Rationale for highly eﬃcient and outdoor-stable terpolymer solar cells †

,


Introduction
The development of non-fullerene acceptors (NFAs), especially those of the ITIC family 1-3 and the Y-series, 4,5 which combine tunable energy levels and near-infrared (NIR) absorption has triggered a rapid advance in the performance of organic photovoltaics (OPV), endowing OPV to arise as a promising contender for 3rd generation photovoltaic technologies. [6][7][8][9][10][11][12] Meanwhile, polymer donors have continued to play a crucial role since, when paired with NFAs, the polymer donor:NFA (D:A) bulk hetero-junction (BHJ) requires a balance between phaseseparation and domain crystallinity to ensure efficient charge generation and collection, resulting in high power conversion efficiencies (PCEs). [13][14][15][16][17][18][19][20][21] However, commercialization of OPV requires not only high efficiency, but also long-term (outdoor) stability, specifically in typical PV operating conditions such as continuous illumination, heat, and day-to-night temperature changes. 22,23 NFAs are prone to aggregate when excess energy such as light and heat is supplied. 24,25 Importantly, the stability of polymer:NFA OPV is predominantly governed by the polymer donor:NFA BHJ framework. 9,26,27 In this regard, advancing NFAs matched with (photo)stable polymer donors is of utmost importance to improve the competitiveness and facilitate the commercialization of OPV.
Random terpolymerization is an emerging strategy to tailor polymer donors toward enhanced photovoltaic performance and long-term stability. 28,29 Improvements originate from the incorporation of a second donor or acceptor moiety, which modifies the polymer structure from a D-A type copolymer to a D 1 -A-D 2 -A or D-A 1 -D-A 2 type terpolymer. [30][31][32][33][34][35][36][37] The second donor (or acceptor) moiety influences the optical absorption, energetics, and blend morphology. 30 Yang and co-workers demonstrated that incorporating as little as 5% of a second donor moiety thieno [3,2-b]thiophene (TT) into the polymer PBDB-T increases the absorption coefficient and optimizes the thin film morphology. This enhanced both short-circuit current density (J SC ) and fill factor (FF) in both PC 71 BM and m-ITIC-based terpolymer donor devices. However, the introduction of excessive TT has been detrimental to the photovoltaic performance of both PC 71 BM and m-ITIC-based terpolymer donor devices.
To unleash the full potential of terpolymer photovoltaics, improving the J SC and FF only is not sufficient, since the opencircuit voltage (V OC ) needs to be maximized as well. Yang et al. reported that incorporating 5% of a second electron-deficient building block (iBDD-Si) into the donor polymer PM6 lowers the energy level of the highest occupied molecular orbital (HOMO), increases the absorption coefficient, and optimizes the molecular packing, thereby enhancing the device PCE from 15.15% to 17.01% due to simultaneously improved V OC , J SC , and FF. 32 In addition to PCE enhancements, random terpolymerization can also improve the stability of solar cells. Kettle and co-workers designed and synthesized a terpolymer (P3HT:PTB7-Th) system, namely, P1. 37 Devices based on P1:PCBM exhibited remarkably improved photostability compared to P3HT:PCBM and PTB7-Th:PCBM devices, indicating that the random terpolymerization strategy has the potential to deliver more stable OPV. However, clear design rules for terpolymer photovoltaics that lead to enhanced PCE and stability, as well as the precise mechanism leading to the efficiency enhancement are unclear at present.
Herein, we summarized the representative best-performing terpolymer solar cells in the past 5 years (Table S1, ESI †) and identified key rules for highly efficient and outdoor-stable terpolymer photovoltaics. Besides, we elucidate these key rules by demonstrating two series of terpolymer photovoltaic devices based on the parent donor polymers D18 and PM6. These rules are: (i) introducing a strong electron-deficient or electron-rich moiety that broadens absorption and down-shifts the HOMO level, enhancing the J SC and reducing energy losses; (ii) using small contents of the third component to maintaining the molecular orientation and exciton diffusion length while optimizing phase-separation and domain crystallinity to reduce the bimolecular carrier recombination, thereby enhancing the FF; (iii) manipulating the D/A miscibility by introducing a third comonomer to obtain a more thermodynamically stable thin film morphology, in turn leading to improved outdoor stability. To elucidate the key rules, we selected the D18 and PM6-based copolymers with 5% strong electron-deficient moiety of estersubstituted thiazole (E-TZ) and found that the third component can effectively blue-shifted the absorption, down-shifted the HOMO level, and fine-tuned the morphology, in turn leading to a reduced bimolecular carrier recombination and enhanced J SC , V OC , and FF. Furthermore, the introduction of 5% TZ (the abbreviation of E-TZ) improved the morphological stability, resulting in improved outdoor operating stability of D18-based terpolymer photovoltaics under the harsh local outdoor conditions at KAUST in Saudi Arabia (highest/lowest air temperature during daytime/nighttime: 48 1C/22 1C). Our work unveils the working principle of terpolymer photovoltaics and provides a rationale to explain and achieve highly efficient and outdoorstable terpolymer photovoltaics.
The normalized UV-Vis absorption and steady-state photoluminescence (PL) spectra are shown in Fig. 1b. The introduction of the second acceptor moiety (TZ) into the polymer backbone slightly blue-shifts both the UV-Vis absorption and PL spectra. More specifically, when increasing the content of TZ from 5% to 50% in D18, the optical bandgaps of the terpolymers increase from 2.01 eV (D18(5%TZ)) to 2.03 eV (D18(50%TZ)), as estimated at the intersection of the normalized UV-Vis absorption and PL spectra of the neat polymer and terpolymer films (Fig. S4, ESI †).
The HOMO levels of D18, the D18-based terpolymers, and NFA used in this work were determined by photo-electron spectroscopy in air (PESA). Unlike cyclic voltammetry (CV) obtain the energy level from oxidation and reduction potential of materials, often in solution, PESA can quantify the HOMO level on thin-film surface. 38,39 The introduction of TZ gradually lowers the HOMO levels of the D18-based terpolymers due to the high electron-deficiency of TZ. 35,36 The lowest unoccupied molecular orbital (LUMO) levels were approximated based on the measured HOMO levels and optical bandgaps (from the intersection of normalized absorption and PL) of the corresponding materials (Fig. S7a, ESI †). To confirm the generality of broadened absorption spectra and deepened HOMO levels caused by the introduction of the electron-deficient moiety, 5% to 50% TZ were also incorporated in PM6 as parent copolymer. As shown in Fig. S3, S5 and S7b (ESI †), PM6-based terpolymers exhibited a virtually similar evolution as the D18based terpolymers.
Next, the PL lifetime of neat D18 and D18-based terpolymers was characterized by time-resolved photoluminescence (TRPL) spectroscopy. The data analysis details have been described in our previous works. 40,41 As presented in Fig. 1c, the D18(5%TZ) neat polymer thin film had a PL lifetime of 373.9 ps, which is higher than that of the parent polymer D18 (196.5 ps) in neat films, implying that terpolymerization with a small fraction (B5%) of a second acceptor can enhance the exciton lifetime, which could be attributed to the finely optimized crystal structure possesses fewer defects. However, when further increasing the TZ content to D18(10%TZ) and D18(50%TZ), the PL lifetime was reduced, implying that excessive TZ may result in severer defects. We note that the PL lifetime of neat PM6 and PM6based terpolymers was increased with the incorporation of TZ (Fig. S9, ESI †), indicating that exciton lifetimes are generally increased by the strong electron-deficient TZ moiety. Fig. 1d summarizes the energy levels, optical bandgaps, and exciton lifetimes of terpolymers based on D18 and PM6, demonstrating To decipher the role of the TZ moiety on the exciton diffusion lengths, we fitted the TRPL kinetics in the ps-ns time range with a rate equation accounting for monomolecular exciton recombination into the ground state and bimolecular exciton-exciton annihilation (EEA). 42 Typically, EEA leads to a fast intensity-dependent decay of the exciton population in the initial time range below 100 ps. Fig. S10 (ESI †) shows the fitting results, which confirm the presence of both bimolecular EEA and exciton recombination. The fitting parameters extracted from fitting the experimental data are presented in Table S5 (ESI †). While the absolute values of the exciton diffusion length cannot be determined without measuring fluence-dependent kinetics, we can still estimate the ratios between the exciton diffusion lengths of D18(5%TZ), D18(10%TZ), and D18(50%TZ) and compare to that of D18, and we find 1.1, 0.8 and 0.8, respectively. The results imply that the incorporation of 5% TZ has little impact on the exciton diffusion length. In fact, the enhanced PL lifetime can facilitate exciton dissociation. 43,44 Next, the molecular orientation and crystallinity of the donor materials in thin films were probed by grazing-incidence wideangle X-ray scattering (GIWAXS). As shown in Fig Table S2 (ESI †), the terpolymers exhibit virtually the same molecular orientation despite the introduction of the TZ moiety. However, their crystallinity was reduced by increasing the TZ content from 5% to 50% in the parent polymer D18. We conclude that the terpolymerization strategy retains the polymers' molecular orientation while gradually reducing their crystallinity. Thus, based on the results from GIWAXS and TRPL, introducing a small fraction of a second acceptor moiety leads to optimized morphology and enhanced PL lifetimes without sacrificing the materials' exciton diffusion length.
Then, we evaluated the impact of terpolymerization on the solar cell device performance. The conventional device architecture of glass substrate/indium tin oxide (ITO)/poly (3,4- Table 1. The D18:Y6 device exhibits a PCE of 17.2% with a V OC of 0.853V, a J SC of 27.17 mA cm À2 , and a fill factor (FF) of 74.1%. Increasing the TZ content from 5% to 50% improved the V OC of the D18-based terpolymer:Y6 devices due to the lowering of the HOMO level in the terpolymers. Furthermore, the D18-based terpolymer:Y6 devices exhibited enhanced J SC and FF when introducing 5% TZ in D18. Incorporating larger TZ contents, however, significantly reduced J SC and FF, indicating a trade-off exists between improving V OC and reducing the device photocurrent (J SC and FF). The external quantum efficiency (EQE) spectra of the D18:Y6, D18 ( The photovoltaic performance improvement of D18-based terpolymer photovoltaics is primarily due to increased V OC and FF. We first focus on the mechanism behind the V OC enhancement in D18-based terpolymer:Y6 devices. Electroluminescence (EL) and sensitive external quantum efficiency (EQE PV ) measurements were carried out on D18:Y6, D18(5%TZ):Y6, D18(10%TZ):Y6, and D18(50%TZ):Y6 blends (Fig. 2a-e). As presented in Table 1 and Fig. S6 (ESI †), the blends' optical bandgaps (E g ) were determined at the intersection of the normalized steady-state absorption and PL emission spectra of the D18:Y6, D18(5%TZ):Y6, D18(10%TZ):Y6, and D18(50%TZ):Y6 blends. The incorporation of TZ (5%, 10%, and 50%) into the backbone monotonically increased E g from 1.447 eV (D18) to 1.450 eV, 1.455 eV, and 1.465 eV. The chargetransfer (CT) state energy (E CT ) of D18:Y6 and D18-based terpolymer:Y6 devices was estimated by fitting the EL data to equation 5 shown in the ESI. † The introduction of 5%, 10%, and 50% TZ monotonically increases E CT from 1.439 eV to 1.442 eV, 1.450 eV, and 1.465 eV, respectively. The total photovoltaic energy loss limiting the Voc in devices consists of both radiative and non-radiative recombination losses (DV OC,rad and DV OC,nr ), which can be estimated by equations 7-10, demonstrated in the ESI. † The introduction of TZ (5%, 10%, and 50%) increases DV OC,rad from 0.330 V to 0.350 V, 0.355 V, and 0.369 V, while reducing the DV OC,nr from 0.256 V to 0.232 V, 0.224 V, and 0.192 V, respectively. The enhanced V OC from terpolymers implies that the introduction of a second strong electrondeficient moiety (TZ) has a positive impact on the nonradiative recombination losses in turn increasing the V OC of terpolymer photovoltaics. However, to support the general versatility of the approach to incorporate another electrondeficient moiety (TZ) to enhance the V OC , TZ of different content was introduced to the donor polymer PM6. Clearly, the introduction of TZ improves the V OC from 0.839 V in PM6:Y6 to 0.871 V in PM6(5% TZ):Y6 to 0.900 V in PM6(50% TZ):Y6 (Table S4, ESI †),  supporting the effectiveness of terpolymerization with TZ as a common approach to enhance the V OC of D18 and PM6-based terpolymer photovoltaics. Next, we investigated the FF enhancement and studied the exciton and charge carrier dynamics to better understand the photophysics of TZ-based terpolymers. The impact of TZ on the exciton dynamics in the D18-based terpolymer:Y6 BHJ active layer was studied by TRPL and PL quenching efficiency (PLQE) measurements. Here, the PLQE is obtained as (1 À t blend /t neat ), where t blend and t neat represent the lifetimes of the blend and neat materials, respectively. The fitting details have been described in our previous works. 40,41 We parameterized the PL decay using a bi-exponential decay function to extract the weighted-average lifetimes. The PL of neat Y6 and blend films    line with the observed change of the J SC and EQE from devices, indicating that the additional acceptor moiety (TZ) influences exciton quenching in terpolymer photovoltaics.
To further reveal the photogenerated charge carrier dynamics, we performed transient absorption (TA) pump-probe measurements over a wide temporal (picoseconds to microseconds) and broad probe energy range (2.2-0.7 eV) using our home-built TA setup. 45 Positive DT/T signals correspond to the material's ground state bleaching (GSB), while negative DT/T signals correspond to regions of photo-induced absorption (PA) from excited states. 39,45,46 Fig. 4a-c shows the ps-ns TA spectra of D18:Y6, D18(5%TZ):Y6, and D18(50%TZ):Y6 blend films after photoexcitation at 650 nm, where mainly Y6 molecules are excited (Fig. 1b). Soon after photoexcitation at 700 fs, a photo-induced absorption (PA) band has emerged in the 1.4-1.45 eV region (orange line in Fig. 4a-c) observed in all three blends, which we assigned to charge transfer states formed in Y6, in agreement with previous studies. [47][48][49][50] Over time, the narrow band at 1.4 eV exhibited spectral broadening and a concomitant redshift. This spectral evolution indicates the formation of a new species, namely, charge carrier, observed by their charge-induced absorption. 47 The TA spectra of neat donor and acceptor materials are given for comparison in the SI ( Fig. S13 and S14, ESI †).

View Article Online
The GSB band of the donor (1.9-2.1 eV) exhibited a slow rise up to 100 ps after an ultrafast charge generation. The exciton generated at the donor/acceptor interfaces can go for ultrafast charge generation, but the exciton generated far away from the D/A interface need diffuse through the pure phases. As a result, diffusion through pure phase delays the charge transfer/slows down the generation, but they do not decrease its efficiency because modern NFAs show long-range exciton diffusion length. 51 In addition to larger diffusion length, Y6 exhibit a larger exciton lifetime, which is much larger than other NFAs 51 larger than the maximum charge generation time reported here. Fig. 4d shows the integrated kinetics of the 1.22-1.3 eV (charge carrier-induced absorption) spectral region. The charge carrier generation rate for D18(5%TZ):Y6 blend films are faster than that of D18:Y6 and D18(50%TZ):Y6 at the same laser fluence, and in line with the exciton (fluorescence) lifetimes obtained by TRPL experiments (Fig. 3e). Fig. 4d compares the charge carrier dynamics of thin film blends. Here, a slightly faster charge generation but slower recombination was observed for D18(5%TZ):Y6 films (orange line), which is in line with the higher PCE of this blend.
To monitor the charge carrier recombination dynamics and quantify the carrier recombination rates, we performed nanosecond-microsecond (ns-ms) TA spectroscopy on the same blends used for the ps-ns TA studies. 39,45 The ns-ms TA spectra showed similar spectral features for all three thin film blends (see Fig. S14, ESI †), yet their kinetics differed. Here, the fluenceindependent decay of the charge-induced absorption at early probe delay times indicates a contribution from geminate recombination of bound charge-transfer states, while the fluence-dependent decay indicates non-geminate recombination of free charges dominates at later times. 52 To quantify the individual process yields and to determine the kinetic parameters we fitted the experimentally-determined dynamics to a previously reported two-pool model (solid lines in Fig. 4e-g). Details of the two-pool model and fit can be found in earlier works and in the ESI. † 52,53 Here, we assumed the charge carrier cross-section to be 1 Â 10 À16 cm À2 , that is, on the order of values typically determined for polymeric materials. 54 The fit to the two-pool model revealed that spatially-separated charges are generated with a yield of 96% and 90% in D18(5%TZ):Y6 and D18:Y6 blend films (Table 2), respectively. Free-charge carriers can be extracted as photocurrent from a device if they escape non-geminate recombination. Interestingly, in D18(50%TZ):Y6 films, the yield of free charge carriers were found to be much lower (76%) ( Table 2), in turn indicating that 24% of the initial charge carrier population is bound in chargetransfer states which subsequently recombine geminately. Clearly, it can be seen that carrier losses due to geminate charge recombination directly impact the obtained photovoltaic device performance.
Apart from a reduced fraction of geminate recombination, D18(5%TZ):Y6 exhibited also a further reduced non-geminate recombination coefficient (b), indicating that free charge carriers are longer-lived and thus can be extracted to a larger extent from the device. In contrast, D18:Y6 and D18(50%TZ):Y6 films exhibited higher b values, pointing to faster non-geminate recombination of free charge carriers and thus less charge extraction. Overall, our intensity-dependent TA results reveal not only the mechanism of carrier recombination but also the fraction of charge carrier losses. It appears that adding 5%TZ to the parent polymers reduced carrier losses in blends with Y6, compared to the control device D18:Y6 (Table 2). However, incorporating higher loadings of TZ into the polymer backbone reduced the charge generation efficiency and increases both geminate and non-geminate recombination of charge carriers, in turn resulting in reduced FF and J SC in devices.
In addition to the excited state dynamics, the molecular orientation, material crystallinity, and D/A phase-separation impact the device FF. 55-63 GIWAXS on thin films was performed to determine the molecular orientation and crystallinity of D18:Y6 and D18-based terpolymer:Y6 thin films and their D/A phase-separation was assessed by grazing incidence small-angle X-ray scattering (GISAXS). As presented in Fig. 5a, b and i, the introduction of TZ caused no obvious change in the molecular orientation, implying the TZ-based terpolymerization does not affect the molecular packing and thus the charge transport pathways are also retained. The corresponding GIWAXS parameters are summarized in Table S3 (ESI †). The D18:Y6 BHJ film exhibits (100) lamellar peaks at q xy E 0.31 Å À1 (coherence length (CCL) = 5.28 nm) and (010) p-p peaks at q z E 1.80 Å À1 (CCL = 2.64 nm). In comparison, the D18(5%TZ):Y6, D18(10%TZ):Y6, and D18(50%TZ):Y6 systems present (100) lamellar peaks at q xy E 0.31 Å À1 (CCL = 4.67, 4.25 and 3.95 nm, respectively) and (010) p-p peaks at q z E 1.77 Å À1 (CCL = 2.55, 2.36 and 2.29 nm, respectively), showing a continuously reduced film crystallinity in both IP and OOP direction when increasing the TZ content. The reduced film crystallinity is consistent with the reduced charge carrier mobility found in D18-based terpolymer solar cells (Table S7, ESI †). However, the D18(5%TZ):Y6 device exhibits an enhanced FF, which can be ascribed to different and more optimal D/A phaseseparation. 35,58 GISAXS was carried out to evaluate the D/A phaseseparation, and the 2D patterns are presented in Fig. 5e-h, and their corresponding IP intensity plots and fitting results based on the Debye-Anderson-Brumberger (DAB) model and fractal-like network are shown in Fig. 5j. X DAB refers to the intermixing D/A domain size and 2R g represents the average pure domain size of the acceptor phase in this work for D18:Y6 and D18-based terpolymer:Y6 samples. The corresponding Table 2 Fit parameters obtained from global fits (two-pool model) to the ns-ms charge carrier recombination dynamics in optimized blend films. Here, f is the fraction of free charge carriers, l + 1 is the apparent nongeminate recombination order, g is the non-geminate decay rate, and b is the effective bimolecular recombination coefficient calculated for a carrier density of 1 Â 10 16  Next, carrier drift-diffusion simulations were used to reproduce the experimentally-measured J-V characteristics. We used optical constants, namely the refractive index n and extinction coefficient k, determined by spectroscopic ellipsometry measurements (Fig. S18, ESI †), hole and electron mobilities measured by SCLC experiments (Fig. S17, ESI †), the bimolecular charge carrier recombination coefficient (b TDCF ) obtained from time-delayed collection field (TDCF) measurements (Fig. S16, ESI †), and the device stack architecture as input parameters ( Fig. S11, ESI †). We note that the bimolecular recombination coefficient obtained from TA on thin film blends differs from the one determined by TDCF. This mismatch is largely due to the n-dependence of b as we explained in our previous work 64 and ESI. † In fact, TDCF appears to give a better estimate of the carrier recombination coefficient in devices as found in our previous works and thus we used the recombination coefficient determined by TDCF for our device simulations. 65 More details about the simulation can be found in the ESI, † and our previous works. 40,65 As shown in Fig. S16 (ESI †), the b TDCF of D18:Y6, D18(5%TZ):Y6, D18(10%TZ):Y6, and D18(50%TZ):Y6 devices are 7 Â 10 À11 , 5 Â 10 À11 , 9 Â 10 À11 and 1 Â 10 À10 cm 3 s À1 , respectively. We note that the trend for b TDCF is in line with the change of b TA (Fig. 4e-g), confirming reduced carrier recombination when 5% TZ is incorporated into the polymer backbone, while recombination increases when introducing a higher content of TZ. Details of the simulation using the commercial software tool Setfos 5.2 (Fluxim AG) and the precise simulation procedure can be found in our previous works. 40,45,65,66 The comparison of the simulated and experimentally-measured J-V characteristics of the terpolymer photovoltaics are presented in Fig. 5k-n, and the relevant photovoltaic parameters are listed in Long-term (outdoor) operational stability is of utmost importance to commercialize PV technology, and an important challenge for OPV. The devices need to sustain prolonged operation under solar irradiance and exhibit high tolerance to heat and temperature changes during day-night cycles. 23 Thus, a thermodynamically stable morphology is required to obtain outdoor-stable OSCs. Previous works have demonstrated that D/A miscibility close to the percolation threshold of small molecules (fullerenes and NFAs) is key to obtaining a  (meta)stable morphology. 67,68 Here, the terpolymerization strategy is employed to fine-tune the D/A miscibility to obtain outdoor-stable OSCs with a thermodynamically stable morphology. The experimental details have been shown in ESI. † The outdoor stability tests for all four systems were performed in the hot and sunny climate at KAUST, Thuwal, Kingdom of Saudi Arabia, a location with high irradiance and temperatures. In fact, the irradiance was as high as 1039.5 W m À2 at noon (Fig. 6a) as measured by a calibrated pyranometer and the highest/lowest air temperatures during daytime/nighttime were 48 1C/22 1C (Fig. 6b). As shown in Fig. 6c, the figures-of-merit of the encapsulated D18:Y6, D18(5%TZ):Y6, D18(10%TZ):Y6 and D18(50%TZ):Y6 devices were tracked every 10 min for 32 days in total, presenting an uptrend of the PGD from sunrise to noon, followed a downtrend from noon to sunset. For a better comparison, the device performance obtained at noon was normalized (accounting also for the irradiance by dividing the irradiance measured at noon) to monitor the outdoor degradation (Fig. 6d). The devices based on D18:Y6 (w donor-acceptor (D/A miscibility) of 59.9 Â 10 À2 K) and D18(5%TZ):Y6 (w donor-acceptor of 50.5 Â 10 À2 K) (Table S10 and Fig. S21a, b, ESI †) exhibited a similar degradation in the first 4 weeks, however, the D18:Y6 devices demonstrated significantly lower device performance after 4 weeks. It appears the miscibility of 50.5 Â 10 À2 K in D18-based terpolymer:Y6 systems deliver a thermodynamically more stable thin film. Furthermore, the D18(10%TZ):Y6 with a w donor-acceptor of 12.5 Â 10 À2 K and D18(50%TZ):Y6 with a w donor-acceptor of 1.3 Â 10 À2 K (Table S10 and Fig. S21c, d, ESI †) exhibited much severer degradation in the first week, implying that the miscibility in D18-based terpolymer devices is different from that in D18(5%TZ):Y6 systems, resulting in less thermodynamically stable films. It is worth noting that the D18(5%TZ):Y6 system, which exhibited the thermodynamically most stable morphology, retained 92.6% initial performance after 32 days of harsh outdoor stressing in Saudi Arabia, which is among the best outdoor stabilities reported so far for 3rd generation photovoltaics. 22,23,[69][70][71] To confirm that the D18(5%TZ):Y6 system is the most thermodynamically stable blend, hyperspectral PL imaging was performed on fresh and aged devices of all four systems tested outdoors. The normalized average PL spectra of the entire mapped area are depicted in Fig. S24a and b (ESI †). The D18 and D18-based terpolymers are very stable due to their intrinsic morphological stability, whereas, NFAs are prone to aggregate under light-soaking or heat exposure. Hence, by tracking the PL intensity of the D18/D18-based terpolymer and the Y6 acceptor domains, we can disentangle which D18/ D18-based terpolymer:Y6 system is the most thermodynamically stable. As shown in Fig. S24a (ESI †), all four systems in fresh devices exhibited a stronger PL of the Y6 phase, indicating optimum blend morphology. 65 After 32 days of outdoor testing, the D18:Y6, D18(5%TZ):Y6, D18(10%TZ):Y6, and D18(50%TZ):Y6 devices all exhibited significantly reduced PL intensity of the Y6 domains due to the aggregation of Y6, 72,73 while the D18(5%TZ):Y6 system still showed comparably intense PL from the Y6 domains (Fig. S24b, ESI †), indicative of a more thermodynamically stable morphology and in line with the outdoor stability data showing 92.6% of the initial device performance was retained after 32 days.

Conclusion
This study summarized a set of rules for high-performance and outdoor-stable terpolymer solar cells and proposed a rationale to explain the working principle behind them. We combined spectroscopic, morphological, and computational approaches to elucidate the key rules for highly efficient and outdoor-stable terpolymer solar cells based on parent polymer donor D18 and PM6. We found that the device's J SC is enhanced and energy loss is reduced in terpolymer devices when incorporating another strong electron-deficient moiety that broadens the material's optical absorption and lowers the HOMO level of the terpolymer. Enhanced device FFs were found when the fraction of the third component of the terpolymer was 5% only, which retained the original D/A framework with delicately optimized phase-separation that reduced the bimolecular charge carrier recombination and further improved the device FF. Highly outdoor-stable terpolymer devices were obtained due to the incorporation of a third component which finetuned the D/A miscibility and led to a thermodynamically more stable morphology. Our findings systematically unveiled the photovoltaic performance and outdoor stability enhancement by monitoring the exciton/charge carrier dynamics and the blends' thin film morphology, providing a rationale for highly efficient and outdoor-stable terpolymer solar cells, pushing them closer to commercial realization.

Conflicts of interest
There are no conflicts to declare. thanks the research grant from Chongqing Funds for Distinguished Young Scientists (cstc2021jcyj-msxm0139). G. L. thanks the Hong Kong RGC Senior Research Fellowship Scheme (SRFS2223-5S01). H. T. acknowledges Prof. Derya Baran for her fruitful discussion and valuable suggestion for revising the draft.