Triphenylamine-Based Push-Pull Molecule for Photovoltaic Applications: From Synthesis to Ultrafast Device Photophysics.

Small push-pull molecules attract much attention as prospective donor materials for organic solar cells (OSCs). By chemical engineering, it is possible to combine a number of attractive properties such as broad absorption, efficient charge separation, and vacuum and solution processabilities in a single molecule. Here we report the synthesis and early time photophysics of such a molecule, TPA-2T-DCV-Me, based on the triphenylamine (TPA) donor core and dicyanovinyl (DCV) acceptor end group connected by a thiophene bridge. Using time-resolved photoinduced absorption and photoluminescence, we demonstrate that in blends with [70]PCBM the molecule works both as an electron donor and hole acceptor, thereby allowing for two independent channels of charge generation. The charge-generation process is followed by the recombination of interfacial charge transfer states that takes place on the subnanosecond time scale as revealed by time-resolved photoluminescence and nongeminate recombination as follows from the OSC performance. Our findings demonstrate the potential of TPA-DCV-based molecules as donor materials for both solution-processed and vacuum-deposited OSCs.

Introduction

The efficiency of bulk heterojunction (BHJ) organic solar cells (OSCs) based on small molecular (SM) donors[1−5] is constantly increasing, with efficiencies of over 10% achieved.[1,6−11] In addition to high efficiencies, SMs demonstrate unique benefits over more conventional polymers such as high purity, batch-to-batch reproducibility, well-defined molecular structure, molecular weight, and so on.[12−14] Moreover, low-molecular-weight SMs make vacuum processability possible,[15−18] which can be utilized to create highly efficient OSCs with low energy disorder to facilitate long-range exciton transport.[19,20]

An important advantage of SMs is flexibility in molecular design,[21] which allows for fine and precise tuning of the chemical and photophysical properties. In the design concept introduced by Tereniziani[22] and Roncali,[23] a triphenylamine (TPA) donor core and dicyanovinyl (DCV) acceptor end groups led to improved solubility, better layer-to-layer stacking in films, and high hole mobility. High performance of >4% in vacuum-evaporated OSCs[24,25] highlights the potential of this approach; however, such molecules potentially suffer from insufficient stability because of the presence of an active vinyl proton in the DCV group. The substitution of the active proton in DCV by an alkyl improves the stability while retaining all other benefits, as has been previously shown for the series of star-shaped[26−31] and linear molecules.[31,32] A similar strategy was very recently undertaken by Bakiev et al.,[33] who reported the synthesis and basic photophysical properties (absorption and photoluminescence spectra in solution) of a TPA-2T-DCV-Me molecule with methyldicyanovinyl as the acceptor group (Figure a). This molecule can be considered to be an asymmetrical analog of star-shaped molecule (SSM) N(Ph-2T-DCV-Me)3,[26] which has already demonstrated the power conversion efficiency (PCE) to be as high as 4.8%.

Figure 1

(a) Chemical structures of the TPA-2T-DCV-Me small molecular donor and [70]PCBM electron acceptor. (b) Frontier orbital energy levels of TPA-2T-DCV-Me and the [70]PCBM acceptor[34] as measured by cyclic voltammetry.

In this work, we extend the synthesis strategy developed recently for the star-shaped[30,31] and linear[31,32] molecules for the preparation of TPA-2T-DCV-Me and present a comprehensive study of the properties of the molecule. We further focus on early-time photophysics of BHJ thin films based on TPA-2T-DCV-Me with [70]PCBM, as explored by ultrafast photoinduced absorption (PIA) spectroscopy and time-resolved photoluminescence (PL). We demonstrate that charge generation proceeds through the charge transfer (CT) state between TPA-2T-DCV-Me and [70]PCBM and that CT state recombination competes with the charge-generation process, which results in ∼50% efficiency of generation of the long-lived charges. This value is weakly dependent on the BHJ composition because of the interplay between electron transfer from TPA-2T-DCV-Me to [70]PCBM on the subpicosecond time scale and diffusion-delayed hole transfer from [70]PCBM to TPA-2T-DCV-Me on the tens/hundred picosecond time scales. For solar cell operation, the optimal donor/acceptor ratio amounts to 1:2 because of the competition between charge generation and nongeminate recombination due to too fine a BHJ morphology. These findings highlight the potential of TPA-2T-DCV-Me for photovoltaic applications with a possibility of manufacturing both solution-processed and vacuum-deposited OSCs.

Experimental Section

Materials

Tetrakis(triphenylphosphine)palladium(0)Pd(PPh3)4, malononitrile was obtained from Sigma-Aldrich and used without further purification. Pyridine, THF, toluene, and hexane were dried and purified according to the known techniques and then used as a solvent. (4-Bromophenyl)diphenylamine was obtained as described in ref (35). 2,5,5-Trimethyl-2-[5′-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)-2,2′-bithien-5-yl]-1,3-dioxane was obtained as described in ref (36). All reactions, unless stated otherwise, were carried out under an inert atmosphere.

Synthesis

The synthesis route toward TPA-2T-DCV-Me (Figure ) used our approach developed recently for the synthesis of star-shaped[30,31] and linear[31,32,36] oligomers and is based on the preparation of precursors with dioxane or dioxolane protective groups followed by a deprotection reaction and Knœvenagel condensation with malononitrile. The extension of this approach presented herein to the synthesis of unsymmetrical molecules such as TPA-2T-DCV-Me proves a high universality of the method as compared to an alternative synthesis approach based on Vilsmeier–Haack–Arnold acylation followed by Knœvenagel condensation as reported recently by Bakiev et al.[33]

Figure 2

Synthesis of TPA-2T-DCV-Me. For a more detailed description of the synthesis and characterization, refer to Supporting Information.

The synthesis of TPA-DCV-Me consists of the following three steps. First, a precursor with a 5,5-dimethyl-1,3-dioxane-protected carbonyl function (3) was prepared via Suzuki coupling between (4-bromophenyl)diphenylamine (1)[35] and 2,5,5-trimethyl-2-[5′-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)-2,2′-bithien-5-yl]-1,3-dioxane (2)[36] in 70% isolated yield. Removing the protective group was successfully achieved by treatment of the ketal (3) with 1 M HCl to give the corresponding ketone (4) in 98% isolated yield. Finally, TPA-2T-DCV-Me was prepared by Knœvenagel condensation of the ketone with malononitrile in 80% isolated yield.

The measured solubility of TPA-2T-DCV-Me in o-dichlorobenzene (ODCB) at room temperature was found to be 25 g/L, which is a factor of 8 higher as compared to its star-shaped analog N(Ph-2T-DCV-Me)3.[31] This fact is explained by weaker intermolecular interactions for the former case due to the unsymmetrical architecture complicating the ordering of molecules in the bulk. The high solubility of TPA-2T-DCV-Me excludes the necessity to complicate the synthesis by the addition of any alkyl chains to triphenylamine or thiophene blocks, which is one of the advantages of such a design concept.

Sample Preparation

For the photovoltaic blends’ preparation, TPA-2T-DCV-Me and [70]PCBM (Solenne BV) were dissolved separately in ODCB at a concentration of 25g/L. A standard ODCB solvent was used to ensure the dissolution of the [70]PCBM acceptor, which is needed to maximize the OSC efficiency. The solutions were stirred on a magnetic stirrer for at least 20 h at 50 °C. The donor solution was then mixed with [70]PCBM solution in volume ratios of 3:1 to 1:20. Solutions with different mixing ratios were again stirred on the magnetic stirrer for at least 2 h at 50 °C. Thin films were then spin-coated (800 rpm, 2 min) on a microscopic cover glass. The optical densities of all films at the excitation wavelength were ∼0.2–0.8. All measurements were performed within 1 month after sample preparation to prevent the degradation effects; no variations in the results obtained were observed during this period.

For the solid solution samples, PMMA (Sigma-Aldrich, Mw ≈ 120 000 g/mol) was dissolved in ODCB at a concentration of 150 g/L. The solution was stirred on a magnetic stirrer at 50 °C for at least 5 h. Then the solution was mixed with an o-dichlorobenzene solution of TPA-2T-DCV-Me with a molar concentration of 1:50, which corresponds to one TPA-2T-DCV-Me molecule per 60 000 PMMA monomers.

Optical

Linear absorption spectra were obtained using a PerkinElmer Lambda 900 UV/vis/NIR spectrometer. Polarization-sensitive ultrafast PIA measurements were performed on a setup based on a Spectra-Physics Hurricane Ti:sapphire laser and two optical parametric amplifiers (Light Conversion TOPAS) operating in the visible (475 to 2600 nm) and infrared (1.2 to 20 μm) regions. The outputs of these amplifiers were used as excitation and probe beams. The wavelengths of the excitation and probe beams were set at 520 nm and 1.3 μm, respectively, in accordance with the linear and polaron absorption spectra of TPA-2T-DCV-Me (Figures and 6).

Figure 3

Absorption spectra of the representative blends with different TPA-2T-DCV-Me/[70]PCBM weight ratios (indicated). For other absorption spectra, refer to SI, Figure S3a.

Figure 6

Polaron absorption spectra of the 1:1 TPA-2T-DCV-Me/[70]PCBM blend at different delays (indicated in the legend) reconstructed from the transients measured at different probe wavelengths after 520 nm excitation. Symbols represent experimental data; solid lines are the fits of experimental data with a Gaussian function. The fitting parameters (central energy E0 and standard deviation σ) are listed in the SI, Table S1.

The polarization of the IR probe beam was rotated by 45° with respect to the polarization of the excitation beam. After the sample, the parallel and perpendicular components of the probe beam (T∥ and T⊥) were selected by two wire-grid polarizers and then detected by two nitrogen-cooled InSb photodiode detectors. The isotropy population signal and photoinduced anisotropy were calculated by the following expressions:[37]

The average angular displacement between the polarization of incoming light and the photoinduced dipole moment was calculated by the following expression, in which α represents the average angular displacement and r0 is the maximal possible anisotropy of 0.4:[37]

PL measurements were performed with a Hamamatsu C5680 streak-camera system with a time resolution of ∼6 ps. The excitation wavelength of 520 nm was produced by selecting a 10-nm-wide portion of the white light supercontinuum generated from a Ti:sapphire laser (Mira) output in a Newport SCG-800 hollow fiber. The time-resolved PL was collected in 90° geometry with respect to the excitation beam. To cut off the excitation light, an OG570 colored-glass long-pass filter was placed in front of the polychromator. No degradation of the samples was observed during the whole measurement time.

Fabrication and Characterization of the OSCs

All devices were fabricated in the normal architecture. The indium tin oxide (ITO)-covered glass substrates (from Osram) were cleaned in toluene, acetone, and isopropyl alcohol. After drying, the substrates were doctor-bladed with 40 nm PEDOT/PSS (HC Starck, PEDOT PH-4083). Photovoltaic layers, consisting of small-molecule donor TPA-2T-DCV-Me and fullerene acceptor [70]PCBM with different D/A weight ratios in ODCB (10 mg mL–1 in total) were bladed on top of the PEDOT/PSS layer. The thickness of the relevant blend films was ca. 80–90 nm. Finally, 15 nm Ca and 100 nm Al were thermally evaporated through shadow masks to form an active area of 10.4 mm2. The current–voltage characteristics of the solar cells were measured under AM 1.5G illumination from an OrielSol1A solar simulator (100 mW cm–2). The light source was calibrated using a reference silicon cell.

Results and Discussion

Thermal, Electrochemical, and Optical Properties

The thermal (Supporting Information (SI), Figure S1) and electrochemical (SI, Figure S2) behaviors of TPA-2T-DCV-Me are very similar to those of its star-shaped analogs.[31,32]

Decomposition temperatures (Td, corresponding to 5% weight losses) of TPA-2T-DCV-Me calculated from the thermogravimetric analysis (TGA) data were found to be 362 and 364 °C in air and under an inert atmosphere, respectively, indicating very high thermal and thermooxidative stability. The phase behavior of TPA-2T-DCV-Me was found to be very similar to that of its star-shaped analog, N(Ph-2T-DCV-Me)3.[31] TPA-2T-DCV-Me as received is in a crystalline state and has a relatively high melting temperature (212 °C vs 270 °C) and melting enthalpy (114 J/g vs 82 J/g) (Figure S1b). However, after melting it becomes amorphous, similar to other TPA-based analogs, and has a glass-transition temperature of 67 °C, which is more than a factor of 2 lower as compared to the star-shaped analog.[31] The difference in the glass-transition temperatures between the two materials can be explained by a significantly higher molecular weight of the star-shaped molecule.

The highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) of TPA-2T-DCV-Me were calculated from cyclic voltammetry (CV) in film as −5.35 and −3.35 eV, respectively. Thus, the electrochemical band gap is calculated to be 2 eV, which is slightly wider as compared to the star-shaped analogs,[31,32] probably due to the presence of a partial conjugation between oligothiophene arms in the star-shaped molecule. It should be noted that the HOMO and LUMO positions in the film are slightly shifted as compared to those in the MeCN–CH2Cl2 9:1 solution (−5.20 and −3.15 eV, respectively)[33] due to intermolecular interactions in the bulk.

Linear absorption spectra of the studied films are shown in Figure . The neat TPA-2T-DCV-Me film exhibits two prominent absorption bands located at ∼365 and ∼520 nm. Similar to the N(Ph-2T-DCV-Me)3 SSM,[38] both bands have a mixed CT and π–π* character:[38−40] in the excited state, the electron density is delocalized over the DCV acceptor and thiophene linkers.[33] Note that intermolecular interactions in the solid film also affect the absorption band position and red shift the absorption spectrum by ∼20 nm as compared to the noninteracting molecules diluted in the PMMA matrix (SI, Figure S3b) as a result of the different polarizability of the environment.[41] The maximum of the low-energy absorption band of TPA-2T-DCV-Me in the PMMA matrix is 12 nm blue-shifted as compared to the N(Ph-2T-DCV-Me)3 SSM (∼494 vs ∼506 nm, respectively, SI, Figure S3). The reason for this is that in a conjugated system the band gap decreases with increasing conjugation length. As the nitrogen atom in the TPA core allows for a partial conjugation,[42,43] the effective conjugation length in SSM is increased compared to that in TPA-2T-DCV-Me, therefore leading to a red-shifted absorption.

In the blends with [70]PCBM, the absorption spectra extend more to the blue wavelengths because of the increased contribution of [70]PCBM absorption. In addition, in the 630–720 nm region, the TPA-2T-DCV-Me/[70]PCBM blends demonstrate an absorption shoulder even though the neat TPA-2T-DCV-Me and [70]PCBM films are transparent (Figure , SI, Figure S4). This most probably indicates the formation of ground-state charge-transfer complexes (CTCs), which leads to the decreased band gap and thus to the red-shifted absorption.[44] A similar effect was previously observed in BHJ blends of SSMs.[38] Overall, the blends efficiently absorb visible light up to ∼650 nm, covering a significant amount of the solar irradiation.

Time-Resolved Photoluminescence Measurements

Because the charge-separation process in the BHJ OSCs typically occurs via intermolecular CT states,[45−47] the branching ratio between the photogenerated excitons and the ones transferred to the CT states can be used as a benchmark for long-lived charge-generation efficiency.[47,48] PL spectroscopy has been proven to be a simple and reliable tool for estimating the efficiency of CT state generation.[47−49] As a figure of merit (FOM), the ratio of CT state PL to residual singlet PL has been proposed.[47,48] A correlation between FOM and OSC efficiency was demonstrated,[47,48] with the conclusion that the blend with the maximized FOM is optimal for OSC functioning.

Because [70]PCBM absorbs a significant number of photons in the visible and near-UV regions, the charges in [70]PCBM-based OSCs are generated from both the donor phase and [70]PCBM phase with comparable efficiencies.[34,38,50−53] Consequently, the definition of FOM[47] should be modified to include both TPA-2T-DCV-Me and [70]PCBM phaseswhere ICT is the intensity of CT PL, I[70]PCBM is the intensity of [70]PCBM singlet PL, and ISM is the intensity of TPA-2T-DCV-Me singlet PL.[47]

Figure a shows the PL spectrum of the 1:5 TPA-2T-DCV-Me/[70]PCBM blend after 520 nm excitation (for other blends, see SI, Figure S5). To calculate the FOM index, it is necessary to factorize the contribution of CT states and the combined contributions of [70]PCBM/TPA-2T-DCV-Me (eq ) in overall PL.

Figure 4

(a) Time-integrated PL spectra of neat [70]PCBM (blue), neat TPA-2T-DCV-Me (green), and the 1:5 TPA-2T-DCV-Me/[70]PCBM (red) blend after 520 nm excitation. The spectra are arbitrarily normalized by the [70]PCBM PL maximum at ∼720 nm. For the spectra of all blends in the 550–900 nm spectral region, refer to Figure S5a. Regions of [70]PCBM/TPA-2T-DCV-Me PL and CT PL are shaded in green and red, respectively. (b) FOM (eq ) for the blends at different TPA-2T-DCV-Me/[70]PCBM contents.

The PL spectrum consists of two bands at ∼680–730 nm and ∼750–850 nm and an extended shoulder at <680 nm. The 680–730 nm band is clearly present in PL spectra of pristine [70]PCBM and therefore is assigned to PL of [70]PCBM. The residual PL from TPA-2T-DCV-Me, which peaks at ∼680 nm,[41] results in the high-energy shoulder. Both contributions are strongly quenched in TPA-2T-DCV-Me/[70]PCBM blends of any concentration (SI, Figures S5, S6), indicating efficient dissociation of both [70]PCBM and TPA-2T-DCV-Me excitons.

The 750–850 nm PL band is present in blends with [70]PCBM as well as in pristine [70]PCBM films but not in pristine TPA-2T-DCV-Me films. Therefore, it could be attributed to either PL of intermolecular CT states or residual PL of [70]PCBM. However, the decay kinetics of [70]PCBM PL integrated in the 690–710 nm region (i.e., where it is well-separated from the 750–850 nm band) and the 750–850 nm band are essentially different (Figure ). [70]PCBM PL is strongly quenched and decays on the sub-100-ps time scale. (Note that the exact decay time is wavelength-dependent as a result of the dynamical spectral relaxation[41,54−56] and is slower in the red flank of [70]PCBM PL.) Photoluminescence in the 750–850 nm region, after the initial fast component corresponding to the remaining [70]PCBM emission, decays noticeably longer (Figure , red curve) . Furthermore, the PL decay of the 750–850 nm band does not depend on the TPA-2T-DCV-Me/[70]PCBM ratio (Figure S5b), indicating an identical decay channel for blends of any composition. Therefore, the 750–850 nm spectral band is assigned to PL of the CT state. Similar CT state dynamics were also observed for the previously studied by PIA absorption[38] N(Ph-2T-DCV-Me)3/[70]PCBM blends (SI, Figure S7).

Figure 5

PL dynamics of the 1:5 TPA-2T-DCV-Me/[70]PCBM blend obtained by integration of the PL in the 810–850 nm region (dominated by CT PL, red line) and in the 690–710 nm region (dominated by quenched [70]PCBM PL, blue line) after 520 nm excitation. The transients are normalized by their maxima. The black lines show triexponential (CT contribution) and biexponential ([70]PCBM contribution) fits. Note the strong PL quenching in the region of [70]PCBM PL.

To calculate FOMs, the PL intensities were time-integrated over the 800–850 nm (the red flank of CT PL) and 650–750 nm ([70]PCBM and TPA-2T-DCV-Me PL) spectral ranges. Figure b shows FOMs calculated for the studied blends at different TPA-2T-DCV-Me/[70]PCBM ratios (blue symbols/line). The FOM dependence exhibits clear peaklike behavior: it rapidly grows until a 1:4–1:5 ratio is reached and decreases further on. Note, however, that because of (i) a high but yet finite spectral contrast between CT states and excitons at the given wavelength and (ii) different PL cross sections of the excitons and CT states, FOM does not provide the actual branching ratio between CT states and photogenerated excitons but serves rather as an estimation of the optimal concentration range.

It does not seem possible to draw any conclusions about the exact pathway of PL decay of the CT state from the PL measurements alone: it can be due to CT state back recombination to the ground state, back electron transfer to the donor/acceptor phase with formation of a singlet or triplet exciton,[57,58] or/and CT state separation into charges. In all scenarios, the CT PL decays in time whereas obviously the latter scenario is more favorable for OSC operation. To directly observe the charge-separation processes, we performed time-resolved PIA measurements, which allow tracking the charge-separation dynamics.[50,59−62] With the PIA experiments, the two scenarios can be easily distinguished: in the case of CT state recombination, a gradual decrease in the PIA signal is expected.[27,59,62]

Photoinduced Infrared Absorption

To track the dynamics of photogenerated charges, the PIA technique was used. In the TPA-2T-DCV-Me phase, photogenerated charges (TPA+ cations) create an additional polaron-induced[38] absorption band in the near-IR region, the magnitude of which is proportional to the population of the photogenerated charges.[63] Depending on the molecule, the photoinduced absorption bands are usually located in the 1–3 μm region, with the band energy inversely proportional to the conjugation length (P1 = C + A/L, where P1 is the photoinduced absorption band energy, C and A are constants, and L is the conjugation length).[38,64,65]

Figure shows the PIA IR spectra for the 1:1 TPA-2T-DCV-Me/[70]PCBM blend at different pump–probe delays. The PIA band is located at around 1.3 μm. Within first 10 ps, the blue shift of the spectrum of about 0.02 eV occurs (Figure , Table S1). Similar behavior was observed in the PIA spectrum in a pristine film of TPA-2T-DCV-Me (SI, Figure S8). This blue shift is attributed to the transient polarizability of the environment and/or the interplay between different photoexcited species (namely inter- and intramolecular CT excitons with different lifetimes).[38] In the 1:1 blend, CT excitons dissociate within ∼100 ps (Figure S6); therefore, the spectral dynamics at earlier times are caused by the residual CT exciton response. Nevertheless, because the spectral dynamics are very minor in blended films, a single probe wavelength was set at the polaron absorption maximum of 1.3 μm.

Photoinduced Absorption Dynamics

Photoinduced charge dynamics in BHJs are determined by both intrinsic (intramolecular) photophysics of the molecules and their intermolecular interactions. The lifetime of the excited state in diluted TPA-2T-DCV-Me molecules amounts to ∼2.1 ns (Figure S9). When the molecules are densely packed in the solid film, the excited-state depopulation becomes biexponential with lifetimes of 135 and 900 ps (Table ) due to the self-quenching effect (Figure a, red; see ref (41) for details).

Table 1

Fit Parameters (Equation ) for the Neat TPA-2T-DCV-Me Film and Blends with [70]PCBM

	offset A0	A1	τ1 [ps]	A2	τ2 [ns]	A3	τ3 [ns]	∑ Ai
neat	0.1 ± 0.1	0.12 ± 0.02	135 ± 8	0.35 ± 0.02	0.9 ± 0.2			0.57 ± 0.12
3:1	0.45 ± 0.1	0.12 ± 0.02	100 ± 20	0.3 ± 0.03	0.8 ± 0.1			0.87 ± 0.15
1:1	0.4 ± 0.1			0.4 ± 0.05				0.8 ± 0.1
1:2	0.25 ± 0.05			0.4 ± 0.05		0.2 ± 0.1	1.5 ± 0.5	0.85 ± 0.2
1:4	0.25 ± 0.05			0.5 ± 0.1		0.25 ± 0.05	5 ± 1	1 ± 0.2
1:5	0.15 ± 0.03			0.5 ± 0.1		0.25 ± 0.05	8 ± 2	0.9 ± 0.2
1:10	0.1 ± 0.02			0.3 ± 0.1		0.35 ± 0.05	5 ± 2	0.75 ± 0.1
1:20	0 ± 0.03			0.4 ± 0.05		0.4 ± 0.05	23 ± 5	0.8 ± 0.1

Figure 7

(a) Isotropic transients for TPA-2T-DCV-Me/[70]PCBM blends at representative [70]PCBM concentrations (as indicated; transients for all concentrations are presented in SI, Figure S10) after 520 nm excitation. The circles represent the experimental data, and the solid lines show the best fits according to eq . All transients are normalized to the number of absorbed photons. (b) Share of long-lived charges (green symbols) and respective contributions from electron transfer (red symbols) and hole transfer (blue symbols) for the blends at different TPA-2T-DCV-Me/[70]PCBM ratios. The contributions of electron and hole transfer processes are derived from the Monte Carlo simulations of the data (Figure S11). Lines are to guide the eye.

When the external acceptor ([70]PCBM) is added to the blend, the dynamics become more complex because now they involve both intramolecular dynamics in the donor and acceptor phases and intermolecular charge separation/recombination processes. For the intermolecular charge transfer, the change Δ in the Gibbs free energy between the excited and charge-separated states is calculated as[66−68]where IPD is the donor ionization potential (in the first approximation, equal to the modulus of the TPA-2T-DCV-Me HOMO energy), EAA is the acceptor electron affinity (approximately the modulus of [70]PCBM LUMO energy), and Eexciton is the energy of the photogenerated exciton (∼2.07 eV, equal to the energy of the 0–0 transition). Thus, ΔG for photoinduced charge transfer in the TPA-2T-DCV-Me/[70]PCBM system amounts to −1 eV. This value well matches ΔG for the P3HT/PCBM system[69] and is somehow higher than the optimal ΔG of −0.3 to −0.8 eV for polymer-based OSCs.[67] Therefore, even though ΔG is clearly sufficient for charge separation, the total charge collection efficiency might be decreased by the system being in the Marcus inverted region, where the increase in ΔG leads to the reduced charge-separation rate.[67]

Figure shows the PIA isotropic transients for TPA-2T-DCV-Me with different fullerene loadings. The amplitude of the PIA signal is proportional to the concentration of charges (holes) at the TPA-2T-DCV-Me molecules and the polaron absorption cross-section. The transients were normalized by the film absorption to obtain the relative value of generated charge per absorbed photon. The maximal amplitude among the transients was arbitrarily assigned to the unity charge yield. To quantitatively describe the charge dynamics, all transients were fitted with the following multiexponential functionconvoluted with a Gaussian apparatus function with standard deviation of σ ≈ 100 fs. Here, A represents the amplitude of the decay (i = 1, 2) or growing (i = 3) component with a lifetime of τ whereas A0 stands for the offset (i.e., the number of long-lived charges). Similar recombination channels were assumed in all blends except for neat TPA-2T-DCV-Me so that τ2 was set as a global fit parameter for both PIA (Figure ) and PL kinetics (Figure ). The fitting parameters are listed in Table .

For all films, the PIA signals build up within the experimental resolution time of 100 fs at a zero pump–probe delay. This is due to the ultrafast formation of the CT exciton in the TPA-2T-DCV-Me phase and/or the instantaneous splitting of the interfacial excitons in blended films.[38,64,70]

Transients for the pristine sample and 3:1 blend demonstrate a fast decay component on the subnanosecond time scale (τ1), indicating that a certain number of photogenerated excitons and/or charges recombine. This decay component diminishes at high [70]PCBM loadings and is also present in the neat TPA-2T-DCV-Me film, which allows us to rule out the possibility of back electron transfer from [70]PCBM to TPA-2T-DCV-Me. Nonetheless, a similar decay component was not found for the solid solution of TPA-2T-DCV-Me molecules in the PMMA matrix (SI, Figure S9)[71] where intermolecular interactions are negligibly low.

Interestingly, PL in the neat TPA-2T-DCV-Me film decays much faster compared to the PIA signal (∼40 ps vs 135 ps, Figure S6), indicating that PIA dynamics are most likely driven not by the emitting excitons that produce PL but by some other species. The fact that the PIA spectrum of neat TPA-2T-DCV-Me (SI, Figure S8) is very similar to the PIA spectrum of the 1:1 BHJ blend (Figure ) points toward the effective formation of polaron pairs[72] separated between neighboring TPA-2T-DCV-Me molecules rather than intramolecular CT excitons. This may be beneficial for OSC operation because the binding energy of the polaron pairs is lower than that of the intramolecular CT excitons due to a longer spatial separation between the electron and the hole and increased screening. Therefore, we attribute the fast decay to the intermolecular recombination of charges that were initially separated by the neighboring TPA-2T-DCV-Me molecules. A similar recombination channel was previously observed in SSM-based blends, suggesting analogous charge recombination dynamics.[38]

For the blends with higher [70]PCBM loading, the initial decay is replaced by the gradual growth of the PIA signal on the picosecond time scale. The amplitude and time of this component increase with increasing concentration of [70]PCBM. Because [70]PCBM absorbs a significant fraction of photons at the excitation wavelength of 520 nm (SI, Figure S4e), it is reasonable to attribute this growth to the diffusion-delayed hole-transfer process.[38,70] This is in line with the quenching of [70]PCBM PL in BHJ blends (Figure S5b). Therefore, in the BHJ films the charges are efficiently generated via both electron transfer from TPA-2T-DCV-Me to [70]PCBM and hole transfer from [70]PCBM to TPA-2T-DCV-Me.

The diffusion-delayed hole-transfer dynamics can be used to extract valuable information about the BHJ morphology,[56] namely, the characteristic sizes of [70]PCBM domains in the mixed [70]PCBM/TPA-2T-DCV-Me phase. Briefly, the domain sizes are retrieved from Monte Carlo (MC) simulations of the exciton hopping diffusion in spherical [70]PCBM domains with the domain size used as a fit parameter (details in ref (56)). The MC simulations perfectly reproduce the diffusion-delayed growth for the whole region of [70]PCBM concentrations (SI, Figure S11a). The extracted sizes of [70]PCBM domains vary from ∼3 nm (3:1 and 1:1 blends) to ∼20 nm (1:20 blend, SI, Figure S11b), which indicates the fine intercalation between the two constituencies.

For all blends, decaying components on subnanosecond time scales are observed. Interestingly, neither the time scales nor the amplitudes of this component depend on the particular TPA-2T-DCV-Me/[70]PCBM ratio. A similar decay was also observed in the PL measurements (Figure ); therefore, it is attributed to the recombination of the CT states. Note that decay of the PIA signal clearly points to the back recombination of the CT states but not to the separation of the CT state into charges. Therefore, charge recombination via the CT state seems to be the main loss channel in the TPA-2T-DCV-Me/[70]PCBM blends. This corroborates our previous finding of charge recombination is SSM-based blends,[27,38] although without having specified explicitly the CT state pathway (SI, Figure S7).

Efficiency of Long-Lived Charge Generation

In the operating OSC, the maximal amount of charges which can be extracted from the singlet excitons is determined by the amount of separated charges generated at the nanosecond time scale. In the TPA-2T-DCV-Me/[70]PCBM system, recombination of CT states competes with charge generation thereby limiting the final amount of long-lived charges. PL measurements estimate the optimal blend compositions (Figure b); however, FOM does not necessarily reflect the amount of separated charges due to overlapping PL spectra of the exciton and CT states as well as recombination of the CT states. Additionally, some excitons can dissociate directly to charges without the CT state as an intermediate, and therefore are not taken into account by FOM. Overall, FOM readily captures the processes of CT states formation and the amount of wasted excitons but does not provide a grasp on the amount of the separated charges. The amount of long-lived separated charges, however, can be derived from the fits of PIA data (Table ) as the sum A0 + A3 of the long-time offset A0 and the ingrowing component A3.

For the neat TPA-2T-DCV-Me film, the number of long-lived charges is very low (∼10%, Table ). This is because most of the intra- and intermolecular CT excitons cannot be separated any further and therefore eventually recombine. Only a small portion (∼10%) of the charges survive, most probably due to either local traps or long-range intermolecular dissociation.

As [70]PCBM is added to the blend, the share of long-lived charges quickly increases to a relatively high level of ∼50% (Figure b, green symbols). The reasons for such a tremendous increase are the following. On one hand, adding the [70]PCBM acceptor opens a new pathway for charge separation as CT excitons from TPA-2T-DCV-Me are split at the BHJ interface (Figure b, red symbols) due to the sufficient energy offsets for electron transfer (Figure b). On the other hand, [70]PCBM also absorbs photons and generates excitons, therefore contributing to the long-lived charges via the hole-transfer process (Figure b, blue symbols).

The total number of long-lived charges is weakly dependent on the TPA-2T-DCV-Me/[70]PCBM ratio, with the small decrease at higher fullerene loadings notwithstanding. This is explained by the interplay between electron- and hole-transfer processes due to fine intermixing between TPA-2T-DCV-Me and [70]PCBM domains. The relatively large size of [70]PCBM domains in BHJs with the highest [70]PCBM content causes the decrease in free charge generation (Figure b) as ∼20% of generated [70]PCBM excitons are lost in 20 nm [70]PCBM domains.[56]

The difference between FOM (Figure b) and the number of long-lived charges (Figure b) is most probably related to the noticeable contribution from the polaron pairs to the separated charges, as was discussed above. For the blends with high [70]PCBM contents, the formation of the polaron pairs is likely to be suppressed due to (i) the smaller size of TPA-2T-DCV-Me domains and therefore the higher probability of the intramolecular CT exciton reaching the interface before forming the polaron pair and (ii) the higher contribution of [70]PCBM excitons to the separated charges.

Photoinduced Anisotropy Dynamics

The initial inter- and intramolecular charge separation within a push–pull molecule plays a significant role in the further generation of separated charges.[72,73] Isotropic PIA dynamics provide information about the intermolecular charge separation but do not shed any light on the intramolecular dynamics. Photoinduced anisotropy, in turn, provides important information on the evolution of orientations of the transient dipole moments as the resulting dynamics reflect the depolarization of excited states of any nature be it due to the charge migration, the isotropic hole transfer process, and/or intrinsic intramolecular depolarization of the excited state. With this in mind, we measured photoinduced anisotropy dynamics in the TPA-2T-DCV-Me-based films and in a solid solution of TPA-2T-DCV-Me molecules diluted in the PMMA matrix (Figure ).

Figure 8

Photoinduced anisotropy of neat TPA-2T-DCV-Me film (red), a 1:2 film of TPA-2T-DCV-Me/[70]PCBM (blue), and a solid solution of TPA-2T-DCV-Me in the PMMA matrix (green) after 520 nm excitation, measured (open symbols) and fitted with biexponential decay functions (solid lines). To reduce the noise, log-averaging of the data was performed; refer to SI, Figure S13 for the unprocessed data. Other blends have similar dynamics (SI, Figure S13).

In the pristine TPA-2T-DCV-Me film, the initial anisotropy value is ∼0.2 (Figure ), which is twice as low as the maximal possible value of 0.4. As the intermolecular interactions are strong in the film, intramolecular CT excitons can be formed after photoexcitation and intermolecular CT exciton formation is possible, which leads to lower anisotropy values. The further decay of the anisotropy within the first ∼50 ps is likely due to the exciton migration within the TPA-2T-DCV-Me film.

To verify these conclusions, the anisotropy dynamics of separated TPA-2T-DCV-Me molecules in the PMMA matrix were also measured (Figure ). A constant anisotropy value of ∼0.35 was observed, which clearly points toward the intermolecular nature of the depolarization process in the pristine films: on the ultrafast time scale, the depolarization is due to the isotropic intermolecular charge separation, while at longer time scales the anisotropy decays because of the exciton migration. This is in sharp contrast to N(Ph-2T-DCV-Me)3 SSM, where the anisotropy decays to ∼0.15 within the first picosecond (SI, Figure S12). The value of 0.15 corresponds to the average angular displacement of ∼40° between the polarization of the incoming light and the photoinduced dipole moment of the polarons (eq ).[37] This is due to the presence of degenerate excited states in the SSM with different distributions of the electron density between the arms, which leads to ultrafast depolarization of the excited states to the mixed state.[38] In contrast, no intramolecular depolarization occurs in TPA-2T-DCV-Me because of asymmetric molecular structure.

For the TPA-2T-DCV-Me/[70]PCBM blends, the initial anisotropy is slightly lower as compared to that of the neat TPA-2T-DCV-Me film (Figure ). This is explained by the noticeable contribution from the interfacial [70]PCBM exciton dissociation via the hole-transfer process, which is essentially isotropic.[70] Further on, the anisotropy decays to zero due to both exciton migration within the TPA-2T-DCV-Me phase and diffusion-delayed hole transfer from [70]PCBM. Overall, in the TPA-2T-DCV-Me-based films not only intramolecular CT excitons but also intermolecularly separated charges are formed after photoexcitation, which leads to the low initial anisotropy and its decay in time.

Device Performance

The early time charge generation essentially determines the maximal number of available charges whereas for the OSC operation charge extraction and nongeminate charge recombination losses are also important. With this in mind, we measured photon-to-charge conversion efficiencies of OSCs (Figure a). The efficiency peaks at 1:1–1:2 TPA-2T-DCV-Me/[70]PCBM ratios (Figure a), which is in contrast with the rather flat, long-lived charge dependence on BHJ composition (Figure b, green line). A low short-circuit current of 5.7 mA cm–2 and the fill factor of 35% (Figure b) strongly suggest severe nongeminate recombination,[74] which is most probably caused by too fine an intermixing of TPA-2T-DCV-Me and [70]PCBM and a reduced number of intercalated pathways for the charges to reach the electrodes. Nonetheless, the open-circuit voltage (Voc) of the OCSs is as high as 0.9 eV, which implies low Voc losses of <0.5 eV. Perhaps this is connected to the effective formation of polaron pairs in TPA-2T-DCV-Me, which reduces the Coulomb attraction between electrons and holes and also diminishes the number of filled interfacial CT states.[75] Low Voc losses are characteristic of the whole family of TPA-based molecules,[27] which potentially makes this molecular design attractive for the manufacturing of OSCs in combination with further morphology optimization and interfacial engineering.

Figure 9

(a) Power conversion efficiency dependence on the TPA-2T-DCV-Me/[70]PCBM ratio. The data represent statistics over 10 devices. The bars represent 5% (95%), the boxes represent 25–75%, the horizontal line inside the box is the median value, and the dot inside the box is the mean value. (b) J–V for the best solar cell based on the 1:2 TPA-2T-DCV-Me/[70]PCBM blend. The device parameters are indicated next to the curve.

Conclusions

In this article, we have reported the efficient synthesis of the push–pull molecule with the TPA donor core and methyl-DCV acceptor[33] by the extension of the universal synthetic platform developed previously for the star-shaped[30,31] and linear molecules.[31,32] Because of its push–pull nature, the molecule demonstrates broad absorption in the UV–vis region, which is slightly blue-shifted as compared to its star-shaped counterpart, N(Ph-2T-DCV-Me)3.

In photovoltaic blends, the charge-generation process competes with geminate charge recombination, which is the main loss channel on the ultrafast time scale. Using time-resolved PL, we have shown that the charges are trapped at the interfacial CT states, which recombine on the subnanosecond time scale, and a half of the charges are lost thereby.

Using ultrafast PIA spectroscopy, we have demonstrated that the intramolecular CT exciton formation and electron transfer occur within the first ∼100 fs, whereas the diffusion-delayed dissociation of [70]PCBM excitons takes up to ∼20 ps. Eventually, both electron- and hole-transfer channels contribute to the long-lived charges with complementary shares at different blend compositions. From the photoinduced anisotropy measurements, we have demonstrated that in the films of pristine molecules depolarization of the excited states occurs due to intermolecular interactions. This is in sharp contrast with symmetrical star-shaped molecules,[38] where extremely fast depolarization occurs due to intermixing of the degenerate excited states.

In the solar cells, charge extraction is limited by the nongeminate recombination that reduces the fill factor of ∼35%. Nonetheless, efficient intermolecular charge separation highlights the advantage of such molecular design, which in combination with device engineering has potential for organic solar cell applications.