Subnanosecond charge photogeneration and recombination in polyfluorene copolymer-fullerene solar cell : Effects of electric field

Influence of electric field on the subnanosecond charge photogeneration dynamics in the polymer solar cell based on polyfluorene copolymer BisDMO-PFDTBT blended with PC61BM was examined with transient absorption spectroscopy. The charge dynamics showed no difference under shortor open-circuit conditions and under a forward bias of 0.79 V (1.6 × 10 V/cm), implying negligible field effects on the subnanosecond dynamics of charge photogeneration/recombination. However, under the reverse biases of −2 V (4.0 × 10 V/cm) and −5 V (1.0 × 10 V/cm), significant enhancement of charge photogeneration and apparent suppression of polaron pair recombination were observed, which agrees with the field-assisted enhancement of external quantum efficiency of the solar cell devices. ©2013 Optical Society of America OCIS codes: (320.7130) Ultrafast processes in condensed matter, including semiconductors; (160.5470) Polymers; (250.2080) Polymer active devices; (300.6500) Spectroscopy, time-


Introduction
The primary charge photogeneration in the bulk heterojunction (BHJ) polymer solar cells involves the formation and diffusion of excitons in the polymer phases, and the subsequent dissociation of excitons at the BHJ interfaces yielding free charge carriers and Coulombically bound polaron pairs (PPs).PPs can further split into free charges or recombine back to the ground or relax to the triplet excited states [1,2].In addition, free charges can be directly photogenerated with significant yields in the bulk polymer phases in ~100 fs [3][4][5][6].Since the filling factor (FF) of an operating solar cell is determined by the comprehensive exciton and charge dynamics under the internal electric field, it is important to examine the influence of electric field on the elementary light conversion processes, especially the PP dissociation and the charge recombination.
To assess the influence of electric field on the dynamics of charge formation and recombination, integrated solar cells biased with external potentials have been investigated by the use of transient photocurrent/voltage [7,8], time-resolved photo-or electro-luminescence [9,10] and optical transient absorption (TA) [11,12] spectroscopies, as well as by using timedelayed collection field (TDCF) and charge extraction by linear increasing voltage (CELIV) [13].Despite the intensive research efforts, the results and conclusions on the field effect remain controversial: Recent TA investigation on the solar cells based on poly-[2,6-(4,4-bis-(2-ethylhexyl)-4H-cyclopenta[2,1-b;3,4-b′]-dithiophene)-alt-4,7-(2,1,3-benzothiadiazole)] (PCPDTBT) and [6,6]-phenyl-C 71 -butyric acid methyl ester (PC 70 BM), as the respective electron donating and the electron accepting materials, claimed negligible field dependence of charge photogeneration and geminate recombination [12].However, TDCF and CELIV studies on the PCPDTBT/PC 70 BM devices confirmed the field-assisted PP dissociation, which facilitates the photogeneration of free charge carriers [13].In the case of poly(3hexylthiophene) (P3HT) based devices, the TA results also showed the field-assisted PP separation in competing with its geminate recombination [11].On the other hand, timeresolved luminescent studies of the blends of poly[2,7-(9,9-dialkylfluorene)-alt-5,5-(4′,7′-di-2-thienyl-2′,1′,3′-benzothiadiazole)] (PF10TBT) and PC 61 BM revealed that upon electrical bias the emission of charge transfer (CT) excitons was substantially quenched and the relaxation kinetics was concomitantly accelerated, indicating that the electric field is indeed influencing on the dissociation of the emissive CT-excitons [10].To the contrary, the biasindependent photoluminescence of the PCPDTBT/PC 61 BM device suggests that the CTexcitons cannot dissociate into free charges under an applied electric field [14].Hence it seems that the aforementioned inconsistency may be related to the variation of photovoltaic materials and the difference in experimental methods.Here, it is worthy of noting that the field-assisted charge photogeneration seems intimately dependent on the morphologies of the photoactive layers: E. g., the field assistance can be enhanced by thermal annealing treatment [11] and by introducing processing additives such as diiodooctane [13] for the respective P3HT/PC 61 BM and PCPDTBT/PC 70 BM devices.
To date, the literature works on the field dependence of the dynamics of charge photogeneration and recombination focus mostly to the ns-ms timeframes.However, the PPs generated directly from photoexcitation live as short as tens to hundreds of picoseconds as a result of dissociation and germinate recombination [3,10,[15][16][17][18][19], and the directly photogenerated PPs may differ from those formed via free charge recombination in the behavior of electric field dependence.The present work is intended to examine the subnanosecond PP dynamics, that is, to focus on the field dependence of the initially formed PPs.We target the polyfluorene copolymer poly([2,7-(9,9-bis-(3,7-dimethyl-octyl)-fluorene)]-alt-[5,5-(4,7-di-2′-thienyl-2,1,3-benzothiadiazole)]) (BisDMO-PFDTBT), a promising light harvester and electron donor giving rise to a power conversion efficiency (PCE) of 4.5% when blended with the electron acceptor PC 70 BM [20].Note that the BisDMO-PFDTBT/PC 70 BM device exhibits moderate FF (~50%) and external quantum efficiency (EQE, ~50%) in the visible spectral region, implying significant dependence of the photocurrent on the internal electric field.Such device is thus suitable for investigating the possible field dependence of the dynamics of PP and polymer radical cation (P + , also referred to as polaron in literature).Because in a nanosecond the free charges would not fully arrive at the electrodes to build static electric fields [21], we biased the solar cell with an external potential to mimic the macroscopic internal electric field of operating device.Furthermore, since the excitation photon fluence used in the TA measurements can be several orders of magnitudes higher than that of the terrestrial solar irradiation, which may introduce unwanted nonlinear effects such as singlet-singlet exciton annihilation and bimolecular charge recombination [12,18,22], we applied low-fluence photoexcitation (~1.0 × 10 13 photons•cm −2 •pulse −1 ) in the TA measurements.Our results have revealed significantly enhanced charge photogeneration and substantially suppressed PP recombination under reverse but not forward potential biases, which agrees with the field dependence of FF and EQE of the BisDMO-PFDTBT/PC 61 BM device.

Device fabrication
The polymer solar cells were constructed on patterned and ITO-coated glass substrates.The ITO surface was modified by spin-coating PEDOT:PSS (Baytron P VP A1 4083) with a thickness of ~40 nm, followed by baking in air at 150 °C for 15 min.BisDMO-PFDTBT blended with PC 61 BM (1:3 w/w) were dissolved in 1,2-dichlorobenzene.Photoactive layers were obtained by spin-coating (3000 rpm, 30 s), and were annealed at 110 °C for 10 min.The typical thickness of the BisDMO-PFDTBT:PC 61 BM layer was ~49 nm as determined by using a surface profilometer (Ambios Technology XP-2).The cathode consisted of 20 nm of calcium and 100 nm of aluminum, which were thermally evaporated to the top of polymer layer with a shadow mask to define an active area of 0.04 cm 2 .The current-voltage (I-V) curves were measured under 100 mW/cm 2 standard AM 1.5 G spectrum using a solar simulator (XES-70S1, San-Ei Electric Co. Ltd.; AAA grade, 70 × 70 mm 2 photo-beam size), for which a 2 × 2 cm 2 monocrystalline silicon cell (SRC-1000-TC-Q; VLSI Standards Inc.) was used for reference.The EQE were measured, with dark currents deducted, by the use of the Solar Cell Spectral Response Measurement System (QE-R3011, Enli Technology Ltd., Taiwan), and the light intensity at each wavelength was calibrated with a standard singlecrystal Si photovoltaic cell.

Time-resolved near infrared absorption spectroscopy
The apparatus with a temporal resolution of 160 fs is briefly described below.An optical parametric amplifier (OPA-800 CF-1, Spectra Physics) pumped by a regenerative amplifier (SPTF-100F-1KHPR, Spectra Physics) provided the actinic laser pulses at desired wavelengths (~120 fs, full width at half maximum).The continuum probe (800~1400 nm) generated from a 3-mm thick sapphire plate was detected after interrogating the excited volume of sample by an InGaAs detector (OMA-V, Princeton Instruments) attached to an imaging spectrographs (SpectraPro 2300i, USA).To ensure that each laser shot excites the sample fully relaxed form the previous excitation, the laser system was run at a repetition rate of 333 Hz.A mechanical chopper (Model 75158, Newport) was set in the pump beam to regulate pump "on" and "off" for a pair of sequential actinic pulses.The crossing angle between the polarization of the pump and probe beams was set to 54.7° (magic angle).To improve the signal-to-noise ratio, each transient spectrum was obtained by averaging 500 individual measurements.The time-resolved measurements were performed in a reflection mode: The pump and probe beams were introduced from the ITO side of the device, passed through the active layer, and then reflected by the aluminum cathode.The polarities of bias potentials, forward (V f ) and reverse (V r ), are defined as that the polarities of the potential source coincide and oppose those of the solar cell, respectively.Figure 1(a) shows the UV-visible absorption spectrum of the BisDMO-PFDTBT/PC 61 BM photoactive layer.The major absorption band at 554 nm (I) is attributed to the optical transition from the ground state to the intramolecular charge transfer state (S 1 ←S 0 ) shifting the electron density from the fluorene unit to the thienyl-benzothiadiazole (TBT) unit of BisDMO-PFDTBT, while the band at 389 nm (II) is ascribed to the absorptive transition to the excitonic state (S 2 ←S 0 ) with the π-electron delocalized over the fluorene-thiophene backbone [23,24].In time-resolved measurements, we applied the bandgap photoexcitation at 610 nm to minimize the photoexcitation of PC 61 BM and the excess energy of the S 1 exciton of BisDMO-PFDTBT, otherwise the former would introduce the diffusive dynamics of fullerene excitation (~100 ps) [25] and the latter would promote the PP dissociation despite a weak contribution [15,26].Figure 1  Figures 2(a) and 2(b) illustrate the TA spectra recorded for the solar cell under short circuit and the forward bias of V f = 0.79 V, respectively, we see two major spectral components in common: The characteristic PP absorption peaking at ~910 nm in the transients at Δt = 0 ps, and the signature P + absorption at ~1050 nm in the transients at Δt = 2.0 ns [3,17,18].Note that the S 1 exciton absorbing at ~1300 nm for the BisDMO-PFDTBT/PC 70 BM blend are rather weak even in the initial transients (0 ps, 0.13 ps), implying the extremely efficient dissociation of the S 1 exciton.This is in accordance to the nanosecondto-subpicosecond shortened fluorescence S 1 -exciton lifetime from the neat to the blend films [19], i.e. the nearly unitary exciton quenching efficiency.The temporal evolution of the TA spectra are similar from the short circuit (Fig. 2(a)) to the forward bias of 0.79 V (Fig. 2(b)).On the other hand, the charge carriers generated by pulsed photoexcitation cannot reach the electrodes within a few nanoseconds to build the macroscopic (internal) electric field, because the timescale of charge collection is 100~1000 ns depending on the hole mobility for a 100nm thick active layer [18].Therefore, the charge dynamics are governed by the diffusive charge translocation and the intrinsic electric fields localized to the polymer-fullerene interfaces.As the result, the spectral dynamics were similar under the conditions of short circuit (Fig. 1(a)) and open circuit (data not shown).However, in Fig. 2(b) the forward bias, comparable to the V OC , provides a macroscopic and static electric field across the photoactive layer, the spectral dynamics also show little difference from those under short circuit (Fig. 2(a)).These results strongly suggest that, under working condition, the macroscopic electric field does not influence the subnanosecond charge dynamics of the BisDMO-PFDTBT/PC 61 BM solar cell.

Results and discussion
From Figs. 2(c) and 2(d), we see that the TA spectra under the reverse biases of V r = −2 V and −5 V resemble closely each other.These transients, however, differ distinctly from those in Figs.2(a) and 2(b): In Figs.2(c) and 2(d), the relative intensity of the transients at 0.13 ps and 1.00 ps are much lower, and those later than 1 ps decay much slower.In addition, the TA amplitudes at Δt = 2.0 ns are considerably higher than those at Δt = 100 ps.The differences between the spectral dynamics in Figs.2(c  Figure 3 shows the kinetics plotted at the respective characteristic wavelengths of 970 nm and 1050 nm preferentially probing PP and P + .It is seen that the kinetics under short circuit are close to those under V f = 0.79 V, whereas those under V r = −2 V and −5 V are similar to each other (Figs.3(a) or 3(b)).The kinetics of PP and P + can be described by the power law that is applicable to the temporal evolution of the charge species in disordered polymeric solids [27,28].We therefore simultaneously fitted the kinetics at two different probing wavelengths for each case of electrical bias to the model function where the positive and the negative exponents, respectively, represent the decay and the formation of charge species [27,28].In curve fitting, the kinetics data before 0.4 ps were not taken into account, because in this temporal regime the kinetics are complicated by the dynamics of exciton dissociation and geminate charge recombination [23].The exponents thus derived are listed in Table 1.

Table 1. Power-law exponents (α) obtained by global fitting of the 970 nm and the 1050
nm kinetics under individual bias voltages (cf Fig. 3).

Bias (V) Exponents
It is seen from Fig. 3 that the kinetics traces recorded under short circuit and under V f = 0.79 V are similar, as also reflected by the similar α 1 (0.20) or α 2 (0.06) between the two different bias conditions (Table 1).However, on going to the reverse biases of −2 V or −5 V, α 1 reduces from ~0.2 to ~0.1, suggesting significantly slowing down of PP depopulation, most probably owing to the field-assisted suppression of PP geminate recombination.In addition, under a reverse bias the absolute value of α 1 is equal to α 2 , implying the decay-torise correlation between PP and P + , i. e. the conversion of PP into P + .Importantly, increase of the reverse bias voltage accelerates the PP-to-P + conversion, as indicated by the increased magnitudes of α 1 and α 2 from V r = −2 V to V r = −5 V (Table 1).The field effects are in accordance to those reported for solar cells based on other types of fluorene copolymers [10,29].
Because both polymer and fullerenes are agglomerated to different degrees [30,31], upon photoexcitation a range of initial e − -h + separations are expected under a given polymerfullerene LUMO level offset (−ΔE L ).In addition, the delocalization lengths of electron and hole, respectively, rely on the molecular structures of polymers and the topology/bulkiness of the acceptor molecules [15,32].According to Braun-Onsager's model for e − -h + escape probability [33], a larger initial e − -h + separation of the PP state leads to a higher dissociation probability.Considering a threshold e − -h + separation (r th ) at which the −2 V bias (4.0 × 10 5 V/cm) can be effective in driving the e − -h + dissociation, we obtained a r th of 9.1 nm (by assuming that the biased field completely counteracts the field at the middle of an e − -h + pair), meaning that for the BisDMO-PFDTBT/PC 61 BM device biased with V r = −2 V, a PP with initial separation larger than 9.1 nm can be converted into free charges by the macroscopic electric field.Under V r = −5 V, the r th reduced to 5.8 nm, and hence the effect of field-assisted dissociation of PP would be more prominent.
As seen from Fig. 3 and Table 1, the PP depopulation is evidently speeded up (α 1 , 0.07→0.11)upon varying V r from −2 V to −5 V, and the formation of P + is accelerated concomitantly (−α 2 , 0.06→0.10).These results once again prove that a reverse bias of −2 V is effective in promoting the PP-to-P + conversion.Here, we note that the internal field of a working device (1.6 × 10 5 V/cm), considerably lower than the field strength under the −2 V bias (4.0 × 10 5 V/cm), is consequently ineffective in the PP dynamics.Regarding the field effect on the dissociation of neutral excitons, either the applied or the intrinsic macroscopic fields may be too weak to be of any help.In this relation, for the ladder-type methylsubstituted poly(paraphenylene) (MeLPPP) film, the field strength for exciton dissociation was reported to be as high as 1.5 × 10 6 V/cm [34].However, recent studies suggest that the microscopic fields localized to the BHJ interfaces may play important roles in the interfacial charge photogeneration.E. g., the interfacial dipole can reduce the driving force of exciton dissociation and thereby improve V OC [35,36].In addition, enhanced charge separation in the organic photovoltaic films doped with the ferroelectric dipole additive poly(vinylidenefluoride-co-trifluoroethylene) (PVDF-TrFE) was attributed to the fieldassisted dissociation of excitons and PPs (field strength localized to the polymer-fullerene interface, 2.4 × 10 6 V/cm), although under working condition the macroscopic internal field showed little effect on the device performance [  Finally, to see whether the device performance can be enhanced with the assistance of electric fields, we measured the EQEs of the BisDMO-PFDTBT/PC 61 BM solar cell under various bias potentials, and the results are presented in Fig. 4. Evidently, the bias at V f = 0.8 V drastically suppressed the EQE with reference to the EQE of short circuit, which is due to the unfavorable charge transport and collection in the μs-ms regime, because in the subnanosecond timeframe the charge kinetics at 0 V resemble with those at V f = 0.8 V (Fig. 3, Table 1).In contrast, the EQE under V r = −2 V is enhanced dramatically, while only slight further improvement is seen for V r = −5 V. (We have determined the I-V characteristics up to −5 V, under which no field-induced damage and irreversible modification of the device were observed.)Most importantly, the substantial enhancement of the EQEs under reverse biases tightly correlates to the field-assisted dissociation of PPs found in the subnanosecond timescale (vide supra).Recent study on the energetics factors driving the PP separation at organic-inorganic semiconductor interfaces has revealed that the band-bending, i. e. the higher shifted binding energy of organic molecules at the heterojunction with respect to that in the organic bulk, facilitates the separation of the photogenerated interfacial PPs, and hence the regeneration of interfacial holes by transferring to the consecutive bulk organic molecules plays an important role in free charge production [38].Similar mechanism may also work in organic solar cells, which is especially important when the effects of macroscopic internal electric field is taken into account.

Conclusions
The present work unambiguously shows that, under working condition, the macroscopic internal electric field of the BisDMO-PFDTBT/PC 61 BM solar cell is ineffective in the subnanosecond charge photogeneration/recombination dynamics.However, a reverse bias strength exceeding −2 V can significantly accelerate the processes of PP-to-P •+ conversion, and concomitantly suppress the PP geminate recombination.Polymer solar cells generally work with low operating voltage (<1 V), meaning relatively low internal field strength.To improve the device performance, it is important to optimize the phase morphologies of the BHJ layers so as to increase the initial e − -h + separation, and to promote the field-assisted PP separation.

Fig. 1 .
Fig. 1.(a) UV-visible absorption spectra of the BisDMO-PFDTBT/PC 61 BM (1:3) photoactive layer in real solar cell device (thicker solid), the neat films of BisDMO-PFDTBT (thinner solid) and PC 61 BM (dashed).Insets are the molecular structures.Arrow points to the excitation wavelength for time-resolved measurements (610 nm).(b) Typical current-voltage (I-V) curve of the solar cells determined under AM 1.5 G illumination () and in dark ().
Figure1(a) shows the UV-visible absorption spectrum of the BisDMO-PFDTBT/PC 61 BM photoactive layer.The major absorption band at 554 nm (I) is attributed to the optical transition from the ground state to the intramolecular charge transfer state (S 1 ←S 0 ) shifting the electron density from the fluorene unit to the thienyl-benzothiadiazole (TBT) unit of BisDMO-PFDTBT, while the band at 389 nm (II) is ascribed to the absorptive transition to the excitonic state (S 2 ←S 0 ) with the π-electron delocalized over the fluorene-thiophene backbone[23,24].In time-resolved measurements, we applied the bandgap photoexcitation at 610 nm to minimize the photoexcitation of PC 61 BM and the excess energy of the S 1 exciton of BisDMO-PFDTBT, otherwise the former would introduce the diffusive dynamics of fullerene excitation (~100 ps)[25] and the latter would promote the PP dissociation despite a weak contribution[15,26]. Figure 1(b) shows a typical I-V curve of the BisDMO-PFDTBT/PC 61 BM device, indicating an open-circuit voltage (V OC ) of 0.79 V. Here, PC 61 BM instead of PC 70 BM was used, which resulted in a relatively low PCE of 1.47%.

Fig. 2 .
Fig. 2. Representative time-resolved spectra for the BisDMO-PFDTBT/PC 61 BM solar cell at indicated bias potentials and delay times.Excitation wavelength was 610 nm (~1.0 × 10 13 photons•cm −2 •pulse −1 ).The depleted sections around 1200 nm are due to the interference from the second order diffraction of the excitation pulses.

Fig. 3 .
Fig. 3. Time-evolution profiles probed at (a) 970 nm and (b) 1050 nm for the BisDMO-PFDTBT/PC 61 BM solar cell at indicated bias potentials (cf Fig. 2. Normalized to the amplitudes maxima at 0.13 ps).Solid lines were obtained by global fitting of the kinetics to the model function