Intrinsic Advantage of Fused‐Ring Nonfullerene Acceptor‐Based Organic Solar Cells to Reduce Voltage Loss

The large voltage loss is a significant disadvantage of organic solar cells (OSCs) compared with inorganic and perovskite solar cells and should be reduced to further improve the power conversion efficiency of OSCs. Herein, the voltage loss in OSCs with a systematically controlled energy difference between the excited and charge transfer states is discussed. The voltage loss is compared between two types of OSCs: narrow‐bandgap donors paired with wide‐bandgap acceptors and narrow‐bandgap fused‐ring nonfullerene acceptors (NFAs) paired with wide‐bandgap donors, to elucidate whether there are any essential advantages in the latter. The first advantage of narrow‐bandgap‐NFA‐based OSCs is their higher photoluminescence quantum yield, owing to their rigid fused‐ring architecture. The second advantage is the ability to achieve efficient charge separation with a small voltage loss.


Introduction
The power conversion efficiency (PCE) of organic solar cells (OSCs) has increased rapidly in recent years, and the PCE of state-of-the-art OSCs now exceeds 19%. [1]A remaining challenge to be overcome to further improve the device performance is to reduce the voltage loss ΔV, which is defined as the difference between the optical bandgap E g and open-circuit voltage V OC (ΔV = E g /qÀV OC , where q is the elementary charge). [2]Recent improvements in the PCE can be mainly attributable to the suppression of ΔV, which was typically more than 0.8 V a decade ago, [3] whereas it now approaches %0.5 V. [1d,k] However, ΔV of the state-of-the-art OSCs remains much larger than those of their inorganic and perovskite counterparts, in which ΔV of less than 0.4 V can be achieved. [4]Therefore, further reducing ΔV as well as understanding the underlying physics regarding ΔV are the top priorities for improving the PCE of OSCs to the same level as those of inorganic and perovskite systems.
Compared with their inorganic and perovskite counterparts, the most significant obstacle to reduce ΔV of OSCs is their significantly larger nonradiative voltage loss ΔV nr .A decade ago, ΔV nr was typically %0.4 V, particularly for conventional fullerene-based OSCs. [5]In contrast, the emergence of novel nonfullerene acceptors (NFAs), such as a perylene diimide (PDI) dimer SF-PDI 2 , remarkably reduced ΔV nr to less than 0.3 V, although the PCEs of SF-PDI 2 -based OSCs were moderate due to the transparency in the near-IR region. [6]Shortly thereafter, the development of various narrow-bandgap fused-ring NFAs, such as Y6 (the chemical structure can be found in Figure 1), successfully achieved a relatively low ΔV nr of less than 0.3 V while maintaining a high (>80%) photovoltaic external quantum efficiency (EQE PV ) in the visible to near-IR region when paired with a wide-bandgap donor polymer, such as PM6.This resulted in the recent success of improving the PCE. [7]Therefore, utilizing narrow-bandgap fused-ring NFAs paired with wide-bandgap donor polymers is the current mainstream of the OSC community, and hence, recent material development is based thereon.However, the essential advantage of narrow-bandgap-NFA-based OSCs over fullerene-based or wide-bandgap-NFA-based OSCs remains unclear.In other words, whether there is an intrinsic advantage for narrow-bandgap-NFA-based OSCs in terms of ΔV and ΔV nr is not elucidated.Verifying this is important for future material development, which is the main aim of this study.
Here, we studied whether there is an essential advantage in ΔV and ΔV nr for narrow-bandgap-NFA-based OSCs (hereafter referred to as h-type OSCs, where "h" stands for the hole transfer) compared with fullerene-based or wide-bandgap-NFA-based OSCs (hereafter referred to as e-type OSCs, where "e" stands for the electron transfer).There is a clear difference in the charge generation process between the e-type and h-type OSCs.2a] In e-type OSCs, an electron in the lowest unoccupied molecular orbital (LUMO) of the donor is transferred into the LUMO of the acceptor.In contrast, in h-type OSCs, a hole in the highest occupied molecular orbital (HOMO) of the acceptor is transferred into the HOMO of the donor.In contrast, ΔV and ΔV nr are determined by the charge recombination dynamics in OSCs.Because the charge recombination process does not depend on which (electron transfer or hole transfer) process the charge carriers were generated through, the difference between the e-type and h-type OSCs in terms of ΔV and ΔV nr is not evident.In this study, we prepared various e-type and h-type OSCs and quantified their voltage losses.We did not observe any essential differences in the voltage loss trend.Rather, the recent success of narrowbandgap fused-ring NFAs can be attributed to their ability to generate charge carriers with a small energy offset between the local excited (LE) and CT states.

Results and Discussion
We prepared 17 devices (nine e-type and eight h-type devices, see Table S1, Supporting Information) using four donor polymers (Figure 1a-c), three fullerenes (Figure 1d-f ), and nine NFAs (Figure 1g-n).The steady-state absorption spectra and cyclic voltammograms of these materials can be found in Figure S1, S2, and Table S2, S3, Supporting Information.For e-type devices, a narrow-bandgap polymer PTB7-Th was used as a common donor material and blended with three fullerenes and six widerbandgap NFAs with different LUMO energy offsets.This means that all the e-type devices exhibited similar E g .Nevertheless, V OC of these devices depended significantly on the choice of acceptor material (V OC decreased with increasing LUMO energy offset), and V OC varied up to 0.38 V, as reported previously (current density-voltage ( J-V ) characteristics can be found in Figure S3 and Table S4, Supporting Information). [8]In contrast, for h-type devices, Y-series acceptors were used as narrow-bandgap fused-ring NFAs and blended with four donor polymers with different HOMO energy offsets.V OC varied up to 0.53 V depending on the choice of donor polymers despite comparable E g among all h-type devices, as was observed in the e-type devices.
Figure 2 shows the EQE PV spectra for these devices (the linear-scale plot can be found in Figure S4, Supporting Information).The experimental data for the e-type devices were obtained from our previous report. [8]Weak shoulders were observed in the EQE PV spectra at lower energies for some devices (plotted in cool colors), which can be attributable to the photocurrent response from the CT absorption. [9]In contrast, no shoulders were observed for the other devices (plotted in warm colors) because the CT absorption was buried under the smeared-out absorption edge of the narrow-bandgap materials owing to the small difference between E g and the CT state energy E CT .
To obtain a deeper understanding of the origin of the large variation in the voltage loss, V OC was divided into several components according to earlier studies, as follows [5a,8,10] where ΔV SQ is the voltage loss due to unavoidable radiative charge recombination (SQ stands for the Shockley-Queisser limit). [11]In contrast, the latter terms (ΔV SC , ΔV r , and ΔV nr ) are additional voltage losses due to the nonideal behavior of real devices.The details of the physical meaning and evaluation procedures are provided in the Supporting Information.V OC rad is the radiative limit of V OC , wherein charge recombination must occur with photon emission.Table 1 summarizes the voltage losses of these devices.
Because ΔV SQ depends only on E g and temperature, all the e-type devices in which E g of PTB7-Th was lower than those of the acceptors, exhibited the same ΔV SQ values.The same was true for the h-type devices in which the optical bandgaps of the Y-series acceptors were lower than those of the donors.ΔV SC s (SC stands for the short-circuit condition) were considerably smaller than the other losses, particularly ΔV nr , because ΔV SC is proportional to the logarithm of the ratio between J SC and J SC SQ .ΔV r and ΔV nr (r and nr stand for the radiative and nonradiative recombination, respectively) varied up to 0.151 and 0.387 V, respectively.As reported previously, both ΔV r and ΔV nr depend significantly on the energy difference ΔE between E g and E CT ; both ΔV r and ΔV nr are expected to decrease with decreasing ΔE. [2,8,12] When ΔE is small, the accurate determination of E CT is challenging and beyond the scope of this study; therefore, we used E g /qÀV OC rad as an alternative quantitative measure for the energy difference.As reported previously, V OC rad is linearly dependent on E CT except when ΔE is negligible. [8]Therefore, E g /qÀV OC rad scales linearly with ΔE. Figure S5a, Supporting Information, shows the variation in ΔV r plotted against E g /qÀV OC rad .A clear positive correlation was observed in the region where E g /qÀV OC rad > 0.35 V, where the EQE PV shoulder due to the CT absorption was observable.As the blackbody emission scales exponentially with decreasing energy, the emergence of the weak shoulder in the EQE PV spectra due to the CT absorption results in a substantial contribution to ΔV r .This led to a positive correlation between ΔV r and E g /qÀV OC rad .In contrast, ΔV r was scattered when E g /qÀV OC rad was less than or approximately equal to 0.35 V.Because the CT absorption in the EQE PV spectra was not observable in this region, ΔV r was governed by the smeared-out absorption edge of the narrow-bandgap materials, resulting in a weak correlation between ΔV r and E g /qÀV OC rad .In this region, ΔV r The experimental data for e-type devices were obtained from our previous report. [8]ower energy tails of EQE PV spectra were extended using the electroluminescence (EL) spectra (details are described in the Supporting Information).
instead depends on the Urbach energy [13] and EQE PV above the E g region of the device.The Urbach energies of PTB7-Th (as a narrow-bandgap material for e-type devices), Y1, Y5, and Y6 (as narrow-bandgap materials for h-type devices) were 27.8, 25.7, 27.3, and 27.8 meV, respectively (details can be found in Figure S6, Supporting Information), [14] indicating that there was no significant difference in the Urbach energy between the e-type and h-type devices.As a result, we found a positive correlation between ΔV r and the maximum EQE PV (Figure S5b, Supporting Information).Importantly, we did not observe any essential difference in ΔV r between e-type and h-type devices; both types of devices followed a general trend, as shown in Figure S5, Supporting Information.Rather, ΔV r tended to be larger for h-type devices owing to their lower E g and higher EQE PV values.Because the Urbach energies of our devices were close to the thermal energy at room temperature (%25 meV), there is minimal scope to reduce ΔV r .In fact, ΔV r s of both types of devices were less than 0.1 V when ΔE was small.In other words, ΔV r cannot be the decisive factor differentiating between the e-type and h-type devices.
As mentioned earlier, the largest source of voltage loss is ΔV nr .Therefore, significant scope exists to reduce ΔV nr even though the PCE of state-of-the-art OSCs approaches %20%.In this study, ΔV nr was determined from the difference between V OC rad and V OC .Alternatively, ΔV nr is related to the external quantum efficiency of electroluminescence (EQE EL ) from an OSC device under forward bias where k B and T are the Boltzmann constant and absolute temperature, respectively.Because CT states predominantly decay nonradiatively to the ground state, [15] EQE EL of typical OSCs is extremely low, resulting in a large ΔV nr .This means that a one-order increase in EQE EL results in an approximately 60 mV decrease in ΔV nr .Figure 3a shows the variation in ΔV nr plotted against E g /qÀV OC rad .A clear positive correlation was again observed when E g /qÀV OC rad > 0.35 V, as reported previously. [8]Importantly, both types of devices followed a general trend in this region.Conversely, ΔV nr of both e-type and h-type devices dropped sharply when E g /qÀV OC rad approached approximately 0.35 V, where E g and E CT were very close in energy.Strikingly, no apparent difference in ΔV nr was observed between e-type and h-type devices even in this region, whereas the smallest values of ΔV nr for e-type and h-type devices were slightly different; the smallest ΔV nr for e-type and h-type devices was 0.185 V (PTB7-Th:IDFBR) and 0.145 V (PM6:Y5), respectively (Figure 3b).Therefore, both types of devices can achieve a low ΔV nr when ΔE is small.Previous studies have pointed out that reducing ΔE leads to the hybridization of the CT state with the LE state, resulting in an enhancement in the (effective) oscillator strength of the CT state, and hence an enhancement in EQE EL . [8,12,16]2b,17] In contrast, because the LUMO energy offset of the PTB7-Th:IDFBR blend is smaller than the HOMO energy offset of the PM6:Y5 blend, difference in ΔE alone cannot explain the difference in ΔV nr between them.Therefore, we concluded that the smaller ΔV nr for the h-type device can be partly rationalized by the difference in the PLQY of the LE states (3.2% for PTB7-Th and 9.5% for Y5) because the large PLQY of the LE state is beneficial for the abovementioned two scenarios.Because recent narrow-bandgap NFAs, such as Y-series acceptors, consist of fused-ring A-D-A architectures, they tend to exhibit higher PLQYs than narrow-bandgap donor polymers owing to  the suppression of vibrational nonradiative deactivation, [18] which leads to higher EQE EL .In contrast, this also implies that e-type OSCs could achieve lower ΔV nr if highly emissive narrow-bandgap donors were developed.In other words, this is not an intrinsic advantage of h-type devices.Emphatically, however, what is critically important here is to achieve a low ΔV nr while maintaining a high EQE PV .As shown in Figure 4, h-type devices tended to exhibit a higher EQE PV compared with e-type devices when ΔE was small.Note that both types of OSCs can generate charge carriers efficiently when ΔE is large. [19]Therefore, the greatest advantage of h-type OSCs in recent studies is their ability to generate charge carriers with a small ΔE, whereas the mechanism of charge separation with a small ΔE is a subject of continuing debate. [20]A possible explanation for the mechanism of charge separation with a small ΔE is the formation of a cascaded energy landscape near the donor:acceptor interface produced by the large quadrupole moment of the A-D-Atype NFAs.This may be an intrinsic advantage for h-type OSCs because the A-D-A-type architecture inevitably reduces the optical bandgap of the acceptor.In contrast, EQE PV of the PM6:Y5 device, which exhibited the lowest ΔV nr , was noticeably low, resulting in a poor PCE of 6.65%.20d,21] Therefore, a top priority for future research is to fully elucidate the mechanism of charge separation with a small ΔE.

Conclusion
In this study, we investigated whether there is any essential difference in the voltage loss between e-type and h-type OSCs.We prepared nine e-type and eight h-type devices.The e-type devices consisted of PTB7-Th as a common narrow-bandgap donor paired with various wider-bandgap acceptors with different LUMO energy offsets.The h-type devices consisted of Y-series NFAs paired with various wider-bandgap donor polymers with different HOMO energy offsets.The radiative and nonradiative voltage losses (ΔV r and ΔV nr , respectively) of the e-type and h-type devices were quantitatively compared.For ΔV r , we did not observe any essential differences between the e-type and h-type devices.Generally, to reduce ΔV r , both the Urbach energy and the energy difference ΔE between E g and E CT should be reduced.Because the Urbach energies of both types of OSCs were similar, ΔV r exhibited a positive correlation against EQE PV , indicating that ΔV r of the h-type devices tended to be larger than those of the e-type devices owing to the higher EQE PV .In contrast, we observed intrinsic advantages for h-type OSCs with respect to ΔV nr .The first (but perhaps not essential) advantage of h-type OSCs is the higher PLQY of the narrowbandgap NFAs owing to their fused-ring architecture.The second (and more essential) advantage is their ability of more efficient charge separation with a small ΔE.Because the mechanism of charge separation with a small ΔE is not yet fully understood, future research should focus on fully elucidating the mechanism of charge separation with a small ΔE.
Device Fabrication: The photovoltaic devices were fabricated on ITO/ glass substrates (Geomatec Co., 1006, 10 Ω sq À1 ), which were cleaned by sonication consecutively in toluene, acetone, and ethanol, followed by UV-O 3 treatment (Nippon Laser and Electronics Lab.) for 30 min.PEDOT:PSS (Clevios, PH500 or AI4083) was spin-coated onto the substrates and dried on a hot plate under ambient conditions.Thereafter, the active layers were spin-coated onto the substrates in a N 2 -filled glovebox (see Table S1, Supporting Information).Subsequently, PFN-Br or PDINO was spin-coated from methanol solution, and 80 nm of Al was thermally evaporated thereon.The devices were encapsulated in the glovebox using a UV curable epoxy for the EL measurements, whereas they were set in a N 2filled chamber for the J-V and EQE PV measurements.In contrast, the samples for the optical measurements were prepared on quartz substrates.J-V and EQE PV Measurements: The J-V characteristics were measured using a DC voltage and current source/monitor (Keithley, 2611B) in the dark and under AM1.5G simulated solar illumination at 100 mW cm À2 .The light intensity was corrected with a calibrated Si photodiode (Bunko-Keiki, BS-520).The EQE PV spectra were measured using a spectral response measurement system (Bunko-Keiki, ECT-25D).
Steady-State Absorption, PL, and EL Measurements: UV-vis absorption spectra were measured using a UV-vis spectrometer (Hitachi, U-4100).PL and EL spectra were measured using a PL spectrometer (Horiba Jobin Yvon, Nanolog) equipped with a photomultiplier tube (Hamamatsu, R928P) and a liquid-N 2 -cooled InGaAs near-IR array detector (Horiba Jobin Yvon, Symphony II).PLQY and EQE EL were measured using an absolute quantum yield measurement system (Bunko-Keiki, BEL-300) with an integrating sphere.A DC voltage and current source/monitor (Advantest, R-6243) were used to adjust the applied voltage.
Cyclic Voltammetry Measurements: Materials were dissolved in acetonitrile/o-dichlorobenzene (9:1 v/v) solutions with 0.1 M tetrabutylammonium perchlorate as the supporting electrolyte.Cyclic voltammetry measurements were performed using a potentiostat (Princeton Applied Research, Potentiostat/Galvanostat Model 273A) at a scanning rate of 5-50 mV s À1 .A Pt mesh, a Ag/AgCl wire and a thinly sliced ITO substrate were used as the counter, reference, and working electrode, respectively.Ferrocene was used as an internal reference.The HOMO and LUMO energies were then evaluated from the onset potential of the first oxidation and reduction peaks (φ Ox and φ Red , respectively) as E HOMO/LUMO = À(φ Ox/Red þ 4.8 À φ Fe/Feþ ), where φ Fe/Feþ is the redox potential of ferrocene/ferrocenium as an internal reference.

Figure 2 .
Figure2.EQE PV spectra of a) e-type and b) h-type OSC devices.The experimental data for e-type devices were obtained from our previous report.[8]Lower energy tails of EQE PV spectra were extended using the electroluminescence (EL) spectra (details are described in the Supporting Information).

Figure 4 .
Figure 4. Maximum EQE PV for OSC devices with E g /q À V OC rad ≈ 0.35 V.

Table 1 .
Summary of voltage losses.