Non-fullerene acceptor photostability and its impact on organic solar cell lifetime

,


INTRODUCTION
In recent years, the rapid development of non-fullerene acceptors (NFAs) has led to significant improvements in the performance of organic photovoltaics, [1][2][3][4][5][6][7][8][9] with power conversion efficiencies (PCEs) reaching more than 18% for single-junction binary devices. 10In comparison to their fullerene counterparts, NFAs offer several key advantages.The ease of adaption of their chemical structures enables a high degree of tunability in their optoelectronic properties.By tuning the band gap, the absorption region can be shifted into the UV, visible, or infra-red regions to target different applications (e.g., for semi-transparent solar cells 11 or indoor light harvesting 12,13 ).In addition, through optimization of their molecular orbital energetics, high open-circuit voltages can be attained with minimal voltage losses. 14It is also foreseen that NFAs may offer simpler synthetic pathways that have the potential to reduce fabrication costs. 3,15ganic solar cells (OSCs) are achieving the efficiencies required for commercial viability.However, device stability requires further improvement to make them a competitive next-generation photovoltaic (PV) technology.5][26][27][28] Illumination under inert atmospheres can also induce a rapid loss of solar cell performance, resulting in a reduction in efficiency of $10%-30% within the first tens to hundreds of hours (often called burn-in), followed by a more gradual loss over the next thousands of hours.Because a substantial proportion of device performance can be lost through this initial rapid burn-in process, significant research efforts have been dedicated to unraveling its origins.For example, photo-induced fullerene dimerization has been identified as a major degradation mechanism for polymer:PC 61 BM OSCs, leading to loss of short-circuit current (J SC ) during illumination. 29We and others have identified the formation of disordered states, leading to increased charge traps and voltage losses upon photodegradation. 30,313][34][35] Encouragingly, the rapid development of NFAs allows the burn-in degradation of some OSC systems to be effectively addressed.For example, multiple NFA-based OSC systems were found to exhibit minimal burn-in losses, and extrapolated OSC lifetimes of more than 10 years have been demonstrated. 30,36,37espite intensive mechanistic studies into this topic, the origins of burn-in are debated.The various degradation mechanisms identified suggest multiple origins that strongly depend on the material system being investigated.However, to date, few studies have been dedicated to understanding how burn-in is influenced by the choice of donor and acceptor components of the active layer.Despite some NFA-based OSCs demonstrating minimal burn-in and good lifetimes, there remains a lack of fundamental understanding of the generality of this improved photostability to other classes of NFAs and donor polymers, especially in terms of the underlying degradation mechanisms and hence the molecular design rules for improving lifetimes of NFA-based OSCs.
To date, the IDTBR 2 and ITIC [5][6][7][8]38 families have been established among the most promising NFAs, achieving OSC efficiencies of more than 12% 39 and 14%, 40 respectively. Fo IDTBR-based NFAs, burn-in free devices have been demonstrated for some systems, 30,34,36 but the generality of this behavior to other systems requires further study.The photostability of ITIC-based acceptors in inert conditions is comparatively less clear, and it seems that several factors can affect device photostability.Some studies have reported that ITIC-based materials may be chemically incompatible with commonly used transport layers. 41,42Another study demonstrated extrapolated operational device lifetimes of up to 10 years for some ITIC derivatives when used with PBDB-T, although these devices still suffered from a small burn-in loss.37 This study also showed device stability to strongly depended on the end groups and side chains of the ITIC-based acceptors, suggesting that further stability enhancements are possible by optimization of the material design of NFAs at a molecular level. 37 Du etal. 43 also highlighted the importance of initial morphology on device stability for polymer:ITIC-4F-based devices, with performance deterioration being linked with polymer reorganization and diffusion-limited aggregation of NFAs.
Herein, we report an in-depth investigation of the photostability of OSCs based on several popular NFAs, namely, O-IDTBR, EH-IDTBR, ITIC, and ITIC-M.5][46][47] OSCs are fabricated and their photostability is systematically tested under continuous white LED illumination in an inert atmosphere.The photostability of the investigated systems is found to be highly dependent on the choice of NFA but relatively insensitive to the choice of donor polymer.Devices using NFAs from the IDTBR family suffer minimal burn-in across all investigated polymers, whereas all devices using ITIC-based NFAs show more severe burn-in and continuous long-term degradation.Using advanced characterization measurements, we find that ITIC and ITIC-M are susceptible to photo-induced chemical degradation.This leads to decreased optical absorption and increased charge trapping, resulting in poor device stability.Such behavior is not observed for O-IDTBR or EH-IDTBR, correlating with the superior stability of these devices.This work represents the first systematic study to address the generality of burn-in behavior in polymer:NFA OSCs.In addition, it highlights the importance of NFA molecular design as a key strategy to improve the photostability of fullerene-free OSCs.

Device characterization
Three donor polymers (PTB7-Th, PffBT4T-2OD, and PBDB-T) [44][45][46] and four NFAs (O-IDTBR, EH-IDTBR, ITIC, and ITIC-M) 2,5,7 are investigated in this work.The chemical structures of these materials are shown in Figure 1, and the full chemical names are provided in Note S1.All binary donor:acceptor combinations were investigated, except for PffBT4T-2OD:ITIC and PffBT4T-2OD:ITIC-M due to the poor initial performance of these systems.The details of active-layer preparation and deposition can be found in the Table S1.To monitor performance during continuous photoexcitation, devices were placed into a nitrogen-filled environmental chamber and illuminated with a 1 sun equivalent intensity white LED array (spectrum shown in Figure S1).The evolution of the photovoltaic parameters during continuous photoexcitation is shown in Figure 2.
All O-IDTBR-and EH-IDTBR-based OSCs exhibit superior photostability upon 120 h of light soaking, undergoing negligible degradation across all photovoltaic parameters.In contrast, the ITIC-and ITIC-M-based OSCs exhibit a more significant loss in device performance, losing up to $30% of their initial performance within the first 24 h of illumination.This is followed by more gradual degradation, primarily caused by a continuous loss of fill factor (FF) and J SC .Remarkably, degradation strongly depends on the choice of NFA yet is relatively insensitive to the choice of donor polymer across all investigated blends.
To investigate the origin of the strong dependence of photostability on the choice of electron acceptor, further characterization was conducted, with PTB7-Th as a common donor polymer.PTB7-Th was selected because of the good performance achieved when blended with all investigated NFAs (Table S2).Although PBDB-T:ITIC and PBDB-T:ITIC-M devices use 1,8-diiodooctane (DIO) during fabrication, the corresponding PTB7-Th devices do not.This excludes DIO-related stability issues [48][49][50][51] from being solely responsible for the poor device stability of the ITIC and ITIC-M devices.
The formation of additional sub-band tail states (also known as shallow trap states) has been previously reported during burn-in for other OSC blends. 30,31,52,53To probe the impact of prolonged continuous photoexcitation on the transport and recombination kinetics of charge carriers and tail-state configurations, transient photovoltage (TPV) and charge extraction (CE) measurements were performed.Devices were measured before and after 24 h of illumination (Figure 3), covering the burn-in period.
As shown in Figure 3A, PTB7-Th:O-IDTBR and PTB7-Th:EH-IDTBR OSCs exhibit no noticeable change in charge carrier density (measured as a function of open-circuit voltage, V OC ) upon degradation, whereas a clear increase is seen for PTB7-Th:ITIC and PTB7-Th:ITIC-M OSCs.This increased charge carrier density is assigned to the additional formation of shallow sub-band tail states, as observed in our previous studies. 30Notably, no clear change in the slope of charge carrier density is seen, suggesting negligible changes in the distribution of the sub-band tail states upon degradation for all investigated OSC systems. 52The effective drift mobility of these NFAs was measured by charge extraction methods under short-circuit conditions, 54 as shown in Figure 3B.The reduction in effective mobility of PTB7-Th:ITIC and PTB7-Th:ITIC-M OSCs upon degradation suggests an increase in hopping steps during transport.The magnitude of the decrease in effective mobility is in good agreement with the increase in charge carrier lifetime measured at open circuit (shown in Figure S2).This suggests that the additional tail states are acting as traps that must first undergo thermally activated detrapping before recombining, thereby leading to the observed increase in lifetime and decrease in mobility.In contrast, no noticeable change in effective mobility is seen for PTB7-Th:O-IDTBR and PTB7-Th:EH-IDTBR OSCs, consistent with their excellent photostability.

Optical studies
To investigate photo-induced changes to the optical properties of the investigated materials, UV-visible spectra of blend and neat films were measured.To probe the intrinsic photostability of the donor and acceptor components, we focused on the investigation of neat films, as shown in Figure 4.All prolonged photoexcitation was carried out under 1 sun equivalent intensity white LED illumination inside a nitrogen-filled glovebox with oxygen and moisture levels below 0.5 ppm.UV-visible spectra of fresh films were measured and subsequently repeated after 1 and 7 days of continuous photoexcitation, corresponding to the burn-in period and beyond.
After 7 days of continuous photoexcitation, no observable change was seen in the absorbance of PTB7-Th (Figure 4A), indicating that the donor polymer was not significantly contributing to the burn-in and subsequent degradation of the OSCs.Both O-IDTBR and EH-IDTBR exhibited excellent photochemical stability with negligible degradation upon photoaging (Figures 4B and 4C), consistent with the outstanding stability of O-IDTBR and EH-IDTBR OSCs.In contrast, a noticeable loss of absorption was observed for both ITIC and ITIC-M after just one day of continuous photoexcitation (Figures 4D and 4E).This continued with increased photoaging time.A slight increase in the absorbance between 430 and 550 nm was also observed (Figure S3).These observations indicate bleaching of the chromophores and disruption/shortening of the conjugation length of ITIC and ITIC-M upon continuous photoexcitation.Such changes in optical properties are in agreement with the observed losses of J SC and efficiency of ITIC and ITIC-M-based OSCs during the burn-in period and beyond.This bleaching is similar to the bleaching caused by photo-induced dimerization of fullerenes, which also results in a rapid loss in J SC of fullerene-based OSCs, 29 and photobleaching of IDFBR, which was found to originate from fragmentation during continuous illumination in both air and nitrogen. 26igure 4 focuses on neat donor and acceptor films, but similar behavior is observed in blend films, although the rate of photodegradation is reduced (Figure S4).The slower degradation in the blend is a result of the overlapping absorption of the donor and acceptor, which means that the acceptor absorbs relatively less light compared with the neat films.In addition, if the degradation mechanism goes via the excited state, as reported previously for other OSC materials, 26,55,56 it would be suppressed in the blend due to charge-transfer (CT) quenching the excited state.

Structural investigations
To identify the molecular origins of such optical property changes, resonant molecular vibrational Raman spectroscopy was used.Raman spectroscopy has been previously used to study the degradation of organic photovoltaic materials, providing key information about photo-induced chemical and conformational changes and their origins. 26,57As shown in Figures 5A-5C, there was no significant change in the Raman spectra of PTB7-Th, O-IDTBR, or EH-IDTBR upon 7 days of photoaging.This corroborates our previous findings of the good photochemical stability of these acceptors when degraded in situ under nitrogen. 26In contrast, both ITIC and ITIC-M exhibit significant changes in their Raman spectra upon photoaging (Figures 5D  and 5E).This is indicative of more severe photochemical degradation of ITIC and ITIC-M compared with O-IDTBR and EH-IDTBR and is in agreement with the UVvisible spectral changes and device stability measurements.Although some subtle changes are observed after 1 day of photoaging for ITIC and ITIC-M (Figures S5  and S6), these changes become more evident with increasing photoaging time, indicating that the associated degradation processes continue beyond the burn-in period.Raman spectra of ITIC and ITIC-M up to 2,250 cm À1 are shown in Figures S7 and S8.
Overall, there is an increase in Raman intensity for both ITIC and ITIC-M when probed at both 457 and 514 nm.This indicates the formation of a new, widerband-gap, Raman active-degradation product that is resonant at these wavelengths, in agreement with the slight increase in the absorption of degraded molecules at these wavelengths (Figure S3).To identify the changes in specific molecular vibrations upon prolonged photoexcitation and understand the nature of this degradation product, the difference spectra between the fresh and the degraded films were extracted and are shown in Figures 5F and 5G.These difference spectra highlight the peaks associated with the degradation products whose intensities increase because of the resonant effect of the degradation product under 457 and 514 nm excitation.Both ITIC and ITIC-M show the same changes in molecular vibrations, signifying the same degradation process is occurring in both molecules.The nature

of these peaks was identified using density functional theory (DFT) calculations (Figures S9 and S10
).There are some important differences between fresh and degraded ITIC: the alkene peak at 1,550 cm À1 and the thiophene peak at 1,425 cm À1 are significantly quenched in the degradation product, along with the nitrile peak at 2,220 cm À1 (shown in Figure S7).In addition, a new peak at 1,580 cm À1 , possibly a new alkene mode, is seen to grow in.The peak at 1,455 cm À1 appears to gain intensity and shift to higher frequencies; this shift to higher frequency is seen for other peaks, including the core phenyl peak at 1,600 cm À1 .These peak changes are also seen when probed at 457 nm (Figure S11).These Raman peak changes indicate important chemical and structural changes of the NFAs upon prolonged photoexcitation.To demonstrate these peak changes more clearly, we conducted in situ accelerated photodegradation of ITIC in a nitrogen-filled chamber.In situ degradation allows us to track peak changes continuously and hence determine the exact nature of molecular structure changes upon  photoexcitation. 20,26It also provides photodegradation accelerated by high-intensity laser irradiation, which allows us to ensure reproducibility on a shorter timescale and provides a more extreme situation of degradation, which assists with interpretation.The baselined and normalized Raman spectra at increasing degradation times are shown in Figure 6A.There is an increase in Raman intensity and photoluminescence (PL) background across the spectrum (Figure S12), similar to the preceding 7 day photoaged spectra.All changes observed after 7 days of photoaging are observed after in situ laser degradation, with some additional changes becoming more apparent: the thiophene C=C peak at 1,250 cm À1 is quenched and shifts to lower frequencies, and the carbonyl peak at 1,705 cm À1 shows a new high-frequency shoulder growing in.If we take the difference spectra at early degradation times, we obtain a spectrum similar to that shown in Figure 5F, indicating the same degradation process is occurring both in prolonged photoaging under white LED illumination and during laser degradation (Figure S12C).
To understand the origin of these changes and their implication to photostability of ITIC, we compare our experimental data to simulated Raman spectra.First, there is an overall good match between the experimental and the simulated spectra.Four regions of vibrations show higher relative intensities in the measured spectra compared with the simulated spectra, but all are present in both (Figure 6B).These regions are the alkene peak at 1,550 cm À1 , which we ascribe to the vinylene linkage; the carbonyl peak at 1,705 cm À1 (simulated peak highlighted in Figure S10); the thiophene region around 1,450 cm À1 ; and the nitrile peak at 2,220 cm À1 (shown in Figure S7).One of these peaks, the alkene at 1,550 cm À1 , shows a large reduction in relative peak intensity upon degradation.When diluted films of ITIC are fabricated by blending with polystyrene, this peak is reduced in relative peak intensity, demonstrating the alkene's sensitivity to intermolecular packing (Figure S13).Upon degradation, the decrease in relative intensity of this peak indicates disruption of ITIC packing.
To understand the other peak changes upon degradation, we explored several degradation products.One proposed mechanism that correlates well with the changes in the measured Raman spectra, alongside the observed photobleaching, is that of photoinduced conformational change affecting the end groups of the molecules.To verify this conformational instability of ITIC as the main cause for its photoaging, we simulated the Raman changes of ITIC with an increasing dihedral angle between the core and the end groups, rotating about the single bond on the vinylene linkage (Figure 6C, dihedral labeled in Figure 6E).As the molecule is made less planar, the carbonyl and alkene peaks shift to higher frequency, accounting for the new high-frequency peaks observed experimentally.Alongside these new peaks, there is a shift to lower frequencies and quenching of the peak at 1,250 cm À1 , quenching of the shoulder of the 1,400 cm À1 peak, and a slight shift of the peak at 1,450 cm À1 , as observed experimentally.This agreement between simulated and experimentally measured spectra confirms that the end groups are responsible for photoaging.Such conformational changes of ITIC from a planar to a twisted structure will shorten the effective conjugation length of the molecule.The decrease in absorption at low energy and the slight increase in absorption at higher energies might reflect such changes (Figures 4 and S3).
Our previous studies show that conformational change can lead to bond breaking, 26 which can also cause shortening of the effective conjugation length.To probe this, we simulated Raman changes when the vinylene linkage is broken and end groups are lost (the simulated core spectrum shown in Figure 6D).The peaks at 1,420 and 1,600 cm À1 show larger shifts to higher frequency than those induced by conformational change, which is again consistent with the experimentally observed changes.In addition, the peaks corresponding to the end groups (i.e., the alkene, carbonyl, and nitrile) are quenched.Therefore, we can draw the clear conclusion that there is a reduction in the effective conjugation length.This loss of conjugation of ITIC and ITIC-M molecules originates from the vinylene linkage between the electrondonating core and the electron-deficient end group, consistent with the reported chemically unstable nature of this linkage. 28,41,42Importantly, we find that even in an inert atmosphere, this unstable vinylene linkage, which is particularly sensitive to intermolecular packing, is vulnerable to conformational change.Such conformational change will eventually lead to irreversible bond breaking, shortening the effective conjugation breaking of ITIC and ITIC-M.It appears that modification of the substituents on the end group, i.e., by methylation, which affects intermolecular packing or electron-withdrawing strength, 7 influences the stability of this interunit region (depicted in Figure S5).This is in agreement with previous studies that indicated methylation of ITIC leads to poorer stability. 37man spectra were also measured for blend films (Figure S14).In the case of blend films, no obvious changes were observed after 7 days of photoaging.This is not too surprising, given the slower rate of degradation within the blends compared with neat films, as evidenced by UV-visible spectroscopy (Figures 4 and S4).For neat films, it was shown that in situ degradation using the Raman excitation laser as the light source caused the same degradation as observed after prolonged photoexcitation under the white LED source (Figure S12), albeit at a faster rate.When blend films were degraded in situ with the Raman excitation laser in a nitrogen environment, the same peak changes were observed in the blend as were observed for the neat films, demonstrating that the same degradation process is occurring in the blend, despite the slower rate of degradation (Figure S15).
Luminescence studies PL and electroluminescence (EL) measurements were also performed on the PTB7-Th:NFA devices before and after 24 h of continuous photoexcitation, covering the burn-in period (Figure 7).PL and EL emission from OSCs can be affected by numerous factors.For example, PL is sensitive to changes that affect the absorption properties of the photoactive materials, exciton diffusion behavior, and exciton dissociation, whereas EL can be affected by changes in the energetics of various layers, carrier mobility, and trapping behavior.All of these properties are in turn sensitive to the photoactive-layer morphology, meaning it is not always straightforward to determine the responsible factor that leads to changes in PL and EL spectra; nevertheless, the techniques can provide complimentary evidence alongside other experimental techniques and may help to suggest potential degradation pathways.
For both PTB7-Th:O-IDTBR and PTB7-Th:EH-IDTBR OSCs, no significant change (less than 10% increase in intensity) in PL emission was observed after 24 h of continuous photoexcitation.Combined with the negligible changes in EL emission of these devices, this provides further evidence that the constituent materials are stable.Due to the sensitive nature of these techniques to small changes in morphology and transport behavior, these data also suggests that there are no obvious changes to the morphology of the photoactive layer during the initial burn-in period.This is consistent with the observed burn-in-free, excellent device stability of these blends and agrees well with previous reports by others of the good morphological stability of the PTB7-Th:EH-IDTBR blend. 34r the PTB7-Th:ITIC device, although there may be a slight increase in PL emission, changes of this kind are relatively small.A larger 3.5-fold increase was observed for the PTB7-Th:ITIC-M device after this period of continuous photoexcitation.The slight difference in the shape of the fresh PTB7-Th:ITIC-M device PL peak results from its low emission, which is slightly distorted by the small amount of emission from the encapsulation glue (glue emission shown in Figure S16).Comparison of the blend emission with neat material PL (Figure S17) indicated that the observed blend PL emission originates from donor and/or acceptor excitons, rather than from CT-state emission.This means that after a period of prolonged photoexcitation, there is an increase in the number of photogenerated excitons that proceed to radiatively recombine within the PTB7-Th:ITIC-M device.Photoluminescence and electroluminescence spectra of PTB7-Th:NFA devices before and after extended photoexcitation (A and B) Photoluminescence (A) and electroluminescence (B) spectra of PTB7-Th:NFA devices before and after 24 h of continuous photoexcitation under the same conditions as were used during stability measurements.The slight difference in the shape of the fresh PTB7-Th:ITIC-M device PL peak results from its low emission, which is slightly distorted by the small amount of emission from the encapsulation glue (glue emission shown in Figure S16).
Several factors may cause such an increase in PL emission.For example, the observed changes in intermolecular interactions and conformational changes of ITIC and ITIC-M outlined previously may inhibit exciton diffusion and/or dissociation, leading to an increase in the number of excitons that radiatively recombine.9][60][61] Morphological changes have been widely reported to be responsible for burn-in degradation for several OSC blends, including demixing of amorphous regions for both polymer:fullerene 32,33,62 and polymer:NFA 34,63,64 devices, as well as crystallization and diffusion-limited aggregation of NFAs. 42,43,62,63Recently, transmission electron microscopy (TEM) imaging of PTB7-Th:ITIC films before and after photoaging under similar conditions to those used in this work suggested that demixing likely occurs in this blend linked with poor miscibility between donor and acceptor. 34Increases in PL emission, as observed here, were also reported and may be explained by the coarsening of domains, leading to a reduced probability of exciton dissociation. 34To probe the possibility of large-scale phase segregation, atomic force microscopy (AFM) measurements were carried out on PTB7-Th:ITIC and PTB7-Th:ITIC-M films before and after 7 days of continuous photoexcitation with a 1 sun equivalent intensity white LED source in a nitrogen environment (Figure S18).No clear changes in the AFM images were observed after 7 days of photoaging.This suggests that there is no obvious large-scale phase separation.
The conformational changes of ITIC and ITIC-M may explain the increase in PL emission upon photodegradation of the blend.Although exciton emission from the neat ITIC and ITIC-M decreases upon photodegradation (Figure S19), the resulting disruption to packing and energetics, as a result of conformational changes, can cause an increase in PL of the blend by hindering exciton dissociation, resulting in more excitonic emission.We also cannot rule out more subtle morphological changes.For example, ITIC has also been observed to reorientate from more face-on to more edge-on orientation during photoexcitation under nitrogen. 37dditionally, diffusion-limited NFA aggregation, leading to the formation of isolated acceptor domains, has also been reported for ITIC-4F. 43Such changes in morphology could also contribute to an increase in PL emission and may act alongside the photo-induced chemical degradation reported herein.
Contrasting the PTB7-Th:O-IDTBR and PTB7-Th:EH-IDTBR devices, a $50% reduction in EL emission is observed for PTB7-Th:ITIC and PTB7-Th:ITIC-M devices.Given the negligible change in series resistance after extended photoexcitation (Figure S20), the injection of current will cause a similar voltage drop for both fresh and photoaged devices.Therefore, the reduction in emission implies the recombination in the photoactive layer is less radiative.This behavior is consistent with our previously discussed results.This decrease in EL emission can be explained by the observed conformational changes to ITIC and ITIC-M.These changes result in disruption to intermolecular packing, intra-and intermolecular charge transport, and increased energetic disorder and trap formation, thereby causing a reduction in EL emission and correlating well with device stability.The EL emission of the neat materials is shown in Figure S21 for reference.

DISCUSSION
Photostability is a widely recognized challenge for OSCs, and a detailed understanding of the degradation mechanisms is required to enable the design of high-performance stable OSCs.Our study addresses the photostability of NFA-based OSCs and investigates the generality of behavior across a range of donor polymers and electron acceptors.We demonstrate that rational design and choice of suitable electron acceptors, instead of donor polymers, might be a more effective strategy to achieve long-term photostability of OSCs.Through a range of characterization techniques, we show that photochemical stability of NFAs must be addressed for good device stability to be achieved.Due to combined excellent photochemical and likely morphological stability, 34 IDTBR-based OSCs can achieve good photostability with minimal burn-in losses. 30,36Unfortunately the maximum efficiencies achieved by IDTBR-based acceptors are slightly lower compared with ITIC-based materials.However, despite good device performance, both ITIC and ITIC-M suffer from more severe photochemical degradation of their molecular structure during continuous photoexcitation in inert atmospheres, leading to poor device stability.
We find that ITIC and ITIC-M degrade in the same way during continuous photoexcitation.We observe several key changes in the Raman spectra that, alongside simulated Raman spectra, allow us to explain the location of this photodegradation, namely, the interunit region, and propose the mechanism behind it.An initial conformational change-specifically, twisting of the end group (the single bond that connects the core to the end groups)-is observed, which can lead to the more extreme situation in which the vinylene linkage breaks.Gas-phase single-molecule DFT simulations of these proposed alterations show similar outcomes in both situations and agree with the changes observed experimentally.Such changes would reduce the effective conjugation length of the molecule, in agreement with the decrease in the main UV-visible absorption peaks of ITIC and ITIC-M upon photoaging (Figures 4D and 4E), alongside a slight increase in higher-energy absorption (Figure S3).These changes would also affect molecular packing, which is observed by the decrease in the relative intensity of the Raman peaks sensitive to intermolecular packing.The methylation of the end group, which affects intermolecular packing or electron-withdrawing strength, 7 may also affect the stability of this interunit region, with ITIC-M appearing to suffer from more rapid degradation compared with ITIC, as shown by the larger changes in both UV-visible (Figure 4) and PL (Figure S19) spectra of neat materials after a period of continuous photoaging.These changes are reflected in the overall Raman intensity increase (Figures 5D, 5E, and S5) associated with the slight increase in absorption at the probing wavelengths (Figure S3), which is indicative of the formation of a degradation product with a wider band gap.In agreement with previous reports, 37 this suggests that the strategy of methylation to improve the performance of ITIC may compromise the photostability of ITIC-based OSCs, although intermolecular interactions and different morphologies in the blend may also affect the rate of photodegradation in devices.
We consider the potential impact of the different 3D packing structures observed for IDTBR and ITIC acceptors on the photo-induced conformational changes.Despite p-stacking distances being similar in their crystal structures, 65,66 the difference in side chains and the resulting stabilizing interactions between them can influence morphological and conformational stability.The flexible aliphatic side chains of IDTBR can pack tightly, 66 maximizing and stabilizing dispersive interactions, whereas the bulky phenyl-hexyl side chains in ITIC are not as conducive to such stable packing.In fact, ITIC has been shown to pack differently in its crystal structure compared with spin-coated thin films, which tend to be more disordered because of these bulky side chains. 67,68These side chains were designed specifically to reduce molecular planarity and inhibit self-aggregation, which can be detrimental to performance. 5However, this disorder and comparatively poorer dispersive interactions of the hexyl-phenyl side chains could make ITIC more susceptible to morphological and conformational changes, leading to poor device stability.Elsewhere, it has been observed that the inclusion of thiophene in the side chains (ITIC-Th) improves crystallinity and packing, 47 which in turn correlates with the observed improved stability of ITIC-Th devices, highlighting the important role that side chain engineering can have with regards to NFA stability. 37thin PTB7-Th:ITIC and PTB7-Th:ITIC-M devices, we also observe a distinctly larger increase in PL emission upon photoaging compared with PTB7-Th:O-IDTBR and PTB7-Th:EH-IDTBR devices, and the increase was especially large for ITIC-M.This PL emission originates from donor and/or acceptor excitons, rather than CT-state emission.Such behavior is explained by reduced rates of exciton dissociation.The conformational change of ITIC and ITIC-M will affect both intra-and intermolecular interactions in blends, in turn affecting associated photophysical processes such as charge generation and transport.The disrupted interactions would hinder exciton dissociation in blends, leading to more excitonic emission of the neat materials, as observed in our work.This result highlights the important interplay between molecular conformational stability and molecular-scale morphological stability, in which initial conformational change can disrupt molecular interactions.
Although no macro-scale morphological changes were observed after photodegradation, we cannot exclude the possibility of other morphological instabilities contributing to these observed PL changes, especially with previous reports highlighting the tendency of crystallization and small-scale aggregation of ITIC-based acceptors during aging 43,47 and that the poor miscibility between PTB7-Th and ITIC can lead to demixing of intermixed domains. 34Such changes may occur in parallel with the photo-induced chemical degradation reported herein, but these changes are not large enough to be probed by AFM measurements.Therefore, our results identify the molecular-scale conformational and morphological changes as critical origins of the photostability of NFAs and NFA-based OSCs.These observations are consistent with the differing stabilities of the investigated devices.Specifically, O-IDTBR and EH-IDTBR are more resistant to photochemical degradation upon continuous photoexcitation in inert atmospheres.Within PTB7-Th devices, the minimal changes in device performance, PL or EL emission, and TPV/CE additionally suggests that the blend morphologies are relatively stable, likely partly because of the stabilizing interactions between the side chains of these acceptors.Consequentially, no obvious changes in charge transport and trapping behavior are observed for these devices, and initial good performance is maintained during continuous illumination.
However, ITIC and ITIC-M appear to suffer photochemical degradation upon continuous illumination in inert atmospheres, associated with twisting about the single bond on the vinylene linkage and possible complete breaking of this link between the core and the end group.These changes reduce the effective conjugation length of the molecules, reducing their absorption and likely affecting packing and intraand intermolecular charge transport.This, in combination with morphological changes that may occur, can explain the poor device stability and increased charge trapping.
When investigating the differences between ITIC and IDTBR, we noted from quantum chemical simulations (Figure S22) that the charge distribution within ITIC shows a strong intramolecular charge-transfer (ICT) character with a strong dipole moment between the electron-rich core and the electron-withdrawing end groups; however, IDTBR shows a weaker charge distribution.Although this strong ICT character may improve charge generation, 69 its impact on photochemical stability is unclear and requires further investigation.
Our study highlights the importance of rational NFA molecular design to improve OSC stability.It is necessary for these design rules to complement those for efficiency optimization in the early stages of material design.We have shown that some of the most popular NFAs lack good stability due to their susceptibility to photochemical degradation during illumination, even in inert conditions.In addition, because of the strong dependence of degradation behavior on the choice of NFA and relative independence of donor polymer within our investigated systems, we consider new molecular design of NFAs as a key strategy to improve the stability of OSCs.We propose the following design rules for improved NFA stability.Primarily, we draw attention to the vinylene linkage between the donor core and the acceptor units within ITIC, which we identified as the main point of degradation.Interestingly, the vinylene linkage between electron-withdrawing benzothiadiazole (BT) and rhodanine groups of IDTBR appears to be less prone to degradation.Therefore, the real weak point may be better described as the interunit region between the donor core and the acceptor components about which rotation can occur.This links with our previous studies on IDTBR, where we show that the core-BT linkage is susceptible to degradation, 26 although it is necessary to explore the generality of this further.To stabilize this interunit region, strong conformational lockers could be applied to resist the photoinduced conformational changes that lead to degradation, such as in the chemically similar Y6 NFA, although the impact on device performance in Y6 has yet to be established. 53In addition, we suggest that to enable stable and resilient packing, bulky side chains should be avoided in favor of alkyl side chains or alternatives with heteroatoms that improve intermolecular interactions.However, we acknowledge that a careful balance may be required to prevent excessive self-aggregation, which can be problematic for highly planar molecules. 5End-group substitution is another important consideration, and we show that end-group methylation may need to be avoided, although more work is needed to understand the role of end-group methylation on stability.Others have suggested that end-group fluorination, rather than methylation, could be an effective strategy to improve stability. 3762][63][64]

Data and code availability
All data reported in this paper will be shared by the lead contact upon request.This paper does not report original code.Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Device fabrication
Devices were fabricated with the inverted architecture (indium tin oxide [ITO]/ZnO/ active layer/MoO 3 /Ag).Pre-coated and patterned ITO-coated glass substrates were cleaned by ultrasonication in Hellmanex III (2% by volume in deionized water), deionized water, acetone, and isopropanol.After cleaning, substrates were with an air gun.Dried substrates were treated with an oxygen plasma.219.5 mg of zinc acetate dihydrate was dissolved in 2 mL of 2-methoxyethanol and 60.4 mL of ethanolamine overnight at room temperature and was then spin-coated in air at 4,000 rpm onto the substrates and annealed at 150 C for 10 min to form a $20 nm ZnO layer.The active layer was then deposited by spin coating in a nitrogen-filled glovebox with oxygen and moisture levels below 0.1 ppm.The details of active-layer preparation and deposition can be found in the Table S1.Subsequently, a 10 nm layer of MoO 3 and 100 nm layer of silver were deposited via thermal evaporation through a shadow mask as the hole-transporting layer and top electrode, respectively.Devices were encapsulated inside a nitrogen atmosphere with a UV-curable epoxy and a glass cover slide.

Stability testing
Encapsulated devices were placed into an environmental chamber with a glass front for performance monitoring under continuous illumination.The chamber was purged with nitrogen for at least 30 min before illumination, and positive pressure of nitrogen was maintained by passing an uninterrupted flow of nitrogen through the chamber for the duration of the measurements.A white LED array was used as the light source, with the intensity of the LED array and device position adjusted such that the J SC , and hence charge density within each device, was approximately equal to that measured under 1 sun AM1.5G illumination.The spectrum of the LED array is shown in Figure S1.The environment temperature was kept below 30 C by a water-cooling system.During stability measurements, current-voltage responses were measured at least once per hour, and devices were kept under open-circuit conditions between these measurements.The stability test for the PffBT4T-2OD devices was paused around the 92-h mark to reconnect one of the devices that had a poor electrical connection.The devices were kept in the dark during and after reconnection until the chamber had been repurged with nitrogen, after which the LED array was switched back on and the measurements were resumed.The light intensity may have differed by a small amount after reconnection because of slight repositioning of the devices.

Transient photovoltage and charge extraction
Charge extraction measurements were used to determine the average charge carrier densities in devices under different illumination levels and different biases (open circuit and short circuit in this study).For each device, the desired light intensity, and consequently the initial bias, was provided by a ring of 12 white LEDs capable of up to 5 sun equivalent illumination.After illumination at the desired intensity, the LEDs were switched off and the device was switched to short circuit.The transient voltage was then acquired with a DAQ card connected to a Tektronix TDS3032B oscilloscope.The voltage transients were converted into current transients through Ohm's law.Then the current transients were integrated to obtain the total charge, Q, which was used to calculate the carrier density, n, in the device.During TPV measurements, the devices were held at open-circuit conditions.The same LED source was used for illumination, and a small optical excitation was provided by a pulsed 532 nm Continuum Minilite Nd:YAG laser with a pulse length of less than 10 ns.This small excitation produced a small voltage transient decay, which was measured on the oscilloscope.The decay was fitted with a mono-exponential to obtain the small perturbation carrier lifetime, which was used to estimate the total charge carrier lifetime within the device.

Film studies
An identical white LED array that was used for the stability tests was also used for the continuous illumination of thin films for all film-based characterization.For film studies, all continuous photoexcitation was performed inside a nitrogen-filled glovebox with oxygen and moisture levels below 0.5 ppm.fans circulated nitrogen over the samples to keep the environmental temperature below 30 C.

UV-visible spectroscopy
A Perkin Elmer Lambda 750 spectrophotometer with an integrating sphere attachment was used for all UV-visible spectroscopy.A bare glass substrate was used for measurement of the reference spectrum.

Raman spectroscopy
A Renishaw in Via Raman microscope in a backscattering configuration with a 503 objective was used to collect Raman spectra.All measurements were conducted in a nitrogen-purged Linkam sample chamber.All measurements were taken with a defocused laser spot with a radius of $10 mm.Raman spectra were collected at various wavelengths using an argon ion laser (457, 488, and 514 nm).Acquisition times and laser powers were optimized to give the best spectra but were kept consistent between samples that are directly compared in the text.Spectrometer calibration was conducted using a silicon reference sample, and background PL was subtracted using a polynomial fit.Accelerated in situ degradation was carried out with a laser spot of 10 mm and approximate power density of 3.2 3 10 6 W m À2 , giving an acceleration factor of $3,0003 compared with AM1.5G solar illumination.

Density functional theory calculations
Density functional theory simulations were carried out using Gaussian 09 software on the Imperial College High-Performance Computing Service. 702][73][74] For frequency calculations, alkyl side chains were simplified to methyl groups to reduce computational time.Structures were optimized to a local minimum energy conformation, with frozen dihedral angles used to simulate molecular conformational changes.Calculated frequencies were corrected using an empirical factor of 0.97 for the frequency of vibration. 75Electrostatic potential distribution was calculated using the Merz-Kollman model with full side chains. 76otoluminescence and electroluminescence spectroscopy For luminescence spectroscopy, a custom-built setup was used.For photoluminescence measurements, a 405 nm laser was used as the excitation source.However, for electroluminescence measurements, a Keithley 2400 was used to apply the voltage to the devices and monitor the current.A current density of 200 mA cm À2 was applied to the cells, which is sufficiently low to prevent device damage.For both electroluminescence and photoluminescence, an AvaSpec-ULS2048x64 spectrometer was used to detect the emission.A 550 nm long-pass cutoff filter was placed between the sample and the spectrometer to remove the excitation light for photoluminescence measurements.

Figure 1 .
Figure 1.Schematic representation of the chemical structures of the donor polymers and non-fullerene acceptors used in this work Chemical structures of the studied donor polymers (PTB7-Th, PBDB-T, and PffBT4T-2OD) and non-fullerene acceptors (O-IDTBR, EH-IDTBR, ITIC, and ITIC-M).

Figure 2 .
Figure 2. Photostability of investigated devices (A-D) Normalized PCE (A), V OC (B), J SC (C), and FF (D) of investigated devices during continuous photoexcitation under 1 sun equivalent intensity white LED irradiation.Devices were kept in a nitrogen atmosphere for the duration of the test.

Figure 3 .
Figure 3. Energetics and effective charge carrier mobilities of PTB7-Th:NFA devices (A) Measurements of the accumulated charge density at open-circuit voltage as a function of illumination intensities, determined by charge extraction.(B) Effective drift mobility as a function of charge carrier density determined by charge extraction measured at the short-circuit current.Devices labeled ''aged'' were exposed to 24 h of continuous illumination under the same conditions as the stability measurements.

Figure 4 .
Figure 4. UV-visible spectra of donor and acceptor films before and after extended photoexcitation PTB7-Th (A), O-IDTBR (B), EH-IDTBR (C), ITIC (D), and ITIC-M (E).Continuous photoexcitation was performed inside a nitrogen-filled glovebox, with a 1 sun equivalent intensity white LED array as the illumination source.All spectra are normalized to the fresh spectrum of the corresponding material.For PTB7-Th, O-IDTBR, and EH-IDTBR, all spectra are overlapping, because there is no change in absorbance during continuous photoexcitation.

Figure 5 .
Figure 5. Raman spectra of neat donor and acceptor films before and after extended photoexcitation (A-C) Normalized Raman spectra of PTB7-Th (A), O-IDTBR (B), and EH-IDTBR (C) films before and after 7 days of continuous photoexcitation.(D and E) Raman spectra of ITIC (D) and ITIC-M (E) films before and after 7 days of continuous photoexcitation.An additional peak at 2,220 cm À1 is shown in Figures S7 and S8.(F) Normalized difference spectra of the fresh and 7 day aged films for both ITIC and ITIC-M.(G) Normalized difference and fresh spectra of ITIC.All spectra were measured under a nitrogen atmosphere with a 514 nm excitation laser.

Figure 6 .
Figure 6.In situ Raman spectra of ITIC during laser degradation and simulated Raman spectra (A) Baselined and normalized in situ Raman spectra taken at increasing laser degradation times under a nitrogen flow at 514 nm excitation.Arrows show the main peak changes upon degradation.(B) Simulated Raman spectra of ITIC in its optimized lowest-energy geometry (Opt) versus experimental fresh ITIC Raman spectra.The main alkene peak is marked with an asterisk.

Figure 7 .
Figure 7. Photoluminescence and electroluminescence spectra of PTB7-Th:NFA devices before and after extended photoexcitation (A and B) Photoluminescence (A) and electroluminescence (B) spectra of PTB7-Th:NFA devices before and after 24 h of continuous photoexcitation under the same conditions as were used during stability measurements.The slight difference in the shape of the fresh PTB7-Th:ITIC-M device PL peak results from its low emission, which is slightly distorted by the small amount of emission from the encapsulation glue (glue emission shown in FigureS16).