm. NATURE PHYSICS 1

2 In all experiments, the delay of the probe pulse is varied in fs steps with a closed-loop delay stage (PI) for a total of 2.5 ps with a 100 ms integration time per step. Transient absorption traces in the absence and presence of pump and dump/impulsive Raman pulses are obtained by employing a double chopping scheme. A homebuilt single-shot prism spectrograph equipped with a CMOS array detector (LW-ELIS-1024A-1394) is used for broadband detection. In the pump-dump-probe experiment on TIPS-pentacene, the diameters of the probe, pump and dump beams were 55 μm, 64 μm and 120 μm, with pulse energies at the sample adjusted to 3 nj, 14 nj, and 210 nj, respectively. In the pump-raman-probe experiment, the diameters of the probe, pump and Raman beams were 42 μm, 77 μm and 78 μm, with pulse energies of 3 nj, 24 nj and 100 nj, respectively. For the measurements on DPO, the diameters of the probe, pump and dump beams were 35 μm, 55 μm and 80 μm, with pulse energies of 3 nj, 30 nj and 80 nj, respectively. CW Raman Spectroscopy Stokes-shifted Raman spectra of TIPS-pentacene films were recorded using a Horiba T64000 Raman system. The excitation was a helium-neon laser line at nm which was focussed on to the TIPSpentacene film with a Leica 50x, 0.55 numerical aperture, long working-distance microscope objective. The laser power at the sample was approximately 130 μw. The sample was placed in a Linkam chamber through which a small flow of nitrogen gas was maintained. The Stokes-shifted Raman spectra of DPO were measured with a conventional Raman spectrometer (PI SpectraPro 300i coupled to a PI PIXIS 256E CCD camera) under 532 nm (Laser Quantum SMD600) illumination. The system was operated in back-scattering geometry using a 40x objective for collection. Data Analysis We followed the same basic procedure to isolate the vibrational coherence signatures in the transient absorption data sets of TIPS-pentacene and DPO. First, the slowly varying electronic population kinetics were subtracted, using a global fit of multi-exponential and high-order polynomial functions. Subsequent checks of the Fourier transform of the kinetics confirmed that this subtraction introduced no new oscillations with frequencies above 400 cm -1, the range analysed in this work. The spectrally resolved frequency content was determined by taking the Fourier transform of these oscillation maps, individually for each probe wavelength. In order to isolate excited state vibrational coherences, the oscillation map from the Dump ON (three-pulse) configuration was directly subtracted from that of the Dump OFF (two-pulse) measurement, and the Fourier transform of the resulting differential map was taken in the same manner. To avoid contamination with the pronounced but short-lived oscillations around t=0 caused by the coherent response of the system, we only analysed vibrational coherence data well after this artefact. 2 NATURE PHYSICS

3 SUPPLEMENTARY INFORMATION Figure S1: Triplet exciton rise kinetic. a Reproduction of the colour map in Figure 2b, showing early-time dynamics in the two-pulse experiment, as well as assignment (right) of the primary spectral features. b Integrated rise kinetic for the T 1 T 2 absorption band ( nm). The sharp peaks around t=0 are due to the coherent artefact, and the underlying electronic population kinetics can be clearly distinguished. The monoexponential rise (black) of 80 fs is consistent with previous reports. NATURE PHYSICS 3

4 Figure S2: Vibrational coherence from impulsive excitation. a Integrated time-trace from the twopulse experiment within the T 1 T 3 band, with a fit (black) to the slowly varying population kinetics. The residual (pale blue), scaled by four for clarity, shows a strong coherent artefact around t=0 followed by small, fast oscillations that persist for picoseconds. b A similar population subtraction performed across the two-pulse spectrum yields a spectrally resolved oscillation map. To avoid contamination by the oscillatory nature of the coherent artefact, in this and all other vibrational datasets the data was truncated from 80 fs onwards. 4 NATURE PHYSICS

5 SUPPLEMENTARY INFORMATION Figure S3: Three-pulse experiment. Raw data from the three-pulse experiment is largely indistinguishable from the two-pulse experiment. Comparison with Figure 2a reveals slight differences in the presence of the dump pulse, for instance in the GSB band at 625 nm from 500 fs onwards. However, these small perturbations do not significantly alter the singlet fission kinetics or the magnitude of the triplet population. Higher dump intensities were avoided to prevent contamination from two-photon excitation. NATURE PHYSICS 5

6 Figure S4: Isolation of excited state vibrations. a Integrated kinetics for a section of the T 1 T 2 absorption band are essentially indistinguishable for Dump OFF and Dump ON configurations. Only a slight separation between the two traces is evident, once the triplet population reaches its maximum. In the Dump OFF Dump ON residual (following removal of the slowly varying dump-induced kinetics), the fast oscillations seen previously are significantly reduced in magnitude and grow until the full dump pulse has passed, up to a peak ΔT/T amplitude of 5 x The residual is scaled 30x for clarity. b The OFF ON residual is built up across the entire spectrum and over the full delay range (up to 2.2 ps). These oscillations are extremely small, approximately 30x weaker than in the two-pulse experiment in Figure S2b. The triplet absorption bands show fast vibrations, most clearly distinguished at 550 nm and > 750 nm, while no such features can be seen in the central GSB region. 6 NATURE PHYSICS

7 SUPPLEMENTARY INFORMATION Figure S5: Direct generation of triplet vibrational coherence. a In a control experiment, vibrational coherence is directly generated in the triplet state. An initial pump pulse at 515 nm (blue, 200 fs) initiates singlet exciton fission and produces a large triplet population. The compressed Raman pulse follows after a fixed 20 ps delay, to ensure any pump-induced vibrations have fully relaxed. The Raman pulse (purple, 10 fs) is resonant with the T 1 T 2 absorption band and generates vibrational coherence in the triplet state. This is effectively equivalent to performing the two-pulse experiment on a population of TIPS-pentacene triplet excitons. b The same primary transient absorption features of triplets and ground state bleaching can be identified as previously. Zero delay here corresponds to the arrival of the impulsive Raman pulse. c Following subtraction of electronic population kinetics, the residual of Pump OFF Pump ON shows strong vibrational activity, confirming the generation of vibrational coherence in the excited state. d The Fourier transform of the oscillatory data in c reveals the same vibrational frequencies, with the same spectral selectivity, identified in the main text. Frequency spectra integrated over the indicated bands are shown in Figure S6. The fast Raman pulse in this experiment directly excites the T 1 T 2 transition, but also generates substantial ground-state vibrational coherence. A neat subtraction of Pump ON from Pump OFF data does not completely remove this background, as the two experiments are fundamentally different: the first is a simple transient absorption measurement on a ground-state film, while the second is a transient absorption measurement on a mixture of ground-state and excited-state molecules. To account for this difference, in the subtraction the Pump ON dataset was scaled by to minimize the residual signature of ground-state vibrational coherence. NATURE PHYSICS 7

8 Figure S6: Spectral selectivity of vibrational coherence. Integration across the spectral bands indicated in Figures 2 and S5 shows the same vibrational modes throughout the dataset. Variation in the relative intensities indicates overlap of ground-state and excited-state vibrational coherence signatures as well as different vibrational selection rules for the T 1 T 2 and T 1 T 3 transitions. The ground-state bands are removed in the three-pulse experiment (dark lines), leaving no discernable activity in the GSB region. The same main vibrational bands are observed through direct coherence generation (dashed) in the reference measurement. The spectra for the GSB and T 1 T 2 bands in the ES coherence and triplet reference experiments are scaled by a factor of 5 for clarity. 8 NATURE PHYSICS

9 SUPPLEMENTARY INFORMATION Comparison to Theoretical Models As discussed in the main text, the key result of this work the existence of a conical intersection between the singlet and triplet pair manifolds was predicted by Zimmerman et al. 2 on the basis of ab initio calculations of pentacene monomers and dimers in a range of geometries. These results were subsequently expanded 3 to evaluate the effect of exciton delocalization, also in excellent agreement with our experimental results. This does not mean that our result uniquely agree with the direct model of singlet exciton fission. A recent computational study by Casanova 4 on clusters of tetracene derivatives similarly found that nuclear motion could drive the process of singlet exciton fission by inducing crossings between the relevant electronic states. This work ruled out the classic two-step mediated mechanism of sequential charge-transfer events but concluded that indirect coupling to charge-transfer configurations was likely to play an important role, i.e. the superexchange variant of the mediated model. Similarly, an ab initio study of tetracene and pentacene dimers by Feng et al. 5 found energetic crossing of the singlet and triplet-pair states is induced by nuclear motion, and while the coupling between initial and final states was nominally direct, the contribution of charge-transfer configurations to both was considered essential to drive the process. It is thus clear that both direct and superexchange-mediated mechanisms could be consistent with the experimental results presented here. A direct comparison with many of the literature models is, however, inappropriate, due to the common approach of using model Hamiltonians to describe the singlet fission process and coupling between the relevant electronic states. This technique requires the explicit selection of a number of physically motivated diabatic states (such as singlet, triplet and charge-transfer configurations). The properties of these states are not a function of nuclear position, a fundamental restriction which automatically excludes vibronic coupling and conical intersections from the resulting descriptions of singlet fission, as was recognised by Berkelbach et al. 6. Our results do rule out a two-step mechanism involving charge transfer states, as this would be inconsistent with our observation of the transfer of vibrational coherence from singlet to triplet pair state. Furthermore, our observation of a clear growth in the triplet exciton population over tens of fs is inconsistent with the fully coherent picture of fission, where all singlets and triplet pairs form a superposition state populated directly by the initial photoexcitation and there is no subsequent growth in triplet population. Critically, the vibronic dynamics in TIPS pentacene during singlet fission are essentially identical to those observed in DPO where the electronic structure is well-established and understood to involve straightforward internal conversion between two electronic states of different symmetry. We emphasize that one-photon excitation into the S 1 state of DPO is forbidden by symmetry, similar to direct excitation into the 1 TT state in pentacene. Strong mixing of S 2 and S 1 states would necessarily NATURE PHYSICS 9

10 result in very low transition dipoles, which are not observed experimentally, in either DPO or pentacene, suggesting that the states are indeed clearly distinguishable in the Franck-Condon region. In summary our experimental results on delocalisation and conical intersections highlight the importance of accounting for nuclear dynamics in describing the ultrafast regime of singlet fission. Future experimental and theoretical studies are clearly called for to fully elucidate the role of vibronic coupling and driving modes in determining the rate and efficiency of fission. References 1. Liebel, M., Schnedermann, C. & Kukura, P. Sub-10-fs pulses tunable from 480 to 980 nm from a NOPA pumped by an Yb : KGW source. Opt. Lett. 39, (2014). 2. Zimmerman, P. M., Zhang, Z. & Musgrave, C. B. Singlet fission in pentacene through multiexciton quantum states. Nat. Chem. 2, (2010). 3. Zimmerman, P. M., Bell, F., Casanova, D. & Head-Gordon, M. Mechanism for singlet fission in pentacene and tetracene: from single exciton to two triplets. J. Am. Chem. Soc. 133, (2011). 4. Casanova, D. Electronic structure study of singlet-fission in tetracene derivatives. J. Chem. Theory Comput. 10, (2014). 5. Feng, X., Luzanov, A. V. & Krylov, A. I. Fission of Entangled Spins: An Electronic Structure Perspective. J. Phys. Chem. Lett. 4, (2013). 6. Berkelbach, T. C., Hybertsen, M. S. & Reichman, D. R. Microscopic theory of singlet exciton fission. II. Application to pentacene dimers and the role of superexchange. J. Chem. Phys. 138, (2013). 10 NATURE PHYSICS