Indium tin oxide (ITO) glass (surface resistivity: 30–60 Ω/sq.), P3HT and PC₇₁BM were bought from Sigma-Aldrich. The ITO glass was cut, then cleaned and dried using the procedure described elsewhere (Otieno et al., 2016). PEDOT:PSS (1.3 wt% dispersion in H₂O), PBDB-T-SF (PCE13) and ITIC-Th, were sourced from Ossila Limited. The donor:acceptor blends were mixed in a ratio of 1:1 to form a 20 mg/ml solution and then stirred overnight to ensure a homogeneous blend solution. The sequential device fabrication was carried out foremost by spin coating PEDOT:PSS onto etched ITO glass substrates. The PEDOT:PSS layers were further annealed at 100°C for 10 min to remove water residue before spin coating the blend mixture on the PEDOT:PSS/ITO layers. The photoactive layer comprising of the polymer blend was spin coated at 2,000 rpm for 1 min. The donor:acceptor/PEDOT:PSS/ITO layers on glass were metallized by thermal evaporation of Al to form the ITO/PEDOT:PSS/binary blend/Al device architecture. The resulting OPV device was annealed over a hot plate for 10 min at 120°C to increase the interconnectivity of the blend network (Otieno et al., 2017).

A Cary 500 UV–Vis-NIR Spectrophotometer was used to measure the optical absorption of the thin films. Both Photoluminescence and Raman spectra were measured using a Horiba LabRAM HR Raman spectrometer with 600 lines/mm grating using a 514.5 nm excitation wavelength from an argon-ion laser. A laser power of 4 μW at the sample was used to prevent laser damage to the thin film. Current density–voltage (J–V) characteristics were measured using a solar simulator (fitted with 150 W Xe lamp) with 1.5 air mass filters and a (HP 4141B DC) source/measure unit under 100 mW/cm² illumination. It is crucial to note that all devices were prepared and characterized in air prompting the slight lower J–V performances witnessed. The thicknesses of the PEDOT:PSS layer, the active layer of various blends and the Al electrode were measured to be about 50, 215, and 300 nm with negligible variations, respectively using a surface profilometer. Over 10 devices were fabricated and characterized for each active layer combination and a standard deviation calculated.

The absorption spectrum of each individual layer prior to blend formation is presented in Supplementary Figure 1. It is seen in the figure that PBDB-T has strong absorption in the range spanning from 500 to 750 nm. Additionally, a less intense absorption peak is evident in the ultraviolet range around 350 nm. The extent of the absorption window within the visible range is much broader for this donor material compared to that of P3HT (550 < λ < 650 nm) thus highlighting the superior spectral broadening characteristics of PBDB-T. This comparison was extended to the two contrasting individual acceptor layers. Supplementary Figure 1 shows that the non-fullerene derivative ITIC-Th has a very large extinction coefficient (k) and an extended absorption characterized by two peaks at ∼575 and 640 nm compared to the fullerene derivative counterpart PC₇₁BM which peaks at ∼474 nm. Based on these results, thin films of the polymer blends of comparable film thickness were subsequently analyzed. The active layer thickness for all blends was approximately 215 nm.

It is evident that upon blend formation (see Figure 2), non-fullerene ITIC-Th based blends show broader and stronger absorption compared to the fullerene, PC₇₁BM based blends. Among the ITIC-Th based blends, the PBDB-T:ITIC-Th blend shows a higher absorption intensity although with a narrower wavelength range compared to P3HT:ITIC-Th. This implies that absorption beyond the bluish-green portion of the EM spectrum is reduced for the PBDB-T based blend layer. However, there is reasonable agreement that all the ITIC-Th based blends have a better light harvesting capability compared to PC₇₁BM based counterparts for the same film thicknesses.

Supplementary Figure 2 shows the steady-state photoluminescence measurements of pristine PBDB-T, P3HT, ITIC-Th, and PC₇₁BM thin films spin coated as single layers on glass substrates. These measurements have been carried using the 514.5 nm laser line as the excitation wavelength. All the films emit in the visible and near infrared range. The normalized PL spectrum (blue line) of Supplementary Figures 3, 4 show a stronger overlap of the emission of PBDB-T and P3HT, respectively, with the absorption of non-fullerene ITIC-Th compared to the fullerene counterparts. This overlap is likely to assist in the energy transfer from the donor to the ITIC-Th via Forster resonance energy transfer (FRET) thus leading to localized absorption. The quenching of the PL emission upon blend formation suggests possible energy transfer between the donor and acceptor material. Figure 3 shows higher PL quenching from the fullerene acceptor-based blends as compared to their non-fullerene counterpart and this is an indication of a more favorable change transfer mechanism than the energy transfer mechanism. The efficiency of the charge transfer is usually dependent on the mean length between conjugated polymer donor and fullerene/NFA molecules. The high PL emission (low PL quenching) experienced with the NFA blends indicates that the donor - non-fullerene interface may exceed the exciton diffusion length (~10 nm). Hence it is also plausible that some excitons cannot rapidly dissociate as the nearest NFA molecule is much further than the exciton diffusion length. The consequence of this is a radiative recombination that gives rise to the intense photoluminescence signal (Motaung et al., 2009) seen in Figure 3. Since the radiative recombination leads to the loss of charge carriers and thus to a reduction in the photocurrent, re-absorption of the radiated photon by the donor cannot be ruled out. Comparing the donor materials for the same acceptor, PBDB-T: PC71BM show better PL quenching, an indication of better charge transfer compared to P3HT: PC71BM. This is in agreement with the difference in J–V performance in section “J–V Characterization Under Illumination.” On the contrary, P3HT:ITIC-Th showed higher PL quenching compared to PBDB-T:ITIC-Th despite poor device performance in section “J–V Characterization Under Illumination.” This high unusual PL quenching in P3HT:ITIC-Th compared to PBDB-T:ITIC-Th could be attributed to either additive photo-absorption in the acceptor and donor leading to the generation of more electrons besides that of the donor or due to photoexcitation of the donor by the emission from the ITIC-Th.

When organic molecules experience electronic transitions induced by light, new forces in the excited state within their nuclear framework occurs such that atoms are displaced from their ground state equilibrium positions. Such motions can be ascribed to Raman-active vibrations (i.e., totally symmetric vibrations) hence ensuring a metastable configuration of the Franck-Condon geometry (Grey, 2019).

Usually, electronic transitions of conjugated polymer materials are mostly π-π^∗ in nature. This means higher frequency skeletal C–C stretching modes in the range of 1,000–1,600 cm^–1 as depicted in Figure 4. This figure shows the Raman spectra of different blends while the in-set is a magnified view of the spectra between 1,350 and 1,550 cm^–1. From Supplementary Figure 3, PBDB-T shows more polymer crystallization from the sharp Raman peaks compared to P3HT donor material ensuring enhanced regioregular domains (Kim et al., 2011; Roigé et al., 2012). Such increased degree of crystallinity is known to induce a larger interchainmolecular interaction leading to more delocalized conjugated π electrons, consequently producing a low band gap and enhanced optical π–π^∗ transition resulting in a more effective photon absorption (Motaung et al., 2009) as depicted in Figure 2. Pristine PBDB-T used as a donor shows Raman modes at 1,427, 1,452, 1,487, and 1,540 cm^–1, which are assigned to the C_α=C_β symmetric stretching mode, antisymmetric stretch mode of C_α=C_β, skeletal stretching mode of C_β–C_β and antisymmetric C_α–S–C_α ring skeleton that depict distortion of the in-plane vibrations of the thiophene ring of PBDB-T (Gao et al., 2018). Other less intense peaks are at 1,075, 1,231, and 1,598 cm^–1 and may be assigned to the stretching mode of C_β–C_alkyl, C–H bending mode, and antisymmetric stretch mode of C_α=C_β, respectively (Kabongo et al., 2016). Upon blending with acceptor materials, there is no noticeable shift in peak position regardless of whether ITIC-Th (non-fullerene) nor PC₇₁BM (fullerene) is used. This indicates that the two acceptor materials are being loaded onto the PBDB-T polymer backbone without much alteration of the conjugation length of the polymer (Falke et al., 2011). However, there is a general increase in intensity upon blending with PBDB-T, an indication of a preresonant Raman effect as a result of some absorption around the 514 nm excitation wavelength within the blend films (Tsoi et al., 2011).

The P3HT donor shows three main peaks at 1,090, 1,379, and 1,450 cm^–1 that can be assigned to the C–H bending mode, the C_β–C_β skeletal stretch mode, and the C_α=C_β symmetric stretch mode, respectively (Falke et al., 2011), as shown in Supplementary Figure 3. The most prominent band of the pristine P3HT is at 1,450 cm^–1 with a FWHM of 27.63 cm^–1. Upon blending with ITIC-Th, the only significant change was a FWHM increased to 37.03 cm^–1. Blending P3HT with PC₇₁BM acceptor material, the prominent peak shifted to 1,458 cm^–1 while FWHM increased further to 40.03 cm^–1. This difference between the polymer/fullerene blend signals is shown in the inset (Figure 4). The shift to a higher wavenumber indicates a further shrinking of the conjugation length along the polymer backbone and increased FWHM indicates a crystallinity decrease of the P3HT polymer. This shift is more pronounced for the P3HT:PC₇₁BM blend compared to the P3HT:ITIC-Th blend and this corroborates the difference in device performance experienced. On the other hand, P3HT:ITIC-Th shows a more intense peak at 1,452 cm^–1 compared to P3HT:PC₇₁BM, an indication of a more ordered structure leading to an increase of the Raman cross-section (Roigé et al., 2012). Such enhanced intensity characteristics are induced by morphological changes associated with polymer crystallization transitions leading to large phase separation that is likely to affect charge transfer as observed in the J-V characteristics.

For a J–V graph measured in the absence of light, the dark current can be categorized into three sections corresponding to various types of charge transport within the device depicted by the semi-logarithmic representation in Figure 5.

In region (i), located mainly below the open circuit voltage of the OPVs (0–0.5 V), the current is mainly being limited by shunt resistance Rsh. The charge carrier injection emanating from the electrodes into the polymer material is usually minimal as a result of the low bias voltage unable to compensate for the internally built-in field (Gadisa, 2006). In this region, P3HT:ITIC-Th shows the highest current which is attributed to the presence of shunt paths for the charge carriers resulting from the poor organization of the two polymers in the blend which lacks phase separation. PBDB-T:PC₇₁BM blends on the other hand show the lowest dark current density in region (i) prompting decreased reverse dark saturation current density leading to enhanced V_OC of the devices as defined by: Voc=nkTq(ln⁡JphJs+1) (He et al., 2010), where n is the ideality factor, k is the Boltzmann constant, q is the elementary charge, T is the temperature in kelvin, J_ph is the photocurrent and J_S is the reverse dark saturation current density. In region (ii), the current is controlled by the diode characteristics; bias saturation current density J₀ and ideality factor n, usually factors attributed to the metal-active layer interface of the device (Servaites et al., 2011). The four devices show almost the same slope in this region and thus similar diode characteristics. Generally, the same donor materials exhibit similar dark current behavior at 1.0 V [region (iii)], an indication of similar carrier mobilities in the blends. Region (iii) is associated with charge transport in the bulk/blend material and corresponds to space charge limited region. Organic polymer blends have low charge carrier mobility due to the presence of traps, impurities and series resistance (Rs) limited currents, thus the transfer rate of injected charge carriers is highly diminished resulting in charge accumulation in most active organic materials (Martens et al., 2000). A steep slope in this region similar to the PBDB-T:PC₇₁BM blend imply high Rs in the device that aid in better charge collection at the electrodes.

Figure 6 shows the current density–voltage (J–V) characteristics of various OPV devices measured under AM 1.5 G illumination with 100 mW/cm² light intensity. The inset shows the energy level orbitals of the polymers while Table 1 gives the extracted device characteristics.

Based on their favorable energy-level alignment as depicted in the inset of Figure 6, all these donor/acceptor combinations demonstrate a high possibility of achieving efficient charge separation. P3HT:ITIC-Th recorded the lowest performance with fill factor (FF) of 34.2%, J_sc of 2.29 mA/cm² and V_oc of 0.51 V resulting in a PCE of 0.61%. Such low Voc is as a result of the small HOMO-LUMO difference of the donor-acceptor compared to other blends since the Voc of OPVs is directly proportional to the difference in energies of the LUMO and HOMO of the acceptor and the donor, respectively (Qin et al., 2016). Generally, P3HT has an ionization potential of ∼–5.16 eV hence devices in which it is used as a donor material experienced lower Voc compared to their PBDB-T counterparts with ionization potential of ∼–5.33 eV. When the donor material was PBDB-T, a blend of PBDB-T:PC₇₁BM gave a greater V_oc of 0.82 V in comparison to PBDB-T:ITIC-Th that gave 0.71 V. This could be attributed to a higher-lying LUMO level (–3.85 eV) compared to that of ITIC-Th (-3.98 eV).

PBDB-T:PC₇₁BM recorded the highest short circuit current density of 10.22 mA/cm² with a fill factor of 53.5% yielding best conversion efficiency of 4.5%. This high current could be attributed to a better charge transfer as witnessed from the high PL quenching discussed in section “Raman Spectroscopy” despite the low absorption intensity compared to other blends. P3HT: PC₇₁BM was second best in terms of current density among the four blends. This finding is also attributed to better charge transfer at the donor-acceptor boundaries.

We further carried out studies on charge generation and separation. Figure 7 shows a plot of photocurrent density J_ph against the effective voltage V_eff where J_ph is defined as J_ph = J_L – J_d. J_L and J_d stand for the current densities under illumination and in the dark, respectively. The effective voltage V_eff was defined as V₀ - V, where V₀ is the voltage at which J_L = J_d.

From Figure 7, the photocurrent is linearly proportional to the voltage at low effective voltages (V_eff < ∼0.1) and thereafter saturates at higher V_eff (>1) where minimal charge recombination is expected as a result of the high internal electric field within the device (Xiao et al., 2018). From this graph, the saturation photocurrent density (J_sat), which is usually only limited by the incident photon flux absorbed by the active layer and is independent of the temperature and bias voltage (Chen et al., 2015), can be calculated and the results are shown in Table 2. PBDB-T:PC₇₁BM experiences a high saturation current (11.08 mA/cm²), an indication of better charge carrier transmission and collection compared to P3HT:ITIC-Th (3.32 mA/cm²). From the derived saturation current, we can derive the maximum exciton generation rate (G_max) which is an indication of the maximum number of absorbed photons, and is given by J_sat = qL G_max, where q is the basic charge and L is the photoactive layer thickness, about 215 nm (Lu et al., 2015). PBDB-T:PC₇₁BM showed the highest maximum exciton generation rate (G_max) of 3.22 × 10²⁸ s^–1 followed by P3HT: PC₇₁BM (3.05 × 10²⁸ s^–1), an indication that binary blends of fullerene derivative PC₇₁BM have high light absorption and harvesting contrary to the UV-vis spectra (with low absorption intensity but broader range) leading to the high J_sc for the devices. This illustrates that absorption within a wider range of the solar spectrum is quite crucial in ensuring more light harvesting by the solar cell devices.

Charge dissociation and collection probability P(E,T) of different devices was thereafter analyzed using a plot of the J_ph/J_sat ratio under short-circuit conditions against the effective voltages (Qin et al., 2016) as shown in Figure 8. The P3HT: PC₇₁BM blend showed the highest value of P(E,T) of 90.62% which is correlated with the highest fill factor of 54.2% recorded in section “Raman Spectroscopy” while P3HT:ITIC-Th gave the lowest value leading to lowest PCE. This indicates that when PC₇₁BM is used as an acceptor material, lower geminate recombination at the donor–acceptor interface is experienced as indicated by the higher charge dissociation probability compared to ITIC-Th counterparts resulting in more efficient charge generation processes hence higher current density.

As shown in Figure 9, all four devices fabricated and tested showed a linear dependence of current density on light intensity in logarithmic coordinates, but with slightly different slopes. The current density J_sc is known to depend on the light intensity (P_light) through a power-law expressed as Jsc ∝ (P_light)^α, for organic photovoltaics devices (Riedel et al., 2004), where α is the gradient of the log-log plot of Jsc against light intensity.

Usually an α value close to one gives an indication of a weaker bimolecular recombination under short circuit conditions. P3HT:PC₇₁BM gave a slope of 0.978, close to the value of 0.971 of PBDB-T:PC₇₁BM. These high values close to 1 indicate that PC₇₁BM based devices are much more capable of efficient sweeping out of charge carriers with negligible bimolecular recombination in comparison to their ITIC-Th counterparts. PBDB-T:ITIC-Th showed an α value of 0.797 while P3HT:ITIC-Th gave a value of 0.695. This indicates that the ITIC-Th based devices have increased bimolecular recombination that competes with charge extraction resulting in reduced current density recorded.

Figure 10 shows an investigation into the dominant recombination from Voc dependence on the natural logarithm of varied intensity of light governed by the relationship Voc ∝ (nk_BT/q)log(P_light), where n represents the ideality factor, Boltzmann’s constant is shown as k_B, T stand for temperature in Kelvin, and elementary charge is represented as q. The gradient of V_oc against ln(P_light) is shown to estimate the extent of trap-assisted recombination within the OPVs. P3HT:ITIC-Th reported the largest slope of 1.459 k_BT/q, an indication that the device experiences stronger trap-assisted or Shockley–Read–Hall recombination compared to other devices. On the other hand P3HT:PC₇₁BM showed the smallest slope of 0.975 k_BT/q indicating a dominance of bimolecular recombination mechanisms with fewer traps.

The FTIR spectral characteristics of the pristine films and blends deposited on a glass substrate are shown in Figure 11. At wavenumber positions in the range of 650–1585 cm^–1, two clustered peaks can be observed which are attributed to the occurrence of C–H band symmetric and asymmetric vibrations (Arulraj et al., 2018).

From the spectra above, the pristine films showed similar peak positions with no significant shift upon blend formation. The peaks around 600–800 cm^–1 are an indication of C–H bending vibrations out of the plane in which the thiophene ring vibrates while that at 903 cm^–1 could be ascribed to C–H stretching vibrations. The peaks observed at 1,000–1,200 cm^–1 are attributed to the COO- asymmetric and symmetric stretching vibrations of the surface acetate species (Suprunov and Ivanov, 1987), while those around 1,200–1,500 cm^–1 are ascribed to C=C symmetric vibrations formation and that of 1,583 cm^–1 is assigned to the C=C stretching mode (Nicho et al., 2004; Motaung et al., 2009). Upon blend formation, little or no shifts to either longer or shorter wavenumbers are seen; however, there is an increase in intensity of most vibrations, especially the aromatic-C–H out plane vibrations and aliphatic-C–H stretching. This suggests that blends differ in relative amount of twisting moiety in the crystalline state which appear highest in the P3HT:PC₇₁BM blend. This is in agreement with the highest value of P(E,T) of 90.62% reported and indicates better lamellae reorientation and high-mobility percolation paths formation with grain boundaries of low angle (Street et al., 2005). P3HT:ITIC-Th shows a great decrease in peak intensity when compared to its pristine components as shown in the inset in Figure 10. This explains the low performance of the device as it is prone to low-mobility percolation paths with wide-angle grain boundaries.

Figure 12 shows tapping mode AFM images of various thin film compositions being studied. The films of ITIC-Th based blends generally showed a smoother surface with an rms of 2.877 nm and 3.198 nm for P3HT:ITIC-Th and PBDB-T:ITIC-Th, respectively. On the other hand, when the fullerene acceptor was used to make PBDB-T:PC₇₁BM and P3HT:PC₇₁BM, increased rms of 4.146 and 3.301 nm, respectively were recorded (Figure 12). Since the AFM images show a topographical image of the height correlations of a 2D surface, it is rather difficult to correlate our rms roughness with the vertical scale morphology as the aspect ratio of the rms roughness to film thickness is very small. The low roughness observed for these blend layers are rather promising owing to the lack of hillock formation which degrades performance.