All-polymer solar cells with over 16% efficiency and enhanced stability enabled by compatible solvent and polymer additives

Considering the robust and stable nature of the active layers, advancing the power conversion efﬁciency (PCE) has long been the priority for all-polymer solar cells (all-PSCs). Despite the recent surge of PCE, the photovoltaic parameters of the state-of-the-art all-PSC still lag those of the polymer:small molecule-based devices. To compete with the counterparts, judicious modulation of the morphology and thus the device electrical properties are needed. It is difﬁcult to improve all the parameters concurrently for the all-PSCs with advanced efﬁciency, and one increase is typically accompanied by the drop of the other(s). In this work, with the aids of the solvent additive (1-chloronaphthalene) and the n -type polymer additive (N2200), we can ﬁne-tune the morphology of the active layer and demonstrate a 16.04% efﬁcient all-PSC based on the PM6:PY-IT active layer. The grazing incidence wide-angle X-ray scattering measurements show that the shape of the crystallites can be altered, and the reshaped crystallites lead to enhanced and more balanced charge transport, reduced recombination, and suppressed energy loss, which lead to concurrently improved and device efﬁciency and stability.


INTRODUCTION
All-polymer solar cells (all-PSCs) are considered as a promising photovoltaic technology due to their exclusive advantages. For instance, organic solar cells with all-polymer active layers can possess excellent stability and robustness together. [1][2][3][4][5][6][7] This type of photovoltaic has experienced two decades of slow development in terms of the power conversion efficiency (PCE) due to the lack of efficient polymeric materials. However, the recent practice of polymerized small molecular acceptors has brought the PCE of all-PSCs to a new level. There have been multiple reports with >10% PCEs within a short timeframe. [8][9][10][11][12][13][14][15][16][17][18][19] To date, the best performing all-PSCs demonstrate PCEs in the range of 15%-16%. [20][21][22][23][24][25][26][27] Nevertheless, compared to the achievements by the small molecule-based solar cells (∼18% PCE), there is still a long way ahead for all-PSCs. [28][29][30][31][32][33][34][35][36] From the experience of small molecule-based solar cells, fine-tuning the morphology and thus the device properties are essential for further improving the device efficiency in a material system that already manifests decent photovoltaic performance. In all-PSCs, it would be even more difficult because the formation of morphology is based on two tightly entangled materials both consisting of long conjugated chains. Particularly, it would be extremely challenging to improve all three main device parameters, that is, opencircuit voltage (V OC ), short-circuit current density (J SC ), and fill factor (FF) concurrently in the all-PSCs with advanced efficiency, and one increase is typically accompanied by the drop of the other(s). But this does not mean that there is no way to do this from the perspective of device engineering. For instance, the use of additives (solvents or solids, small molecules, or polymers) has been proven an effective strategy to promote the photovoltaic performances of solar cells, as reported by many. [37][38][39][40][41][42] In all-PSC systems, traditional solvent additives such as 1,8-diiodooctane (DIO) and 1-chloronaphthalene (CN) are also effective in promoting device performance. In addition to solvent additive, polymers are successfully used as additive in some recent cases as well. Min, [43] Yang, [44] and Huang [45] et al found that the use of polymer acceptor as additive could also efficiently enhance the device performance.
In this work, we employed the all-PSC based on PM6:PY-IT as the target system, which already demonstrated stateof-the-art photovoltaic performance, [21] and judiciously modified the morphology of the all-polymer active layer through fine-tuning the solvent additive and solid additive during film preparation for boosting device performance. Specifically, compared to the control device (1 vol % CN in active blend solution) with a PCE of 14.93%, we optimize the content of CN and utilize a small amount of N2200 as solid additive to fabricate all-PSC devices, which endow the devices with an excellent efficiency as high as 16.04%. The synergistic effect of the solvent and polymer additives improves the photovoltaic parameters of open-circuit voltage (V OC ), short-circuit current density (J SC ), and fill factor (FF) simultaneously. The grazing incidence wide-angle X-ray scattering (GIWAXS) and device physics studies show that the solvent and solid additives can effectively alter the shape of the crystallites, making them grow "taller" (relative to the substrate plane), then enhances the charge transport while suppressing recombination, which also leads to a reduced nonradiative voltage loss. Notably, compared to the control device, the optimal device with the aid of the solid additive also shows increased storage stability and photostability.

RESULTS AND DISCUSSION
The chemical structures of PM6, PY-IT, CN, and N2200 are shown in Figure 1A, and their optical properties, that is, the absorption in neat films and blend films, are shown  in Figures 1B and 1C, respectively. The complementarity between the absorption spectra of PM6 and PY-IT is clearly observed in Figure 1B, which is consistent with previous report. Besides, two main absorption peaks of N2200 are in the regions that are complementary to the absorption of the PM6:PY-IT blend film. N2200 used here is commercially available from 1-Materials; molecular weight is 30 KDa. Depending on the different amount of CN and/or N2200 used during solution/film preparation, we obtain four blend films that are all based on the same PM6:PY-IT: 0.7 vol % CN, 1 vol % CN, 0.7 vol % CN + 3 wt % N2200, and 1 vol % CN + 3 wt % N2200. Figure 1C shows the absorption spectra of these films. The main features are similar among them, but the film processed with 0.7 vol % CN and 3 wt % N2200 shows the highest absorption in the range of 680-750 nm. The absorption coefficients summarized in Table S1 include all details for comparison. In addition to absorption, we also evaluated the tendency of N2200 to mix with the PM6 and PY-IT through surface tension analysis ( Figure S1). The result shows that surface tensions of PM6, PY-IT, and N2200 are 21.45, 20.98, and 21.50 N/m, respectively, which are close to each other, indicating a decent intermixing propensity of N2200 in the blend. [46][47] Then we fabricated a series of devices using a structure of ITO/PEDOT:PSS/bulk heterojunction (BHJ)/PNDIT-F3N/Ag to study the effect of binary additives on the resulting solar cell performance. The current density versus voltage (J-V) characteristics of the devices are plotted in Figure 2A, and the corresponding photovoltaic parameters are listed in Table 1. The PM6:PY-IT-based binary all-PSC with 1 vol% CN (Device II) shows a PCE of 14.93%. By adding 3 wt% N2200 into the blend active layer (Device IV), the efficiency was promoted to as high as 15.72% owing to the simultaneously increased V OC , J SC , and FF. We find that fine-tuning the ratio of CN from 1 vol% to 0.7 vol% (Device I) allows us to improve the V OC and FF, but the PCE is reduced to 14.77% due to the J SC drop. To compensate the J SC and thus increase all three parameters concurrently, we incorporated a small amount of N2200 into the blend and examined the device performance influenced by the interplay between the solvent additive (CN) and the solid additive (N2200). We achieved the highest efficiency of 16.04% together with an excellent FF of 74.9% when 0.7 vol% CN and 3 wt% N2200 (Device III) were used. Notably, the PCE of 16.04% is the superior value for all-PSCs, regardless binary or ternary, singlejunction, or tandem devices, to the best of our knowledge. Figure 2B summarizes the photovoltaic performance, that is, PCE versus FF, of recent all-PSCs with cutting-edge efficiencies (>15%), which highlights the significance of our results.
The external quantum efficiency (EQE) spectra of the devices are shown in Figure 2C. Comparing the J SC s integrated from these EQE spectra and those from the J-V measurement (Table 1) shows that the errors are within 3%. Besides, the efficiency was confirmed by bringing them to another laboratory to test ( Figure S2, supported by The Hong Kong Polytechnic University, Prof. Gang Li). Apart from efficiency, the device stability of all-PSCs is measured. The storage stability and photostability measurements for encapsulated devices pictured in Figure 2F were conducted in air conditions. We plot the change in efficiency with time in Figures 2D (storage stability) and 2E (light soaking), for Devices I and III. The T80 values for the latter are 997 and 250 h, for storage and light soaking stability, respectively, [48] where T80 means the time it costs for device to lose 20% of its initial PCE and was estimated by linear fitting. These results are significantly higher than those for Device I (557 and 194 h), demonstrating that the addition of N2200, despite a small amount, can reduce the decay of the devices. N2200 is supposed to be beneficial to thin film morphology stability of active layers, resulting better device stability. This method was successfully applied in some ternary works. [49][50][51] For storage stability test, the main differences of PCE degradation start from Day 10, indicating the original morphology of PM6:PY-IT was undergoing a more intensive degradation compared with N2200 doped film. This can also be the explanation of light soaking stability differences beginning from 40th h.
The addition of N2200 into the PM6:PY-IT blend leads to a further increased V OC (from 0.932 to 0.947 V) relative to the device processed with CN as additive compared, which draws our interest since no significant blueshifts in absorption and EQE spectra are observed. This indicates that the inclusion of N2200 regulates the recombination and thus the energy loss (E loss ). To investigate the energy loss, we employed Fourier transform photocurrent spectroscopy EQE (FTPS-EQE) and electroluminescence EQE (EL-EQE). Figures 3A and 3B show the testing results of FTPS-EQE and EL-EQE, respectively. E loss of a solar cell can be divided into three parts as described in the equation (details in supporting information): [52][53] ΔE 1 is an inevitable radiative loss for BHJ solar cells working under 1-sun or below, which can be calculated based on the Shockley-Queisser theory. The four PM6:PY-IT devices we studied in this work show the same value of ΔE 1 (0.27 eV); ΔE 2 is the radiative loss below the bandgap due to nonstep-function like absorptance, which also appears to be identical (0.03 eV) for the devices in this work. The last part, ΔE 3 , is the loss due to nonradiative recombination, which is quite different among different types of solar cells and can be calculated via ΔE 3 = -k B T(lnEQE EL ), with k B the Boltzmann constant, T the absolute temperature, and EQE EL the EQE of EL. Compared to the other three counterparts (0.24 V), Device III shows the smallest nonradiative recombination loss (0.23 V), the difference in which roughly matches the difference in the V OC of their devices. This shows that the one of the combined efforts of 0.7 vol% CN and 3 wt% N2200 is reducing the nonradiative recombination.
In addition to V OC , tuning the ratio of CN and adding N2200 also boosted the J SC and FF. To understand the enhancement of these parameters, we first plot the photocurrent density (J ph ) as a function of the effective voltage (V eff ) ( Figure 2C) to estimate the exciton dissociation (η diss ) and collection efficiency (η coll ). [54][55][56] J ph is defined as the difference between the current density under illumination (J L ) and the dark current density (J D ). V eff is the absolute value of V 0 -V appl , where V 0 refers to the voltage when J L = J D , and V appl is the applied voltage. At high V eff , almost all excitons are separated and extracted, and J ph reaches saturation (J sat ).η diss and η coll are defined by J SC /J sat and J max /J sat , respectively, in which J max is the current density at the maximal output point. (η diss , η coll )s of all-PSCs we focus on are (92.8%, 82.1%), (91.6%, 78.5%), (93.3%, 83.7%), and (92.5%, 80.3%) for 0.7 vol% CN, 1 vol% CN, 0.7 vol% CN & 3 wt% N2200, and 1 vol% CN & 3 wt% N2200 treated blends, respectively. This trend agrees with the trend of the FF of their devices. Next, we measured the steady-state photoluminescence (PL) spectra of the polymer neat film and the blend films ( Figure 2D) to compare the quenching efficiency. [57][58][59] The PL quenching efficiencies are 74.1% for the Device I, 70.3% for Device II, 78.2% for Device III, and 76.3% for Device IV, consistent with analysis above.
Furthermore, we measured the J-V characteristics of the devices under different light intensities to study the change in recombination. [60][61] Figure 2E delivers the relationship between J SC and light intensity (P light ), and Figure 2F shows that between V OC and P light . Theoretically, V OC is linearly proportional to ln(P light ), and the slope is equal to nkT/q, where n is the ideality factor, k is the Boltzmann constant, T is the temperature, and q is the elementary charge. Fitting the data in Figure 2E provides us with the values of the ideality factor, which are 1.197, 1.254, 1.147, and 1.236 for Device I, II, III, and IV, respectively, also consistent with the trend of the FF. This implies that the trap-assisted recombination and monomolecular recombination are least dominant in the optimal active layer (Device III). To assess the bimolecular recombination, we fit the curves of J SC versus ln(P light ). The slope from the fitting (S) is theoretically less than 1, and the closer it is to unity, the weaker the bimolecular recombination is in the device. The S values from fitting the data in Figure 2F are 0.972, 0.960, 0.983, and 0.972 for Device I, II, III, and IV, respectively, suggesting that our best device also has the weakest bimolecular recombination.
Based on electrical characterization and recombination analysis, we then turn our attention to the morphology of the active layer as it is directly related to the processing solvents and materials composition during film preparation. We first performed atomic force microscopy measurements on the films of the four blends ( Figure S3), which show similar surface morphology. We then carried out GIWAXS experiments. [62][63][64] The 2D patterns of the blend films are demonstrated in Figure 4A, while the corresponding intensity profiles in the in-plane (IP) and out-of-plane (OOP) directions are shown in Figure 4B. The peaks corresponding to the lamellar packing of all four blends are located at 0.30 Å −1 , and the intensity is in the IP direction, but the crystalline coherent lengths (CCLs) are 50.0, 58.3, 40.9, and 54.9 Å for Device I, II, III, and IV, respectively. The π-π stacking peaks, found in the OOP direction, are all located at 1.67-1.68 Å −1 . In contrast to the CCL of the lamellar packing, the CCLs for the π-π stacking peaks in the OOP direction  20.4, 19.1, 23.8, and 19.7 Å, for Device I, II, III and IV, respectively. These data suggest that either using less CN or adding N2200 into the blend can suppress the IP lamellar packing while promote π-π stacking. In other words, the combined effect of the optimized amount of CN and N2200 is to alter the shape of the crystallites, reducing the initially overly packed lamellar while enhancing the π-π stacking in the OOP direction, rendering "taller" and face-on crystallites.
To figure out N2200's role in thin film morphology tuning, we evaluated the neat film GIWAXS results for PM6, PY-IT, and N2200 ( Figure S4). PM6 exhibits a (100) peak at 0.29 Å −1 alongside IP direction, whose CCL is 59.5 Å, meanwhile a (010) peak of 1.72 Å −1 with a 17.66 Å CCL in OOP orientation. PY-IT displays a lamellar peak ∼ 0.38 Å −1 (CCL = 43.4 Å) on IP direction, and an OOP located π-π stacking peak (q ∼ 1.65 Å −1 ) with 18.84 Å CCL. These two materialmerged blend films demonstrate reasonable molecular packing features as described above. In contrast, N2200 contains two IP peaks at 0.26 and 0.47 Å −1 , which might undermine the IP packing through doping. Furthermore, the (010) peak with strong intensity can be beneficial to OOP orientated ππ stacking. This should benefit to charge transport in the active layer as the vertical charge transport is the desired pathway for charge extraction. To investigate the effect of such morphology change, we fabricate hole-only ( Figure 4C) and electron-only ( Figure 4D) devices and measure the J-V characteristics to evaluate the charge carrier mobilities. The hole (μ h ) and electron mobility (μ e ) fitted using the spacecharge limited current (SCLC) model are 8.23 and 4.63 × 10 −4 cm 2 V -1 s -1 for Device I, with a hole/electron mobility ratio (μ h /μ e ) of 1.78. In comparison, the μ h and μ e for Device II are 7.61 and 4.55 × 10 −4 cm 2 V -1 s -1 , and the μ h /μ e is 1.67, which indicates that reducing the CN ratio leads to slightly lowered mobilities but more balanced hole charge transport. Besides, μ h for Device III and Device IV are 8.79 and 7.98 × 10 −4 cm 2 V -1 s -1 , while the μ e for are 5.22 and 4.89 × 10 −4 cm 2 V -1 s -1 , respectively. Therefore, the overall result of CN and N2200 is enhancing hole and electron transport simultaneously, and balancing them in the meantime, which is consistent with the morphology result and the J SC , FF of the devices.

CONCLUSION
In summary, we combined the polymer additive, N2200, with the solvent additive, CN, during film preparation of the PM6:PY-IT-based all-PSCs, which improved the PCE from 14.93% to 16.04%, the highest value for this kind photovoltaics. Reduced nonradiative loss, more efficient charge generation, and lower recombination were enabled by finetuning the proportion of additives, which led to concurrently promoted V OC , J SC and FF. Morphology study revealed that these improved photovoltaic performances are correlated with the change in the crystallite molecular packing, i.e., from a "wide and short" shape to a "narrower but taller" form, which increased and balanced charge transport. Besides, optimized devices show enhanced stability compared to the control device, exhibiting T80 values as high as 997 and 250 h for storage and light soaking stability, which are also at cutting-edge level for state-of-the-art polymerized small molecular acceptor-based all-PSCs. Overall, this work successfully pushed the efficiency of all-PSCs up to > 16%, which strongly enhances the competitiveness of this device platform.