Report

Coherent ultrafast charge transfer in an organic photovoltaic blend

See allHide authors and affiliations

Science  30 May 2014:
Vol. 344, Issue 6187, pp. 1001-1005
DOI: 10.1126/science.1249771

Pull, pull, pulling electrons along

Organic photovoltaics operate by transferring charge from a light-absorbing donor material to a nearby acceptor. Falke et al. show that molecular vibrations smooth the way for this charge transfer to proceed. A combination of ultrafast spectroscopy and theoretical simulations revealed an oscillatory signal in a model donor/acceptor blend that implicates carbon-carbon bond stretching in concert with the electronic transition. This vibrational/electronic, or vibronic, process maintains a quantum-mechanical phase relationship that guides the charge more rapidly and directly than an incoherent migration from donor to acceptor.

Science, this issue p. 1001

Abstract

Blends of conjugated polymers and fullerene derivatives are prototype systems for organic photovoltaic devices. The primary charge-generation mechanism involves a light-induced ultrafast electron transfer from the light-absorbing and electron-donating polymer to the fullerene electron acceptor. Here, we elucidate the initial quantum dynamics of this process. Experimentally, we observed coherent vibrational motion of the fullerene moiety after impulsive optical excitation of the polymer donor. Comparison with first-principle theoretical simulations evidences coherent electron transfer between donor and acceptor and oscillations of the transferred charge with a 25-femtosecond period matching that of the observed vibrational modes. Our results show that coherent vibronic coupling between electronic and nuclear degrees of freedom is of key importance in triggering charge delocalization and transfer in a noncovalently bound reference system.

The currently accepted model for the basic working principle of a bulk-heterojunction organic solar cell (1, 2), comprising a conjugated polymer donor and an electron acceptor material, relies on four elementary steps: (i) photon absorption, creating a spatially localized, Coulomb-bound electron-hole pair (exciton) in the donor phase; (ii) exciton diffusion to the donor/acceptor interface; (iii) exciton dissociation at the interface leading to the formation of a charge-separated state (3, 4), often called charge-transfer exciton or polaron pair; and (iv) dissociation of the polaron pair into free charges and their transport to the electrodes.

In this work, we focused on the dynamics of the primary light-induced steps, (i) and (iii), which lead to a charge-separated state in organic photovoltaic (OPV) materials and represent the key process in OPV cells. Over the past years, charge photogeneration has been investigated in several technologically relevant materials, such as blends of polyphenylene-vinylene (5, 6), polythiophene (7, 8), or low band gap polymers (9, 10) with fullerene derivatives. In all of these systems, it is now accepted that charge separation is an ultrafast process occurring on a sub-100-fs time scale. So far the experimental studies on charge photogeneration in OPV materials have mainly been described within the framework of an incoherent transfer model (11, 12), giving a rate constant for the transfer process. These rate constants may be enhanced by hot exciton dissociation (10, 13). Recently, several theoretical studies have simulated the electronic structure (1416) and charge transfer in this class of systems (1719) by ab initio and/or model approaches and point toward an important role of vibronic quantum coherence for the charge separation (17, 18). In biological (2022) and in some prototypical artificial (23, 24) light-harvesting systems, quantum coherence phenomena have recently been observed experimentally, and this has marked a breakthrough in the description of the primary processes of energy and charge transfer in macromolecular complexes. However, still very little is known about the role of quantum coherence at room temperature in the earliest stage of the dynamics in technologically relevant OPV materials. Recent experiments found evidence for an ultrafast long-range charge separation in such systems but could not differentiate between coherent and incoherent charge-transfer models (25).

We studied the ultrafast optical response of a reference OPV material system by combining high time-resolution pump-probe spectroscopy and time-dependent density functional theory (TDDFT) simulations. We observed that the ultrafast electron transfer from the polymer triggers coherent vibrational motion of the fullerene and constitutes the primary step of the photoinduced charge-separation process.

We investigated thin films of the conjugated polymer poly-3-hexylthiophene (P3HT), the fullerene derivative [6,6]-phenyl-C61 butyric acid methyl ester (PCBM), and P3HT:PCBM blends with 1:1 mixing ratio in weight, prepared by spin coating from chlorobenzene solutions. Such blended films are a prototypical material for OPV cells (26, 27), and power conversion efficiencies of up to 5% have been reported (28). The absorption spectrum of a P3HT:PCBM blend (Fig. 1A, solid line) appears mainly as a linear superposition of those of the two separate components (Fig. 1B, solid lines) (29), indicating that no direct charge-transfer transitions occur in the ground state. Weak absorption features below the lowest excitonic resonance (30, 31) may reflect defects or charge-transfer exciton transitions associated to specific local configurations.

Fig. 1 Linear optical and electronic properties of the materials.

(A) Normalized thin-film absorption cross section of P3HT:PCBM 1:1 ratio in weight (solid) together with the theoretical 4T:C60 absorption spectrum (dashed). No signature of ground-state mixing in the blend is observed. (B) Normalized experimental absorption cross sections of PCBM (solid blue) and pristine P3HT (solid red) and the theoretical absorption of C60 (dashed blue) and 4T (dashed red). (C) Molecular structure of the P3HT:C60 blend system. The sulfur atoms belonging to the thiophene rings are depicted in yellow, whereas the carbon atoms are shown in gray. (D) Calculated electronic Kohn-Sham levels of (from left to right) C60, 4T, and 4T:C60 blend.

In order to gain insight into the primary photoinduced charge-transfer dynamics of the blend, we performed ultrafast spectroscopic studies on such thin films by using a two-color pump-probe spectrometer providing independently tunable pulses (32). Because the charge-carrier photogeneration in P3HT:PCBM blends is essentially independent of temperature (33), all experiments were performed at room temperature. The overall time resolution of the setup is better than 15 fs. Pump pulses centered at 540 nm resonantly excite the Embedded Image absorption band of P3HT (30, 34), whereas broadband probe pulses monitor the transient absorption in the blue-to-green wavelength range.

Figure 2A shows a measurement of the pump-induced change in optical transmission (Embedded Image) for the P3HT:PCBM blend as a function of probe wavelength Embedded Image and pump-probe delay Embedded Image. A similar measurement for the pristine P3HT film is reported in the supplementary materials. For probe wavelengths between 500 and 525 nm, the signal is in both samples dominated by ground-state bleaching of the optically excited exciton transition in the polymer. In the shorter wavelength range, the dynamics are substantially different. There, the exciton bleaching of the polymer becomes less prominent, and we have, in the blend, access to an additional stimulated emission signal from the fullerene or intermediate states. In this region, the blended sample displayed an additional and fast-decaying component on a 100-fs time scale (supplementary materials), which can be assigned to an ultrafast charge transfer from the polymer to the fullerene. This time scale is in agreement with previous reports on similar blends (5) and with the results of detailed pump-probe measurements covering the probe wavelength range between 550 and 1400 nm. These results (supplementary materials) show that a substantial fraction of all photogenerated excitons in the blend undergo rapid charge separation on a 50- to 70-fs time scale.

Fig. 2 Charge transfer dynamics of the P3HT:PCBM blend.

(A) Experimental differential transmission (Embedded Image) map of the P3HT:PCBM blend as a function of probe delay and probe wavelength. The pronounced oscillations in the ΔT/T signal reflect coherent vibrational wave-packet motion initiated by the short pump pulse. (B) Fourier transform spectra of the ΔT/T dynamics of the blend (left) and pristine P3HT (right). The spectral intensity is amplified by a factor of 4 for λ = 498 to 485 nm and by a factor of 20 for λ = 485 to 470 nm. (C) Integrated Fourier transform spectra for λ = 520 to 498 nm (top) and λ = 492 to 485 nm [bottom, dashed lines in (B)] of the blend (black) and pristine P3HT (red). The dashed vertical lines indicate the frequency of the P3HT C=C stretch mode (red) at 1450 cm−1 and pentagonal-pinch mode of the fullerene (black) at 1470 cm−1.

For both the pristine P3HT and the blend samples, the Embedded Image map (Fig. 2A and fig. S2) shows a pronounced oscillatory contrast throughout the entire visible range. We further analyzed these Embedded Image data by taking the Fourier transform of the oscillatory component after subtraction of a slowly varying background. In the 500- to 520-nm wavelength region (Fig. 2, B and C), we observed, for both the polymer and the blend, the characteristic C=C stretching frequency of the polymer (1450 cm−1, corresponding to a vibrational period of 23 fs). For shorter probe wavelengths, the behavior became more complex: In the pristine film we saw almost no contrast, whereas in the blended film we detected a strong oscillatory component at a higher frequency of 1470 cm−1. This frequency corresponds to the pinch mode dominating the Raman spectrum of the PCBM film (35, 36). Also in the 470- to 480-nm probe wavelength range, we found vibrational characteristics of the fullerene. There, we saw a weak vibrational mode at 1289 cm−1, corresponding to the T1g(3) mode of PCBM. Control experiments on pristine PCBM films (fig. S3) did not show any evidence of coherence, thus ruling out direct PCBM excitation as a cause for the observed oscillations at 1470 cm−1. These results are difficult to reconcile with an incoherent charge-transfer model, which predicts a gradual and monotonous buildup of charge on the fullerene acceptor on a 100-fs time scale that will not trigger coherent motion on a faster time scale. Also, they cannot be interpreted in terms of an incoherent charge transfer taking place within less than one vibrational period (23 fs) because such a fast transfer is neither seen in the Embedded Image data in the supplementary materials nor consistent with finding essentially the same linewidths in the absorption spectra of the pristine P3HT film and the blend. Instead, they provide evidence for a coherent charge transfer mediated by strong vibronic couplings between polymer and fullerene.

To analyze the experimental observations, we have performed first-principle calculations (25) on the simplest possible model of the experimentally studied thin-film blend, a periodic crystal of charge transfer dimers. PBCM has been substituted with a C60 molecule, and the alkyl side chains on the polymer have been removed in order to reduce the numerical complexity, checking that this does not affect the ground-state properties of the system. By using density functional theory (DFT) at the local density approximation (LDA) level, we relaxed the ground state of the system and found an equilibrium distance of 3.2 Å between 4T and C60. The Kohn-Sham electron energy levels of the 4T:C60 unit are shown in Fig. 1D together with the electronic structure of the isolated C60 and 4T, all aligned to vacuum. The orbital localization shows a highest occupied molecular orbital (HOMO) with 96% localization on the thiophene chain and a lowest unoccupied molecular orbital (LUMO) almost fully localized on the fullerene. The lowest-lying available single-particle thiophene excitation corresponds to a HOMO-to-LUMO transition with 94% localization on the polymer. Steady-state optical absorption spectra (37, 38) of the blend and the isolated components are depicted in Fig. 1, A and B, respectively. The absorption cross sections of both the isolated moieties and of 4T:C60 are in good agreement with the experimentally observed ones. They consist of two distinct absorption bands, centered at 528 and 354 nm, that correspond to thiophene and fullerene excitations, respectively. This calculation supports the picture of a system composed of moieties that are weakly interacting in their ground states.

TDDFT simulations of the dynamics of the photoexcited 4T:C60 model system were performed by imposing periodic boundary conditions in order to mimic the experimental configuration of a blended thin film. We interrogated the system dynamics by assuming an initial instantaneously excited electronic state corresponding to the removal of an electron from the polymer HOMO and the creation of an electron in the polymer LUMO. We chose as initial condition a Maxwellian distribution of random nuclear velocities to approximate the experimental room-temperature environment. We observed that an electron is transferred to C60 with 60% probability within 97 fs, in good agreement with experimental findings (7, 8). Moreover, the charge-transfer probability, taken as the spatially integrated excess charge density on the C60, oscillated in time with a period of about 25 ± 4 fs (Fig. 3A). This period approximately matches the oscillation frequency observed in the experiments. A useful insight into the charge delocalization mechanism is gained by looking at the first few time-dependent Kohn-Sham eigenvalues above the 4T HOMO (Fig. 3C). The continuous purple line represents the 4T LUMO, occupied at time 0 by one electron removed from the 4T HOMO (purple broken line). Its energy undergoes pronounced oscillations in antiphase with the C60 LUMO. At each crossing of the two levels, the charge density is free to move through the system (Fig. 3C), causing a transient peak in the current flowing between 4T and C60 and back (Fig. 3A). Instead, the current flow is suppressed when 4T and C60 LUMOs are energetically detuned, resulting in a periodic variation of the current flow and thus an oscillatory modulation of amount of charge transferred to the C60 moiety. In the long term, the charge will be localized on the acceptor both by a nuclear rearrangement and by energy dissipation. This regime, however, is out of the scope of the present calculations. Our simulations thus predict that vibronic coupling is necessary for charge transfer to occur and indicate that this coupling is responsible for dynamically driving 4T and C60 LUMOs in resonance, explaining the coherent oscillations of the transferred charge. In fact, the electronic excitation remains fully localized on the polymer when keeping the ions in fixed positions (fig. S4).

Fig. 3 Simulation of the charge transfer dynamics.

(A) Charge-transfer dynamics in a crystal of 4T:C60 aggegrates after impulsive 4T excitation at time zero. After 97 fs, the charge-transfer probability from 4T to C60 is 60%. Strong oscillations of the transfer probability, with a period of about 25 ± 4 fs, are the signature of coherent charge-transfer dynamics. (B) Time dynamics of the molecular dipole. The z component, oriented along the axis from the 4T to the C60, oscillates in phase with the displaced charge. (C) Time-dependent Kohn-Sham eigenvalues. The purple lines refer to the 4T HOMO (broken) and 4T LUMO (solid) levels; the green lines to the lowest unoccupied C60 levels, respectively. (D to F) Snapshots of the simulated time evolution of the charge density in the coupled 4T and C60 LUMOs. Initially (D), the charge is completely localized on the 4T chain. As time evolves, it delocalizes between 4T and C60 [(E) 57.3 fs]. After 98.7 fs (F), the charge density is shared between 4T and C60.

The charge separation is further analyzed by examining the time-dependent dipole moment of the system (Fig. 3B). Its z component is oriented along the axis from the polymer to the C60 and oscillates in phase with the displaced charge, whereas the y component shows weakly anticorrelated oscillations along the polymer chain. This again points to vibronic coherence and a periodic charge flow between polymer and fullerene. The dynamics of the charge separation process are nicely visualized by displaying the time evolution of the electronic density projected onto the coupled LUMO orbitals of the blend (Fig. 3, D to F, and movie S1). This charge density is created by photoexcitation at time zero and is initially fully localized on the polymer (Fig. 3D). As the time evolves, the charge density delocalizes between polymer and C60, and the degree of fractional charge on both moieties displays anticorrelated temporal oscillations. At the end of the simulation, it is shared between donor and acceptor (Fig. 3F).

In the simulations, we also analyzed the time-dependent ionic displacements (movie S2) and found that the photoexcitation of 4T promotes vibrational motion of the C60 with similar oscillation period as is found in the charge transfer probability. This is consistent with the experimental observation of collective vibronic coherence of the fullerene triggered by impulsive photoexcitation of the polymer.

We recently studied the charge-transfer dynamics in a very different system, a supramolecular carotene-porphyrin-fullerene triad (24), a model system for artificial light harvesting. The physical nature of this system, a covalently bound molecular complex in solution, is fundamentally different from the one studied here. Nevertheless, we find similar charge oscillations, with a period of about 30 fs, when studying the charge transfer between porpyhrin and fullerene. We believe that the key to the similarity of those values can be found in Fig. 3C. It shows that the energies of the relevant LUMOs of the polymer and fullerene vary in time with a period corresponding to that of the vibrational mode to which the electronic state is strongly coupled. The probability for charge transfer is large whenever the polymer and fullerene LUMOs are transiently brought into resonance. Hence, the amount of transferred charge (Fig. 3A) oscillates at the (average) period of the vibrational mode(s) most strongly coupled to the electronic system.

In case of the P3HT:PCBM blend, these are the C=C stretch mode of the polymer at 1450 cm−1 and the pentagonal pinch mode of PCBM at 1470 cm−1, corresponding to an oscillation period of ~23 fs. This agrees reasonably well with the period of 25 ± 4 fs that we deduce from Fig. 3, A to C. This period is slightly shorter than the 30 fs that we saw in the case of the triad.

We suggest that the similarity of oscillation periods is a direct consequence of strong vibronic coupling. The time evolution of the charge distribution is modulated with a period matching that of the vibrational modes that are most strongly coupled to the charge excitations. Because all carbon-based organic systems have strong vibrations in the 1000 to 1500 cm−1 range, it is likely to experimentally find oscillation periods between 20 and 30 fs. Our results suggest that, despite the very different microscopic properties of the triad and the P3HT:PCBM blend, the coherent charge-transfer dynamics are in both cases governed by strong vibronic coupling.

A consistent and general picture of the elementary photoinduced charge-transfer process in the P3HT:PCBM blend emerges from our detailed experimental and theoretical results. Optical excitation locally creates an electron-hole pair on the polymer moiety. The strong vibronic coupling between electronic and nuclear degrees of freedom promotes a delocalization of the optically excited electronic wave packet across the interface. Both the electronic density and the nuclei display correlated oscillations on the same time scales, which are essential for an ultrafast charge transfer from the donor to the acceptor. The observation of coherent electron-nuclear motion in a noncovalently bound complex, averaging over a macroscopic ensemble of P3HT:PCBM moieties with variable environment and interfaces, is strong evidence for the dominant role of quantum coherences in the early stages of the charge transfer dynamics in this class of OPV materials.

Supplementary Materials

www.sciencemag.org/content/344/6187/1001/suppl/DC1

Materials and Methods

Supplementary Text

Figs. S1 to S10

References (3948)

Movies S1 and S2

References and Notes

  1. Acknowledgments: Financial support by the European Union project CRONOS (grant number 280879-2), the Deutsche Forschungsgemeinschaft (SPP1391), the Korea Foundation for International Cooperation of Science and Technology (Global Research Laboratory project, K20815000003), and the Italian Fondo per gli Investimenti della Ricerca di Base (Flashit project) is gratefully acknowledged. C.A.R. and E.M. acknowledge the Partnership for Advanced Computing in Europe (project LAIT) for awarding us access to supercomputing resources at CINECA, Italy, and useful discussions with C. Cocchi and Y. Kanai. S.M.F. is grateful for a Ph.D. fellowship from Stiftung der Metallindustrie im Nord-Westen. C.L. and G.C. acknowledge support from the European Community (Seventh Framework Programme INFRASTRUCTURES-2008-1, Laserlab Europe II contract no. 228334); A.R. acknowledges financial support from the European Research Council (ERC-2010-AdG-267374), Spanish grant (FIS2010-21282-C02-01), Grupos Consolidados (IT578-13), Ikerbasque. G.C. acknowledges financial support by the European Research Council (ERC-2011-AdG no. 291198). C.L., S.M.F., G.C., C.A.R., and E.M. initiated this work. S.M.F. prepared the samples. S.M.F., D.B., M.M., E.S., and A.D.S. performed the ultrafast spectroscopy experiments. C.A.R. and M.A. performed DFT and TDDFT simulations. S.M.F., A.D.S., D.B., G.C., and C.L. analyzed and discussed the experimental data. C.A.R., M.A., E.M., and A.R. analyzed and discussed the theoretical data. C.A.R., E.M., G.C., A.D.S., and C.L. designed the paper. All authors discussed the implications and contributed to the writing of the paper. The authors declare no competing financial interests.
View Abstract

Stay Connected to Science

Navigate This Article