During the last decades, the power conversion efficiency (PCE) of organic bulk-hetorojunction (BHJ) solar cells based on donor/acceptor blends surpassed the 10% threshold, mainly due to the discovery of novel materials as well as device structure engineering (Liu et al., 2014; He et al., 2015; Holliday et al., 2016; Huang et al., 2016; Spyropoulos et al., 2016; Zhao et al., 2016). Polymers and/or small molecules, used as donor materials, in combination with fullerene derivatives, used as acceptor, are the common active components in BHJ devices (Zhang et al., 2014; Lu et al., 2015c; Min et al., 2015; Squeo et al., 2015). Due to the narrow absorption of the donor materials, one of the main challenges in order to further boost the PCE of organic solar cells is to achieve better absorption match to the solar irradiance spectrum. In this regard, two main concepts have been developed: tandem and ternary organic solar cells (Ameri et al., 2009, 2013a,b; Li et al., 2013; You et al., 2013; Spyropoulos et al., 2014; Lu et al., 2015b; Yang et al., 2015; Cheng et al., 2016; Goh et al., 2016; Keawsongsaeng et al., 2016; Lee et al., 2016; Nian et al., 2016). The former is based on a complex multi-layer stack with the main challenge of designing a robust solution-processed intermediate layer. The latter, made of two donors and one acceptor, mixed together in a unique solution, overcomes the complexities of the tandem device architecture, maintaining the easy processability of a single-junction organic BHJ solar cell. To date, polymers (Lu et al., 2014, 2015a; Gasparini et al., 2015b; Yang et al., 2015), small molecules (Zhang et al., 2015), dyes (Ke et al., 2016), quantum dots (Itskos et al., 2011), and fullerene derivatives (Cheng et al., 2014) have been adopted as “guest” in the polymer-fullerene “host” system. In addition to the need for donor materials with the complementary absorption, one of the key points to surpass the performance of binary cells in ternary devices is to find donor materials with compatible physical and chemical nature (Yang et al., 2015). This can prevent the formation of recombination centers or morphological traps, which deteriorate the photovoltaic properties.

Here, we report a ternary organic solar cell system processed in air that shows a pronounced sensitization effect, resulting in a PCE of more than 4.6%. As sensitizer, we incorporate the near infrared (NIR) polymer P1 containing alternating pyridyl[2,1,3]thiadiazole between two different donor fragments, dithienosilole and indacenodithienothiophene (IDTT), into a host system of indacenodithieno[3,2-b]thiophene,2,3-bis(3-(octyloxy)phenyl)quinoxaline (PIDTTQ) (Gasparini et al., 2015a) blended with [6,6]-phenyl C₇₀ butyric acid methyl ester (PC₇₁BM). Indeed, in order to have components with a similar chemical nature in the ternary blend system, we used two polymers with the same backbone IDTT unite for the host as well as the guest donors.

The polymer P1 was synthesized by Stille-type aromatic cross-coupling reaction of a stoichiometric balance ratio of the distannyl derivative of para-hexyl-phenyl substituted IDTT (M2) and 4,4′-(4,4-bis(2-ethylhexyl)-4H-silolo[3,2-b:4,5-b′]dithiophene-2,6-diyl)bis(7-bromo-[1,2,5]thiadiazolo[3,4-c]pyridine) (M1), in the presence of tris(dibenzylideneacetone)dipalladium(0) (Pd₂dba₃) and tri(o-tolyl)phosphine (P(o-tol)₃) as the catalytic system (Scheme 1). After soxhlet extraction the polymer was obtained from the o-dichlorobenzene fraction with a number average molecular weight Mn of 36,800 g mol⁻¹ and a polydispersity index of 3.3.

Figure S1 in Supplementary Material shows the absorption spectrum of P1 in DCB solution and as solid. The copolymer for both cases shows a single band in the high energy region, which is assigned to a localized π−π* transition and another absorption band in the low energy region (up to 1,000 nm), which is assigned to an intramolecular charge-transfer transition. The maximum of the NIR absorption band of P1 in the solid state is bathochromic shifted (738 nm) in comparison to the corresponding UV–VIS solution (695 nm). The optical band gap energy estimated from the absorption edge of film spectrum was estimated to be 1.87 eV. Based on the onsets of the oxidation and reduction peaks in cyclic voltammetry (CV) measurements, the electrochemical highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) energies were estimated to be −5.34 and −3.71 eV, respectively, corresponding to an electrochemical band gap energy of 1.63 eV (Gedefaw et al., 2016).

Next, we analyzed the device performances of the ternary devices. The device architecture used in this work is based on ITO/ZnO/active layer/MoOx/Ag. PIDTTQ [its lower molecular weight version of PIDTTQ-LMW (Gasparini et al., 2015a)] has been previously presented in the literature. All the solution-processed layers are doctor bladed in air. Figure 1A depicts the energy levels, measured with CV of the polymers and the fullerene derivate. Figure 1C shows the current density–voltage characteristics of the binary PIDTTQ:PC₇₁BM (1:2 wt/wt) as well as ternary PIDTTQ:P1:PC₇₁BM (different composition) under 1 sun illumination (100 mW cm⁻²). In agreement with previous reports, binary cells delivered a PCE of 4.3% with an open circuit voltage (V_oc) of 0.84 V, a short circuit current (J_sc) of 8.62 mA cm⁻², and a fill factor (FF) of 60%. Adding 15 wt% of the NIR sensitizer delivers the highest short-circuit current, reaching 10.60 mA cm⁻², V_oc of 0.84 V, and FF of 52%, increases the overall efficiency of the ternary system 4.6% under 1 sun conditions. As shown in Table 1, J_sc increased monotonically by increasing the amount of P1, due to the better harvesting of the ternary system in the NIR region, in the best case, an improvement in J_sc of ~20% is achieved in the ternary system PIDTTQ:P1:PC₇₁BM (0.85:0.15:2). Notably, the V_oc obtained in the ternary cells is identical to the binary PIDTTQ:PC₇₁BM, reflecting an energy cascade between the HOMO and LUMO energy levels of the three components (Figure 1A). Unfortunately, we observed a continuously decreased in FF by introducing higher amount of P1, which indeed inhibits the higher improvements of the ternary device performance compared to its reference.

We further measured external quantum efficiency (EQE) spectra of OPV devices made from PIDTTQ:P1:PC₇₁BM and PIDTTQ:PC₇₁BM (Figure 1D). Photoaction spectra of active layers with increasing P1 content show improved photoresponse particularly around 800 nm, i.e., the enhancement in J_sc originates dominantly from the NIR polymer absorption regime. We note that the integrated EQE for these devices matches the measured short circuit current within a margin of 5%.

In order to shine light into the mechanism, we performed photoluminescence (PL) measurements. PL is widely used in ternary BHJ solar cells to discriminate between energy and charge transfer between host and guest materials (Lu et al., 2014, 2015a; Gasparini et al., 2015b). In principle, if the charges are transfer from the wide to the low band gap material, the PL of the host should decrease while the PL of guest material should not increase. On the other hands, if the energy transfer is the main mechanism, a quenching of the host PL is associated with an increase of guest PL. Moreover, in order to have an energy transfer, the absorption of the guest polymer should overlap with the emission of the guest. The inset of Figure 1B confirms the aforementioned requirement. Thus, we mixed together PIDTTQ and P1 in different weight ratio. As depicted in Figure 1B, the PL of PIDTTQ is quenched of 53 and 79% upon introduction of P1 (85–15 and 50–50, respectively). In addition, we observed a clear enhancement of P1 PL compared to the pristine one, indicating an efficient energy transfer.

Before getting insight into the optical and electrical behavior of the binary and ternary devices, we performed intermittent contact mode atomic force microscopy (AFM, Figure 2). In agreement with our previous study (Gasparini et al., 2015a), the topography of PIDTTQ:PC₇₁BM layer shows spherical features with domains of ~100 nm. Interestingly, we observed similar morphology for ternary active layers. We calculated a root mean square roughness of 0.46, 0.51, and 0.60 nm for PIDTTQ:PC₇₁BM, PIDTTQ:P1:PC₇₁BM (0.85:0.15:2) and P1:PC₇₁BM blends, respectively. Thus, the low FF calculated for PIDTTQ:P1:PC₇₁BM cannot be ascertained to changes in the microstructure upon addition of the guest sensitizer.

In order to understand the lower FF obtained in the ternary BHJ solar cells, we first studied the charge transport properties by employing the technique of photoinduced charge carrier extraction by linearly increasing voltage (photo-CELIV) (Mozer et al., 2005; Clarke et al., 2015; Min et al., 2015). From the measured photocurrent transients, the charge carrier mobility (μ) is calculated using the following equation: (1)μ=2d23Atmax2[1+0.36Δjj(0)]ifΔj≤j(0), where d is the active layer thickness, A is the voltage rise speed A = dU/dt, U is the applied voltage, t_max is the time corresponding to the maximum of Δj of the extraction peak, and j(0) is the displacement current.

Figure 3A shows the transient recorded by applying a 2 V/60 μs linearly increasing reverse bias and a delay time (td) of 10 µs. Analysis of the photo-CELIV traces extracts charge carrier mobility values 1.13 × 10⁻⁴ cm² V⁻¹ s⁻¹, 8.54 × 10⁻⁵ cm² V⁻¹ s⁻¹, 7.53 × 10⁻⁵ cm² V⁻¹ s⁻¹, and 7.42 × 10⁻⁵ cm² V⁻¹ s⁻¹ for PIDTTQ:PC₇₁BM (1:2)-, PIDTTQ:P1:PC₇₁BM (0.90:0.10:2)-, PIDTTQ:P1:PC₇₁BM (0.85:0.15:2)-, PIDTTQ:P1:PC₇₁BM (0.80:0.20:2)-based devices, respectively. We also calculated the charge mobility of P1:PC71BM (Figure S5 in Supplementary Material), and we found a value of 5.02 × 10⁻⁵ cm² V⁻¹ s⁻¹. The lower FFs obtained in the ternary cells are in agreement with the lower charge carrier mobility calculated with photo-CELIV technique. We then analyze the lifetime of charge carriers by employing transient photovoltage technique (TPV) (Shuttle et al., 2008). The samples were connected to the terminal of an oscilloscope with the input impedance of 1 MΩ and illuminated with a continuous background laser to control the V_oc. A small optical perturbation was applied using a blue laser (λ = 405 nm). The pulse intensity was controlled to keep the height of the photovoltage transient smaller than 10 mV resulting in a voltage transient with amplitude ΔV ≪ V_oc. The measured transient decays show the form of single exponentials, as expected for the pseudo-first-order kinetic (Hamilton et al., 2010; Heumueller et al., 2015). (2)dΔVdt∝dΔndt=−keff=−ΔnτΔn, where V is the photovoltage, t is the time, Δn is change in the density of photogenerated carriers due to the perturbation pulse, k_eff is the pseudo-first order rate constant, and τ_Δn is the carrier lifetime. The Figure 3B depicts normalized photovoltage decays as a function of time for the binary and ternary devices at 1 sun condition by exciting with a blue laser (λ = 405 nm). As reported in Table 2, the lifetime of charge carriers is similar for PIDTTQ:PC₇₁BM (1:2), PIDTTQ:P1:PC₇₁BM (0.90:0.10:2), PIDTTQ:P1:PC₇₁BM (0.85:0.15:2), 6.72, 7.35, and 7.23 µs, respectively, suggesting that these ternary blends are not limited by the short lifetime of charge carriers. Otherwise, a reduce τ is observed for the PIDTTQ:P1:PC₇₁BM (0.80:0.20:2) based ternary system (4.72 µs). Thus, with the combination of photo-CELIV and TPV techniques, we were able to calculate the mobility-lifetime product (μτ). As collected in Table 2, μτ decreases from 7.59 × 10⁻¹⁰ cm² V⁻¹ to 6.72 × 10⁻¹⁰ cm² V⁻¹, 5.44 × 10⁻¹⁰ cm² V⁻¹, and 3.50 × 10⁻¹⁰ cm² V⁻¹ for PIDTTQ:PC₇₁BM (1:2), PIDTTQ:P1:PC₇₁BM (0.90:0.10:2), PIDTTQ:P1:PC₇₁BM (0.85:0.15:2), PIDTTQ:P1:PC₇₁BM (0.80:0.20:2)-based devices, reason of the poorer transport properties in the ternary blends compared to the binary BHJ devices.

Understood the limitation in the transport properties, we then focus on the better ability of the ternary system in the photogeneration by employing charge extraction (CE) and photoinduced absorption (PIA) spectroscopy (Salvador et al., 2012; Löslein et al., 2013; Gasparini et al., 2015b). In CE measurements, the samples were connected to the terminal of an oscilloscope with the input impedance of 1 MΩ and illuminated with a continuous background laser to keep it in the V_oc condition (Heumueller et al., 2015). In order to study the transient decay, we used a nanosecond switch that shifts the cell from open circuit to short circuit condition, allowing the calculation of charge density (n) by integrating the curves, respectively (Figure 4A) (Heumueller et al., 2015; Gasparini et al., 2016). In agreement with the J_sc values obtained, we calculate n as 2.97 × 10¹⁶ cm⁻³, 3.02 × 10¹⁶ cm⁻³, 3.66 × 10¹⁶ cm⁻³, and 1.15 × 10¹⁶ cm⁻³ for PIDTTQ:PC₇₁BM (1:2), PIDTTQ:P1:PC₇₁BM (0.90:0.10:2), PIDTTQ:P1:PC₇₁BM (0.85:0.15:2), PIDTTQ:P1:PC₇₁BM (0.80:0.20:2) based devices, respectively (Table 2). The ability of charge generation in the BHJ solar cells is also studied with PIA spectroscopy. We employed steady-state and frequency-dependent PIA spectroscopy at a pump energy of 2.33 eV to gain further insight into the underlying photophysical steps of the sensitization process. Figure 4B depicts the PIA spectra of binary and ternary devices measured at 60 mW cm⁻² pump intensity at 10 K. All spectra show a pronounced transmission minimum (bleach) around 1.81 eV and a PIA feature around 1.24 eV. Contrary, the ternary blends show a novel bleaching feature around 1.58 eV. We associate the two transmission maxima at 1.81 and 1.58 eV to the photobleaching of the electronic ground states of PIDTTQ and P1, respectively. As shown in Figure 4B, higher polaron signal is observed for the ternary PIDTTQ:P1:PC₇₁BM (0.90:0.10:2), PIDTTQ:P1:PC₇₁BM (0.85:0.15:2)-based devices compared with the binary cells, confirming the higher photogeneration ability by adding 10–15% of P1 into the host PIDTTQ:PC₇₁BM system.

In conclusion, we reported a novel ternary system with a clear contribution in the incident photon-to-current efficiency in the NIR region. A J_sc improvement of around 20% was obtained for PIDTTQ:P1:PC₇₁BM (0.85:0.15:2) ternary devices compared to PIDTTQ:PC₇₁BM binary cells. However, the low FF limited the performances of the ternary BHJ solar cells. We studied the transport mechanism of the organic solar cells by employing photo-CELIV and TPV techniques. We found that by adding P1 into the host system of PIDTTQ:PC₇₁BM the μτ product is reduced, explaining the lower FFs. Despite the poorer transport properties, the complementary results of CE and PIA spectroscopy showed an improved charge generation in PIDTTQ:P1:PC₇₁BM (0.85:0.15:2) ternary solar cells, leading to a PCE of more than 4.6%.

Materials and instruments: all reagents and starting materials were received from commercial suppliers and used without further purification. Anhydrous toluene was purchased from Sigma-Aldrich. Monomers M1 (Gasparini et al., 2015a) and M2 (Welch et al., 2013) were prepared according to literature procedures. The reactions were carried out under argon with standard and Schlenk techniques. Gel permeation chromatography (GPC) measurement was carried out at 135°C on a Waters Alliance 2000 GPC System equipped with PL-Guard and PL-mixed-B columns using trichlorobenzene as solvent. The UV spectrum was recorded on a Jasco V-670 spectrometer at room temperature. The HOMO energy level was determined by atmospheric pressure photoelectron spectroscopy using a photoelectron spectrometer model AC-2. The optical band gap was calculated using the formulae Eg = h(c/λ) + 0.3 eV. Cyclic voltammetry was executed in chloroform with 0.1 M (n-Bu)4NClO₄ against a standard calomel electrode. The electrochemical HOMO and LUMO energy levels were calculated using the formulae HOMO = −(Eox + 4.7) eV and LUMO = −(Ered + 4.7) eV, respectively.

NG, TA, and CB conceived and developed the ideas. NG designed the experiments and performed device fabrication, electrical characterization, and data analysis. NG performed photo-CELIV, TPV, CE, and PIA measurements. AG-R synthesized the polymer under the supervision of AK, AA, GP, VG, CC, SA, US, SB, and AP-C performed CV measurements. NG and MP performed PL measurements and contributed to revision the manuscript. The projects were supervised by TA and CB.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.