Discerning Blend Microstructure and Charge Recombination for Stable Biorenewable-Based Organic Photovoltaics

Abstract

The power conversion efficiency of organic photovoltaics (OPV) has recently surpassed 20%. However, the degradation mechanisms affecting blends based on these materials require urgent attention to improve the stability of such devices towards the long timescales necessary for commercialization. In this work, we evaluated the degradation of OPVs based on sustainable and scalable donors poly[(thiophene)-alt-(6,7-difluoro-2-(2-hexyldecyloxy)quinoxaline)] (PTQ10) and poly[(5-fluoro-6-((2-hexyldecyl)oxy)benzo[c][1,2,5]thiadiazole)-alt-thiophene] (FO6-T) blended with Y-family NFAs with different side-chain lengths processed from biorenewable 2MeTHF for PTQ10:Y12 and FO6-T:Y12 and from chloroform for FO6-T:Y6 blends. Superior stability is observed for FO6-T:Y12 with an extrapolated T80 of over 2000 h under LED illumination, and a more stable trend under metal halide lamps illumination compared to the other blends. By analyzing the thin film microstructure using Atomic Force Microscope (AFM), a significant phase separation is observed in the Y6-based blend, compared to PTQ10:Y12 and FO6-T:Y12, and a clear red-shift in the UV–vis profile. The superior stability of the FO6-T:Y12 blend is attributed to less morphological degradation upon aging and the increased number of photogenerated charges upon degradation. Finally, through a series of light intensity and temperature-dependent J–V characterizations, we evaluated the recombination mechanisms.

1 Introduction

The power conversion efficiency of organic photovoltaics (OPVs) based on bulk heterojunctions (BHJ) has recently surpassed 20%, driven by the development of novel highly crystallizable non-fullerene acceptors (NFAs) in the last decade.^{[1, 2]} However, these devices still suffer from several drawbacks that hinder commercialization, in particular, challenges in scalability and poor operational lifetimes. The stability of OPVss has been extensively studied in fullerene-based systems,^[3-5] and new research has emerged in recent years regarding the degradation pathways of NFAs, but their stability in devices remains poor.^[6-8]

Recently, there has been a lot of progress in improving the stability of OPVs^{[9, 10]} increasing the time it takes the power conversion efficiency (PCE) to drop below 80% (T₈₀). Various approaches have been employed to achieve such high values, examples including surface acid treatment with 2-phenylethylmercaptan (PET) of the electron transporting layer (ETL) reaching a T₈₀ of 4000 h under continuous illumination at maximal power point,^[11] adding insulating polymer matrices,^[12] using vacuum-assisted thermal annealing,^[13] and introducing UV-filtering buffer layers.^[14]

In short, several degradation mechanisms affect OPVs, resulting from one or more environmental stressors, such as water, oxygen, light, heat, and mechanical stress. We usually distinguish three different degradation conditions: Thermal degradation^[15] refers to performance losses induced by exposure to different temperatures, which often causes changes in the active layer's microstructure. Since the optimal morphology of the active layer (AL) is not usually at the thermodynamically equilibrated state, external stressors such as heat can induce changes in the AL's microstructure, in the form of a demixing of the components decreasing the donor: acceptor surface area, resulting in performance losses upon degradation.^[7] Photostability, refers to the performance stability under light in an inert atmosphere.^[16] Lastly, photochemical degradation considers the exposure to light as well as ambient air containing oxygen and humidity, which often leads to chemical degradation of the molecules themselves rather than microstructural changes.

Upon illumination, most devices undergo an initial phase referred to as “burn-in” degradation, where the PCE drops significantly over a short time, usually within a few hours.^[17-19] The origins of burn-in are manifold, and include blend de-mixing, trap-state formation, photo-oxidation, and interlayer degradation, among others. Burn-in due to morphological instabilities in fullerene-based systems has been extensively studied. Regarding degradation mechanisms in NFA-based solar cells several reports on understanding the degradation pathways during burn-in for NFAs in the ITIC, IEICO, IDFBR, and IDTBR series have been reported.^[20]

So far, the photostability of Y-series NFAs has been demonstrated to be strongly dependent on side-chain length, branching, and positioning leading to more stable systems when larger and bulkier chains are used.^[9] Furthermore, it has been shown that end-group functionalization also affects thermal and photo-stability^{[21, 22]} and can suppress photodegradation pathways in outdoor conditions (heat and light) by inducing structural confinement, which increases stability.^[23]

Aside from NFAs, the photostability and degradation pathways of polymer donor materials for OPVs have also been studied in the past few years. In a study by Wang et al., PTQ10 has demonstrated improved photostability compared to PM6 and D18 in blend with Y6, which they attribute to a light-induced dihedral twisting in the backbones of the latter two polymers.^[6] With sustainability in mind, we also recently reported the photostability of OPVs based on low synthetic complexity polymers, using PTQ10:Y12 and FO6-T:Y12 blends in biorenewable solvents, and showed that FO6-T-based devices maintained a high efficiency throughout the degradation cycle (1000 h under LED light illumination, 100 mW cm⁻², at maximum power point (MPP) tracking, whereas PTQ10-based ones suffered a 20% reduction in PCE after 350 h.^[24]

In this work, we evaluate the photodegradation mechanisms of blend devices made using either two Y-NFAs with fluorinated end-groups, 2,2′-((2Z,2′Z)-((12,13-bis(2-butyloctyl)-3,9-diundecyl-12,13-dihydro-[1,2,5]thiadiazolo[3,4-e]thieno[2″,3″:4′,5′]thieno[2′,3′:4,5]pyrrolo[3,2-g]thieno[2′,3′:4,5]indole-2,10-diyl)bis(methanylylidene))bis(5,6-difluoro-3-oxo-2,3-dihydro-1H-indene-2,1-diylidene))dimalononitrile (Y6) and 2,2′-((2Z,2′Z)-((12,13-bis(2-butyloctyl)-3,9-diundecyl-12,13-dihydro-[1,2,5]thiadiazolo[3,4-e]thieno[2″,3″:4′,5′]thieno[2′,3′:4,5]pyrrolo[3,2-g]thieno[2′,3′:4,5]thieno[3,2-b]indole-2,10-diyl)bis(methanylylidene))bis(5,6-difluoro-3-oxo-2,3-dihydro-1H-indene-2,1-diylidene))dimalononitrile (Y12), and two low synthetic complexity polymers, poly[(thiophene)-alt-(6,7-difluoro-2-(2-hexyldecyloxy)quinoxaline)](PTQ10) and poly[(5-fluoro-6-((2-hexyldecyl)oxy)benzo[c][1,2,5]thiadiazole)-alt-thiophene (FO6-T).^[25] These particular polymer donors have been selected in this study due to their high solubility limits in non-chlorinated solvents, their low synthetic complexity, and their ability for high photovoltaic performances when blended with NFAs, making them upscaling compatible. Unencapsulated OPVs were degraded under LED light and metal-halide lamp (MHL) illumination source at 100 mW cm⁻² intensity in a nitrogen environment, while kept at their MPP at 45 °C. Under LED degradation, OPVs were more stable when compared with the MHL degradation, with the FO6-T:Y12 OPVs reaching a T₈₀ of over 2000 h. In contrast, FO6-T:Y6 OPVs demonstrated a much faster degradation under MHL, and upon further investigation, a clear red-shift in absorption was observed which was assigned to aggregation as confirmed with Atomic Force Microscopy (AFM). Light-induced Electron Paramagnetic Resonance (L-EPR) Spectroscopy revealed that degrading the FO6-T: Y12 blend films for 1000 h increases the number of photogenerated charges in the system under light excitation compared to the fresh blend, whereas in FO6-T:Y6 based blends the number of photogenerated charges after aging is reduced, aligned with the stability results. Finally, by performing temperature-dependent characterization of the light intensity dependence of V_oc (suns-V_oc) of fresh and aged OPVs, we investigated the shape of the density of states (DOS) in the blends. We concluded that interface recombination pathways become dominant in degraded Y6-based devices, while the recombination in Y12-based devices is consistent with a power law DOS, suggesting that recombination occurs predominantly in the bulk, as observed before.^[15]

2 Results and Discussion

2.1 Maximum Power Point Tracking

Unencapsulated inverted structure OPVs based on ITO/ZnO/AL/MoOx/Ag were placed in a nitrogen-filled metallic chamber and kept at their MPP for over 1000 h under metal-halide lamp (MHL) illumination, 1 Sun, under ISOS-L-1 protocol at 45 °C.^[26] For evaluating indoor stability, the same experimental conditions were used with the light source replaced by LED lamps of the same light intensity (100 mW cm⁻²). The spectra of both light sources are shown in Figure S1 (Supporting Information). The performance of the devices before and after degradation under both light sources is summarised in Table S1 (Supporting Information), while their J–V characteristics and EQEs are shown in Figure S2 (Supporting Information). The values of the shunt and series resistances R_sh and R_s, respectively, of the fresh and aged devices under MHL illumination, are summarized in Table S2 (Supporting Information).

Under both light degradation conditions, the devices demonstrated a similar degradation behavior, where FO6-T:Y12 remains more stable compared to the PTQ10:Y12 blend and both show better stability than a blend containing Y6 as the acceptor i.e. (FO6-T:Y6), as presented in Figure 1. However, despite the similar degradation patterns between MHL and indoor light, the T₈₀ of the systems differs significantly. It is worth noting the UV part in the MHL spectra (Figure S1, Supporting Information) that would have an additional impact on the OPV degradation, due to the impact of UV irradiation on the properties of ZnO, which is the electron transport layer in the device.^{[27, 28]} Despite the initial burn-in, FO6-T-based OPVsdemonstrate a stable performance over time, suggesting superior stability compared to PTQ10-based ones. The absence of NIR light in the LED sources reduces the burn-in (most likely due to reduced heating in the sample) with FO6-T:Y12 exhibiting an extrapolated T₈₀ over 2000 h under indoor light (Figure S3, Supporting Information). Overall, under both light conditions, the most stable devices were those based on FO6-T:Y12, followed by PTQ10:Y12 and FO6-T:Y6, suggesting that the short side chains in Y6 have a significant impact on device degradation despite FO6-T being a very stable donor.

Figure 1

Inverted structure OPVs (inset) degradation in a nitrogen-filled chamber while kept at their MPP under a) an MHL light source at 45 °C and b) an LED light source, containing FO6-T:Y12, PTQ10:Y12, and FO6-T:Y6 showing the Power at MPP (Pmpp) evolution over time.

For the Y6-based OPVs, the highest-performing system contains 1,8-diodoctane (DIO) as a liquid additive, as we have reported previously.^[25] Since the Y12-based blends did not contain additives, we also evaluated the stability of FO6-T:Y6 OPVs without the additive. Consistent with previous reports,^[29] the blend with the additive demonstrates a faster degradation rate than the OPVs without it (Figure S4, Supporting Information), but even without additive, Y6-based devices degrade faster than Y12-based devices.

In terms of photovoltaic performance, it is clear from Figure S2 and Table S1 (Supporting Information) that devices based on Y12 aged under LED light are less susceptible to a loss of fill factor (FF) compared to the blend based on Y6, which sees a strong reduction in FF of 20% of the original value. All the devices degraded under MHL showed a drastic loss in FF, as discussed in the following section, with the FO6-T:Y6 blend also suffering a more significant loss in short circuit current density (J_sc) compared to the Y12 blends (Figure S4a, Supporting Information). These trends are in line with the values of R_sh (Table S2, Supporting Information), which decrease in all blends after MHL degradation. It is also clear from Table S2 (Supporting Information) that the FF drop in the FO6-T:Y12 and PTQ10:Y12 is more strongly correlated to an increase in the R_s compared to FO6-T:Y6, which is mainly correlated to a reduction in R_sh.

2.2 Degradation of Thin Films

To further understand the differences between the two polymeric donors and the NFAs upon light and thermal degradation, we performed optical and morphological characterizations of fresh and aged (1000 h under MHL at 45 °C) thin films. Figure 2a–c demonstrates the absorption profiles of FO6-T:Y12, PTQ10:Y12, and FO6-T:Y6 with DIO fresh and degraded thin films. The comparison between their absorption spectra revealed that the absorption onset of the NFA red-shifts by ≈20 nm for the Y6-containing blends, suggesting increased aggregation of the acceptor during degradation (Figure 2c).^[30] For the blends containing Y12, such a shift is not observed, suggesting that the microstructure remained similar throughout the degradation process. These results are consistent with the picture that the longer side chains in Y12 may be preventing the disruption of the blend's morphology and improving stability consequently, in line with what has been previously reported.^[23] We further confirm this by analyzing the surface morphology of the blends using atomic force microscopy (AFM), as shown in Figure 2e,d. The aged FO6-T:Y6 films show a significant disruption in their morphology, compared to FO6-T:Y12 and PTQ10:Y12. Both Y12-containing blends reveal a similar morphology before and after degradation, whereas a significant phase separation is observed in the FO6-T:Y6 with a clear sign of agglomeration. The profound level of phase separation and aggregation-induced photodegradation is likely to be primarily responsible for the degradation of the OPVs and FF loss.^{[31, 32]}

Figure 2

Thin film characterization of fresh and aged films (1000 h under MHL in nitrogen, 45 °C): UV–vis profiles of a) FO6-T:Y12, b) PTQ10:Y12, and b) FO6-T:Y6 (DIO). AFM topography of c) fresh and d) aged thin films of FO6-T:Y12, PTQ10:Y12, and FO6-T:Y6 (DIO). The scale bar is 200 nm.

The glass transition temperature (T_g) of pristine (Figure S7, Supporting Information) and blend films (Figure 3; Table S3, Supporting Information) was determined through Fast Scanning Calorimetry (FSC), in order to understand the impact of the T_g on the photostability of the blends. Here, we note that at temperatures below T_g, the materials undergo physical aging, i.e., structural reconfigurations in the glassy material that result in the reduction of its enthalpy (ΔH).^[33] Consequently, aiming at detecting physical aging phenomena, spun cast blends were subjected to the thermal protocol included in Figure S6 (Supporting Information). Figure 3a–c shows the raw heat-flow-rate versus temperature curves for the heating scans applied to fresh and aged samples (referred to as “reference” and “aged” respectively in Figure S6, Supporting Information). When films are subject to isothermal aging below the T_g, an endothermic overshoot shows up in the “aged” heating scans, distinguishing it from the “reference” heating scan, which suggests that the sample underwent physical aging during the isothermal step. Conversely, when annealing temperatures are higher than the T_g, no overshoot is detected, and “aged” and “reference” curves superimpose. The area contained in the overshoot corresponds to the enthalpy relaxed during the physical aging process, ΔH. Thus, the annealing temperature at which ΔH equals to 0 corresponds to the upper limit of the T_g (see Figure 3d–f). The T_g value thus obtained for the FO6-T:Y6 blend was ≈200 °C, while that for the FO6-T:Y12 and the PTQ10:Y12 blends was found ≈180 °C (Table S4, Supporting Information). An inverse relationship between T_g and stability is obtained, and the T_g of the blends is influenced from the higher T_g component; the NFA (Figure S7, Supporting Information). This relationship has been observed before and the lower Tg for Y12 is attributed to the longer internal side-chains in Y12 versus Y6.^[23] It is worth noting that the fabrication temperature and the maximum temperature reached during degradation are far below the T_g of the blends, so all aging data should refer to the regime where structural reorganization is expected.

Figure 3

DSC thermogram from fast scanning calorimetry for the a) FO6-T:Y12, b) FO6-T:Y6, and c) PTQ10:Y12 blends aged at different temperatures. The two thick lines for each temperature correspond to heating scans referred to as “aged” (the higher of the two) and “reference” in Figure S6 (Supporting Information). The number beside the right-hand axis indicates the temperature at which the sample was aged for each plot. d–f) relaxed ΔH versus aging temperature (T_aging) FO6-T:Y12, b) FO6-T:Y6 and c) PTQ10:Y12 blends.

2.3 Light-Induced EPR

To gain insights into the impact of film degradation on the formation of photoexcited species,^{[34, 35]} we carried out light-induced continuous-wave electron paramagnetic resonance spectroscopy (L-EPR). Fresh and aged (under MHL) samples were measured upon steady-state LED excitation of 530 and 780 nm to selectively photoexcite the donor and acceptor species, respectively.^[6] The signal intensity is proportional to the photogeneration efficiency and polaron recombination lifetime, therefore a higher signal intensity correlates to longer-lived photogenerated species and/or a more efficient charge generation.^[36] Upon illumination, thin films containing Y6 species have a lower EPR intensity signal after degradation (Figure 4a), independently of the illumination conditions employed to generate radical species (Figures S8 and S9, Supporting Information). This effect is also exacerbated for FO6-T:Y6 blends when compared to PTQ10:Y12 blends. By further analyzing the EPR signals and using a calibration curve, it is possible to quantify the number of spins (see Experimental Section for further discussion). Spin concentration in the order of 10¹⁶ cm⁻³ was obtained for all samples (Figure 4c), but interestingly, upon degradation, all blend systems behave differently. FO6-T:Y6 blends with and without DIO show a decrease in their spin concentration from ≈4 × 10¹⁶ to ≈1 × 10¹⁶ cm⁻³, whereas the FO6-T:Y12 blend increases from ≈1.3 × 10¹⁶ to ≈3.4 × 10¹⁶ cm⁻³, and PTQ10:Y12 remains at ≈1.4 × 10¹⁶ cm⁻³. This effect was observed when samples were excited with the 530 and 780 nm LED (Figure S10, Supporting Information). Aging has a negative impact on Y6-based blends, for which phase separation and the formation of aggregates (Figure 2) lead to increased recombination losses. The increase in the spin concentration and EPR signal intensity after sample degradation has been observed in pristine polymers,^{[37, 38]} but not, to our knowledge, for donor: acceptor blends. The data suggest that FO6-T:Y12 promotes a longer lifetime of photogenerated species than the other blends, which may potentially be beneficial for the long-term stability of devices in real-world operation conditions.

Figure 4

L-EPR signal intensity was recorded at 40 K, upon excitation at 780 nm (exciting the acceptor) for a) FO6-T:Y6 and b) FO6-T:Y12 blends of fresh and aged (MHL) thin films. c) Spin concentration of the FO6-T:Y6, FO6-T:Y6 with DIO, FO6-T:Y12, and d) PTQ10:Y12 films calculated from EPR signals.

For a deeper understanding of the recombination processes in FO6-T:Y6 and FO6-T:Y12 fresh and aged blends, the transient response of light-induced EPR is analyzed, upon turning on and off the light illumination (Figures S11–S14, Supporting Information).^{[36, 39]} The results are consistent with steady-state light-induced EPR (Figure 4), for which organic thin films with efficient charge generation and slow recombination (e.g., via trap states) are expected to display strong signals.^[40] For FO6-T:Y6 films, changes in radical formation and recombination time are observed upon aging (Table S5, Supporting Information). This correlates with the disruption in morphology observed that may lead to faster recombination processes at the interfaces.^[38] For FO6-T:Y12, instead, negligible variations in radical generation and recombination times are observed upon aging, with little variation in the charge generation times when the LED source was replaced from 530 to 780 nm, suggesting that the morphological change is not drastically affecting the recombination.^[41]

To understand this mechanism better, we then performed a series of electrical characterization in the OPVs by varying the temperature and light intensity in an effort to map the differences in the DOSof the studied blends.

2.4 Temperature & Light Intensity Dependent J–V Characterisation

The recombination mechanisms and the shape of the DOS were further investigated from devices kept in the dark under nitrogen (“fresh”) and two sets of aged devices; where for the first set of aged devices, the OPVs were degraded for up to 330 h (PV parameters are presented in Figure S2 (Supporting Information). The devices show similar degradation rates compared to the ones presented in Figure 1 that were degraded for 1000 h). For the second set of aged devices (“indoor”) we used the same set of OPVs degraded under an LED light source (100 mW cm⁻² intensity) for 1000 h as shown in Figure 1.

To analyze the DOS, we follow a technique introduced by Saladina, et al.^[15] making use of temperature-dependent suns-Voc characteristics (details in the Supporting Information). Suns-Voc measurements at different temperatures were recorded for all three active layer combinations (PTQ10:Y12, FO6T:Y12, FO6T:Y6) in OPVs under each aging condition outlined above (fresh, indoor, MHL). The J–V characteristics of all devices are shown Figures S15–S23 (Supporting Information). Using Equation (1) which defines the ideality factor n_id in terms of light intensity (Φ) and V_oc and Equation (2), which defines the Urbach energy (E_U) in terms of n_id, we calculated the temperature-dependent ideality factors and the Urbach Energies from the dependence of V_oc on light intensity at different temperatures. The results are shown in Figures S24–S27 (Supporting Information). The model for Equation (2) assumes a mixed DOS, in which the trapped electrons from an exponential DOS recombine with mobile holes from a Gaussian DOS:^[15]

$$$$\begin{equation}{n_{id}} = \frac{e}{{{k_B}T}}{\left( {\frac{{d\ln \Phi }}{{d{V_{oc}}}}} \right)^{ - 1}}\end{equation}$$$$(1)

$$$$\begin{equation}{E_U} = {k_B}\;T\left( {2{n_{id}} - 1} \right)\end{equation}$$$$(2)

Figure 5 shows the E_u as a function of V_oc which we can consider as representative of the quasi-Fermi level splitting (for any given temperature and light intensity). When E_u(V_oc) is temperature independent, all data points collapse onto a single line, which is consistent with the recombination processes being dominated by trap states in the bulk. In systems where the DOS is strictly exponential, the E_U would be independent of the V_OC. Experimentally, however, we observe that E_U(V_OC) has a non-zero slope following $${E_u} = \frac{{( {E - {E_0}} )}}{\xi }$$ (where E  =  eV_oc, and ξ is negative) which means that the exponential DOS is only a local approximation of the real DOS. Deeper into the gap, Eu becomes larger, corresponding to a shallower slope of the DOS, which consequentially extends deeper into the gap than would be expected from a pure exponential. The real shape of the DOS for E < E₀ can be described with a power-law according to Equation (3) (See Supporting Information for detailed derivation):

$$$$\begin{equation}DOS\left( E \right) = {\left( {{E_0} - E} \right)^\xi }\;\end{equation}$$$$(3)

Figure 5

a–c) Urbach energy versus V_OC for fresh and aged OPVs under indoor light conditions and MHL for a) FO6-T:Y12, b) PTQ10:Y12, and c) FO6-T:Y6 with DIO. The data were fitted with $${E_u} = \frac{{( {E - {E_0}} )}}{\xi }$$ (lines) including the 95% confidence interval (shaded area). Crosses were excluded in the fitting. d–f) DOS calculated from the fitting parameters E₀ and ξ according to the relation DOS(E)  = (E₀ − E)^ξ . The shaded area shows the upper and lower bound of the DOS when varying the fitting parameters within their margin of error. This procedure for error propagation was chosen due to the high non-linearity of the relationship. The black straight dashed line represents an example of an exponential DOS with E_u = 26 meV and E₀ = 0.7 eV.

Here, E₀ might be interpreted as the mobility edge and ξ represents how fast the number of trap states decays into the gap. We note that Equation (3) is only applicable for a limited range of E, and it is not expected to accurately describe the DOS where the tail merges with the bulk DOS. In Figure 5d–f we show the DOS of the different systems that were calculated using Equation 3. The parameters E₀ and ξ were obtained from fitting the data in Figure 5a–c with the linear E_U (V_OC) =  (eV_OC − E₀)/ξ, and are shown in the Table S6 (Supporting Information).

For all the material systems analyzed in this study, the fresh devices followed this clear, monotonic relationship between E_U and V_OC, suggesting that recombination occurs predominantly in the bulk controlled by a mixed DOS; where the electron traps deep in the gap follow the power-law described in Equation (3). This behavior is in agreement with previous studies in similar OPVs, based on PM6:Y6 bulk heterojunctions.^[15]

The degradation under MHL leads to a shift in the E_u (V_oc) to lower V_oc in all devices, which is due to the initial burn-in (drastic drop in V_oc during the first 10 h) observed in the MPP tracking. Both FO6-T-based systems show a ≈0.2 V reduction in V_oc, whereas the PTQ10:Y12 a ≈0.3 V, which is consistent with the V_oc evolution over time (Figure S4, Supporting Information). The indoor light degradation, on the other hand, does not lead to any shift in the E_u (V_oc) in the FO6T:Y12 device, consistent with the device V_oc under aging, and only to a minimal shift in the PTQ10:Y12. In the shape of the DOS, these V_oc shifts are reflected by a shift in E₀, which shifts the point at which Equation (3) diverges. The rate of decay (describing the number of deep traps) is determined by the exponent ξ (shown in Table S6, Supporting Information). It is evident that PTQ10:Y12 shows a greater increase in the magnitude of the DOS between fresh and MHL degraded than the FO6-T:Y12 OPVs, which can be attributed to increased traps in the bulk, as calculated by E₀. Next to the calculated DOS in Figure 5e,f, we show a hypothetical exponential DOS, which highlights that the traps extend more deeply into the gap than they would in the pure exponential scenario.

In all the aged Y6-based devices, however, there is not only a V_oc drop, but E_u (V_oc) also becomes temperature-dependent, with a steeper slope as the temperature decreases (Figures S27 and S28, Supporting Information). The steeper slope in the E_u (V_oc) indicates the presence of more deep bulk traps. However, the fact that the data points deviate from a single line implies the breakdown of at least one of the assumptions in the model. Therefore, the fitting and DOS calculation is not meaningful for those data (Attempted fitting of the aged devices is shown in Figure S24, Supporting Information). One possible explanation for this behavior is that recombination in the bulk is no longer the dominant pathway, but that interfacial recombination at the contacts instead becomes significant. Given that only the Y6-based devices are affected, we rule out that this effect is caused by the degradation of the contact interlayers themselves, which would affect all devices equally. Together with the strong phase separation seen in the AFM topography of those devices (Figure 2d,e) and the red shift in the absorption spectra of the aged samples, we can hypothesize that the clustering of large Y6 domains leads to increased interfacial recombination at the hole-collecting contacts.

3 Conclusion

In this work, an extensive analysis is conducted to evaluate the degradation of two NFAs, Y6 and Y12, and two low-synthetic complexity polymeric donors, FO6-T and PTQ10, when implemented in OPVs. Devices were degraded in a nitrogen-contained chamber, under MPP tracking, from an LED-light source (with intensity of 100 mW cm⁻²) and an MHL (with intensity of 100 mW cm⁻² and 45 °C). Under both degradation lighting conditions, FO6-T:Y12 is the most stable system, with an extrapolated T₈₀ upon LED degradation of over 2000 h. FO6-T:Y6, on the other hand, shows faster degradation, morphological instability, and a red-shift in the absorption profile of the aged films. Moreover, FO6-T:Y12 OPVs showed superior stability over PTQ10:Y12 OPVs, which is attributed to the increased number of photogenerated charges present in the blend upon exciting either the donor or the acceptor after thin film aging under MHL of 100 mW cm⁻², as observed by the L-EPR. Finally, a light-temperature J–V characterization revealed that all fresh OPV devices show that recombination primarily occurs in the bulk controlled by a mixed DOS, but upon aging FO6-T:Y6 OPVs show that interface recombination is the dominant mechanism, and the two Y12-based OPVs maintain the bulk recombination with the PTQ10:Y12 demonstrating a higher number of bulk traps compared to FO6-T:Y12.

4 Experimental Section

Device Fabrication

PTQ10 (Mw 46 Kg mol⁻¹, Đ 2.16 measured under 150 °C in 1,2,4-trichlorobenzene), Y6, and Y12 were purchased from Brilliant Matters. FO6-T was synthesized with Mw 69 Kg mol⁻¹, Đ 2.9 as previously reported,^[42] measured using an analytical Agilent Technologies 1200 series GPC using an IR and UV detector in 150 °C in 1,2,4-trichlorobenzene.

The OPVs were fabricated using an inverted architecture with the structure ITO/ZnO/AL/MoO_x/Ag using pre-patterned indium tin oxide (ITO) on glass. The ITO substrates were cleaned in sequential sonication rounds of distilled/detergent water, acetone, and isopropanol. They were treated with an oxygen plasma treatment for 8 min prior to device fabrication. The ZnO was spin-coated at 4000 rpm for 40 s and annealed at 180 °C for 10 min in ambient air environment, from a solution based on 219.5 mg of zinc acetate dehydrate, 2 ml of 2-methoxyethanol and 60.4 µL of ethanolamine. The AL blends had a total concentration of 18 mg ml⁻¹ for PTQ10:Y12, 22 mg ml⁻¹ for FO6-T:Y12, and 20 mg ml⁻¹ for FO6-T:Y6, with the Y12-based blends having a donor: acceptor (D:A) ratio of 1:1.2 in the biorenewable 2-Methyltetrahydrofuran, and FO6-T:Y6 with a D:A ratio of 1:1.5 in chloroform as previously reported in the literature. In the case of FO6-T:Y6, the DIO FO6-T:Y6 blend had 0.5% v/v concentration of 1,8-diiodooctane (DIO). Solutions containing Y12 and based on 2meTHF were heated at 55 °C for 15 min prior to deposition. The blends were deposited using spin-coating from the following conditions: i) FO6-T: Y12 at 3000 rpm for 45 sec, ii) PTQ10:Y12 at 2500 rpm for 45 sec and iii) FO6-T:Y6 at 2000 rpm for 40 sec. followed by annealing at 100 °C for 10 min. The MoO_x was used as the hole transporting layer, which it was deposited by  thermal evaporation with a thickness of 10 nm, followed by 100 nm of Ag as the top electrode through shadow masks forming devices with active area of 0.045 cm²

OPV Characterization

Current–voltage measurements were recorded with a 2400 Keithley Source–Measure unit using an Oriel Instruments Solar Simulator with a xenon lamp, calibrated with a Newport silicon cell to ensure AM1.5G. The EQE was measured in air with a Quantum Design PV300 system.

Device Degradation

The OSCs were placed in a nitrogen-filled metallic chamber and kept at their maximum power point (MPP) for 1000 h under MHL illumination, according to the ISOS-L-1 protocol and 45 °C, and under indoor illumination using an LED light source with the spectrum shown in Figure S1 (Supporting Information).

Optical and Morphological Characterization

UV–vis measurements were conducted on thin films on glass substrates using a UV-1601 Shimadzu spectrometer. The AFM images were obtained using an Agilent Technologies Keysight 5500 SPM in tapping mode.

L-EPR

Light-induced EPR measurements were collected on an EMX X-band CW-EPR spectrometer (Bruker) equipped with an ER 4122SHQE high-sensitivity cylindrical resonator (Bruker) and a closed-circuit Helium cryostat (Cryogenic Ltd.). The temperature was controlled using a LakeShore 350 ITC (Lake Shore Cryotronics, Inc.) and a Cernox sensor. Thin films (1.1 × 1.1 cm²) were prepared on PET substrates, as described in the “Device Fabrication” section, and placed inside an open-end quartz tube (Wilmad Labglass) with inner dimater of 3 mm, and outer 4 mm.). Samples were illuminated through the optical access at the front of the EPR resonator (height = 2.5 cm, width = 2.5 cm) with two Thorlabs laser diodes (λ = 530 nm with power density of 12.5 mW cm⁻² and λ = 780 nm with power density of 32.4 mW cm⁻²); a home-build support was used to hold the light sources in position during the measurements. All steady-state spectra were acquired under constant light irradiation. The EPR measurements were performed at a microwave frequency of ≈9.37 GHz and a temperature of 40 K using a microwave power of 0.02 mW, corresponding to strictly non-saturating conditions, a field modulation of 1 G at 100 kHz, a magnetic field scan rate of 0.6 G s⁻¹, a conversion time of 81.92 ms and a time constant of 20.48 ms; the acquisition of each spectrum lasted 84 s and 16 spectra were averaged for each sample. Spin quantification was performed by calibrating the spectrometer against a set of Cu(H₂O)₆ standards (Figure S10c, Supporting Information). These samples were prepared by dissolving anhydrous CuSO₄ in water/Ethylene Glycol (70:30) at concentrations of 200, 150, 100, 50, and 10 µm; 2 m NaClO₄(aq) was added and the pH was adjusted to 1.22 with HCl to prevent the aggregation of copper and obtain single-component spectra.^[43] The solutions were loaded into quartz tubes, and the fill height was adjusted to match the height of the device films, namely 1.1 cm. The fill height corresponds to a sample volume of 77.8 µL; this number was used to convert concentrations into spin numbers. The Cu(II) standards were measured at a temperature of 80 K using a microwave power of 2 mW, a field modulation of 10 G at 100 kHz, a magnetic field scan rate of 8.9 G s⁻¹, a conversion time of 163.84 ms and a time constant of 40.96 ms. When applying the obtained calibration curve to perform spin quantification on the films, the different measurement conditions were taken into account by rescaling the double integral by the modulation amplitude, the square root of the microwave power, and the reciprocal temperature. The spin concentration in the film samples was eventually obtained by dividing the spin number obtained from the quantitative analysis by the volume of the film, estimated to be 1.45 × 10⁻⁵ cm³. Time-domain experiments were performed by cycling the light source on and off for 25 cycles, with the duration of each cycle set to 5 s. The corresponding traces were measured at the magnetic field corresponding to the maximum of the steady-state spectra using a microwave power of 0.02 mW, a field modulation of 4 G at 100 kHz, and a time constant of 1.28 ms; the signal intensity was monitored every 5 ms. The transients corresponding to the individual on/off cycles were averaged after the acquisition. EPR data analysis and fittings were carried out using the function pepper of the EasySpin MATLAB toolbox^[44] (version 5.2.36).

Suns-V_oc Measurements

For the sun's V_oc measurements, the solar cell devices were mounted in a Linkam probe stage (HFS600E-PB4). The devices were illuminated with a ring of white UMILE LEDs. The light intensity of the LEDs was calibrated by matching the V_oc and J_sc of the devices under 1 sun condition. J-V curves were measured using a Keithley 236 source measure unit. Measurements were recorded starting at room temperature (300 K) after which the device was cooled down stepwise to 120 K with a cooling speed of max 20 K min⁻¹. During the cooling process before each measurement, the devices were illuminated with UV light for at least 1 min to exclude a UV light soaking effect associated with ZnO. At each temperature, the J–V s were then measured for a range of different light intensities, ranging from 1 to 1 × 10⁻⁴ sun equivalent; re-measuring at 1 sun equivalent at the end of every light intensity scan, as well as re-measuring at 300 K after reheating the device to check whether the devices have degraded throughout the measurement process.

Acknowledgements

E.M., Z.Q., and J.M. contributed equally to this work. J.P. acknowledges financial support from the Engineering Physical Science Research Council (EPSRC; EP/V057839/1 and EP/X52556X/1). F.E. and J.N. thank the European Research Council for support under the European Union's Horizon 2020 research and innovation program (Grant No. 742708). J.N. thanks the Royal Society for the award of a Research Professorship. J.S.M. and J.N. thank the UK Engineering and Physical Sciences Research Council for funding via the ATIP program grant (EP/T028513/1). C.D. thanks the Deutsche Forschungsgemeinschaft for support (DFG Research Unit FOR 5387 POPULAR, Project No. 461909888). J.M thanks MICINN for the grant Ref. PID2021-126243NB-I00. The EPR measurements were performed at the Centre for Pulse EPR at Imperial College London (PEPR), supported by the EPSRC grant EP/T031425/1.

Conflict of Interest

The authors declare no conflict of interest.

Open Research

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.