A review of nonfullerene solar cells: Insight into the correlation among molecular structure, morphology, and device performance

Nonfullerene acceptors (NFAs) lead the continuous development of organic solar cells (OSCs) with competitive efficiency over 19%. Design and synthesis of novel photovoltaic materials are effective methods to improve the OSCs performance, which can regulate the optoelectric properties, such as energy level, absorption spectra, charge transport, and so on. So far, hundreds of NFAs have been reported. Meanwhile, it has been demonstrated that intrinsic morphology of active layer is partially determined by the chemical structures of NFAs. Hence, only in‐depth understanding of the relationship between different structures of NFAs and morphology can guide the molecular design of NFAs for highly efficient OSCs. Herein, we review some state‐of‐the‐art NFAs according to their functional moieties, that is, arene core, end group and side chain, and discuss the relationship between molecular structure, morphology and device parameter. Additionally, the challenges and prospects for further development of OSCs based on NFAs are briefly considered. This review brings a unique insight into structure–function correlation in this field, which may help to rapidly develop efficient OSCs.


| INTRODUCTION
Developing clean and renewable energy has become an urgent goal for the problem solving of global environmental and energy issues. Organic solar cells (OSCs) have always been regarded as a potential option of green energy sources, which attract plenty of attention among solar energy fields. With rapidly increasing power conversion efficiency (PCE) approaching silicon-based solar cells, OSCs present the merits of light weight, flexibility, semitransparency, and large-area printing, which promise the potential commercializing value and unique application scenarios. [1][2][3][4] At present, the state-ofthe-art single-junction OSCs have achieved over 19% PCE for bulk heterojunction (BHJ) structure. [5][6][7] However, the performance of OSCs still falls behind that of siliconbased and perovskite solar cells. Therefore, further improving PCE is the primary aim of OSCs research.
The synthesis of novel photovoltaic materials is one of the most effective ways to improve the PCE of OSCs. 2,8 In the early stage, fullerene based acceptors were widely used to blend with polymer donors for efficient OSCs due to its sphere-like structure leading to favorable isotropic charge mobility. 9 However, this structure leads to inferior absorption coefficient and morphology stability, which is not conductive to obtain a high device performance. 10,11 Thus, the community devotes to create a new type of acceptors with favorable photoelectric property. In 2015, a benchmark acceptor ITIC was reported by Zhan et al., 12 which is comprised of a central electron donating unit and two electron accepting units (known as acceptor-donor-acceptor [A-D-A] type nonfullerene acceptors [NFAs], and it presents broad near-infrared absorption and comparable device performance [PCE = 6.8%]). The structure of NFAs bringing large possibility on structure engineering to tune the photoelectric property. 13,14 Since then, these kind of small molecular acceptors have started a new era for OSCs. 10,[14][15][16] So far, its derivative IT-4F has been reported the highest PCE of 14.4% in PBDB-T-2Cl:IT-4F blend by Hou et al. 17,18 Later in 2019, a novel central unit dithienothiophen [3.2-b]-pyrrolobenzothiadiazole (TPBT) was designed by Zou et al., who further synthesized TPBT-based acceptor Y6. 19 This state-ofthe-art molecule exhibits outstanding PCE over 15% with polymer donor due to its improved light harvesting at near-infrared range and high electron mobility. Modified NFA L8-BO based on the Y6 structure obtained PCE of 19.6% in ternary blend leading to the efficiency of OSCs. 6 On account of representative ITIC and Y6 molecules, numerous molecular engineering has been proposed to modify the core, end group and side chain, which yields plenty of derivatives. [20][21][22][23] These modified derivatives with varied absorption and energy levels were combined with suitable donor materials, thus leading to continuous efficiency breakthrough. 10,15 Apart from the photoelectric properties of NFAs, the morphology of the active layer has a significant influence on device performance as well. As we know, molecular structure has evident impacts on morphology, such as molecular packing, aggregation as well as phase separation. [24][25][26] Recent studies have demonstrated that the ideal morphology should possess the following characteristics: (1) relatively small domain size, evaluated to be 10-20 nm, allows efficient diffusion of excitons to the donor/acceptor (D/A) interface within their limited lifetime; (2) higher domain purity prevents the recombination of exciton and carriers; (3) ordered molecular packing and face-on orientation are beneficial to charge transport; (4) favorable vertical phase separation structure provides transport pathways facilitating charge transport and extraction. [27][28][29][30][31][32][33][34] Considering the diversity of advanced NFAs, it is necessary to correlate the chemical structure of NFAs with morphology of active layer.
In this review, we summarized representative NFAs from the perspective of molecular structure engineering. Innovation and modification on core, end group as well as side chains are illustrated with commonly used strategies. The modified chemical structures are correlated with the morphology of active layer, consequently performance of OSCs. In addition, ITIC-based and Y6-based NFAs are specially focused, indicating favorable tunability of molecular engineering. This review was completed by presenting challenges and perspectives for the future of OSCs based on NFAs.

| ARENE CORE
The electron-donating fused-ring core is a critical moiety of the NFA in determining the optoelectronic property and intermolecular interaction, consequently affecting photovoltaic performance. Because fused-ring core is the largest coplanar structure in the NFA, which induces the formation of π-π intermolecular packing, thus facilitates the charge transport. 35 Furthermore, the absorption range of NFAs can be broadened by intramolecular charge transfer (ICT) between electron-donating cores and electron-withdrawing end groups. Therefore, the fused-ring core in NFAs should be in the spotlight.

| Core size
Core extension is usually considered as an effective method to regulate molecular packing. Zhan et al. compared two series of fused ring electron acceptors (FREAs), the cores of both series consist of two terminal thiophene, thieno [3,2-b]thiophene, or dithieno[3,2b:2′,3′-d] thiophene rings, as shown in Figure 1A. The grazing incident wide-angle X-ray scattering (GIWAXS) measurements illustrate the molecular packing of pure F5IC, F7IC, F9IC, and F11IC films in Figure 1B and 1C. Pure F5IC presents an almost amorphous scattering pattern, while the molecular packing gets obviously stronger as the conjugated area increases. Especially for F11IC, scattering signals of both lamellar packing and backbone packing can be recognized. Thus, an impressive electron mobility enhancement is able to be measured from 8.1 × 10 −5 to 1.4 × 10 −3 cm 2 V −1 s −1 . When blended with polymer donor FTAZ, the blend film with larger core size acceptor presents stronger packing same as the pure film and weak phase separation ( Figure 1D), which results in increased short circuit current density (J SC ) from 14.88 to 20.20 mA cm −2 , consequently PCE enhancement (from 5.6% to 11.7%) as shown in Figure 1E. 36 However, it should be noted that the overlarge core size will restrain molecular stacking in some cases. The counterpart group F6IC, F8IC and F10IC replace the central benzene ring with thieno[3,2-b]thiophene (TT). The GIWAXS results for the pure film show that the enlarging core size, from 6-ring to 8-ring, can enhance the π-π stacking, while F10IC exhibits no further enhancement even slightly decrease than F8IC. On the other hand, the degree of side-chain lamellar stacking apparently decreases as the core size enlarges. This is because the overlarge core size may inhibit the molecular diffusivity, which prevents intermolecular stacking. The phase separation of PTB7-Th:FXIC blends were further characterized by grazing incident small-angle X-ray scattering (GISAXS) measurements, indicating 54.2, 18.5, and 14 .8 nm domain sizes for PTB7-Th:F6IC, PTB7-Th:F8IC, and PTB7-Th:F10IC. Owing to the balanced  crystallinity and suitable domain size, PTB7-Th:F8IC based device presents 10.9% PCE with obviously higher J SC (25.12 mA cm −2 ), which is the best system among PTB7-Th:FXIC. 37 The central benzene ring replaced by TT unit has become an efficient method to facilitate the electrondonating ability as well as intermolecular interaction for the synthesis of FREAs. Zhan et al. compared the tris (thienothiophene) (3TT) core acceptor FOIC to IT core counterpart, ITIC3. 38,39 According to the GIWAXS results, it is clear to figure out that FOIC presents a much stronger π-π stacking peak, which indicates the apparent effect of 3TT core on enhancing intermolecular packing. After blending with polymer donor PTB7-Th, the more ordered packing of FOIC remains in terms of the larger crystalline coherence length (CL) 1.9 nm, while the CL of ITIC3 blend is 1.6 nm. In addition, FOIC induces a smaller domain size (18 nm) and higher relative domain purity in the blended film. The improved molecular packing and phase separation lead apparently higher J SC (24%) fill factor (FF) (67.1%) as well as PCE (12%) for PTB7-Th:FOIC blend. 40 One more similar structure comparison was reported by Lin et al. but shows less distinct result for IHBT-2F and IDBT-2F. Both of them are amorphous though TT unit was replaced the central benzene ring. However referring to FOIC and ITIC3, two symmetric benzothiophene located at the two sides of core obviously decrease the intermolecular interaction comparing to two TT units. In the blend film, PTB7-Th:IHBT-2F exhibits a relatively higher crystallinity than PTB7-Th:IDBT-2F and more compact π-π stacking in the in-plane direction. 39 Employing naphthalene core is an effective strategy to enlarge the core size of FREAs as well. Zhan et al. compared benzene-and naphthalene-cored small molecular acceptors IDIC1 and IHIC1, respectively. It is interesting to figure out the GIWAXS results of two pure films that IDIC1 shows stronger (100) peaks in both in-plane and out-of-plane directions, while IHIC1 displays more distinct (010) peak in the out-of-plane direction. As for the blend films, the (010) peak of FTAZ:IHIC1 is sharper than the counterpart as well as larger CL value (2.2 nm vs. 1.9 nm), which indicate stronger π-π stacking in IHIC1 blend. The more favorable molecular packing of IHIC1 benefits the charge transportation, resulting in higher J SC (14.2 mA cm −2 ) and FF (66.4%) in FTAZ:IHIC1 based devices. Consequently, FTAZ: IHIC1 blend contributes to a better PCE (8.91%) compared to FTAZ:IDIC1 blend (7.05%). 41 However, excessive aggregation may lead inferior performance, which can be found in perylene diimide (PDI) based acceptors with overlarge core size. 15,42,43 Referring to Narayan's work, twisted perylene dimer can apparently disrupt ordered stacking compared to planar perylene molecule. The corresponding devices show significant J SC improvement from 0.8 to 9.5 mA cm −2 for twisted perylene dimer solar cells due to largely reduced aggregation. This results in great PCE enhancement from 0.13% to 2.78%. 44 According to this part, it is obvious that enlarging the area of conjugated plane is an efficient way to improve the molecular packing, especially π-π stacking. Furthermore, enlarging core size with heterocycle moieties such as thiophene will emphasize this effect due to their strong intermolecular interaction. 14,16 However, it is necessary to note that excessive core size will lead overlarge aggregation and molecular packing, which is definitely detrimental to the device performance. The appropriate core size should take into account of the synergetic effect of end group and side chain with balanced ordered packing.

| Heteroarene core modification
Heteroarene core modification is a useful way to regulate the ICT and optical band gap of NFAs. Zhu et al. reported a new Y-series acceptor AQx-1 with quinoxaline moiety (Figure 2A). 45 They further removed the electron-donating methyl groups on side substituents and synthesized AQx-2 ( Figure 2A). As shown in Figure 2B, both AQx-1 and AQx-2 prefer face-on orientation, while AQx-1 presents stronger π-π stacking with CL of 2.3 nm at 1.70 Å −1 . The authors noted that AQx-2 presents two separating lamellar stacking peaks, while AQx-1 only exhibits a major peak with a shoulder peak. In blend films, AQx-1 presents excessive aggregation in PBDB-TF:AQx-1, which would impede exciton diffusion. Meanwhile, PBDB-TF:AQx-2 blend shows well-defined phase separation according to transmission electron microscope (TEM) pattern. Therefore, the OSCs based on PBDB-TF:AQx-2 blends indicate higher PCE of 16.64% with J SC of 25.38 mA cm −2 and FF of 76.25% in terms of the J-V curves in Figure 2C. 46 One more case comes from a heteroarene core modification on Y6 by Zhu et al. They designed N-substituted acceptor SN, in which pyridine with methyl was introduced to create an asymmetry core structure. Compared with Y6, SN shows reduced but compact molecular stacking. Meanwhile, SN presents obviously red-shifted absorption and higher-lying highest occupied molecular orbital (HOMO) level. Such properties of SN can be not only used to improve device performance in PM6:Y6:SN ternary system, but also fabricated semitransparent solar cells with a good average visible transmittance (AVT) over 20%. 47,48

| Isomerization
Among the OSC materials synthesis, isomerization is a common strategy to improve photovoltaic properties as we know in investigation for fullerenes and PDIs acceptors. Zhan et al. designed two isomeric FREAs FNIC1 and FNIC2 which have the same side chains and end group but different core. Both of these two FREAs show quite favorable performances over 10%, while their molecular packings are distinctly different. The weak scattering features of FNIC1 in 2D GIWAXS pattern suggest its poor crystallinity, while obvious (100) and (010) scattering peaks at q r = 0.36 Å −1 and q z = 1.80 Å −1 can be observed in FNIC2 film. Such distinct difference indicates isomerization has obvious impact on molecular crystallinity. Additionally, in PTB7-Th based blend film, FNIC2 can cocrystallize with polymer donor, which results in higher mobility. The PTB7-Th:FNIC2 blend also shows smaller domain size of 21.6 nm. Consequently, the ultimate PCE of PTB7-Th:FNIC2 based device are higher than PTB7-Th:FNIC1 counterpart (13%-10.3%). 49 Yang et al. designed an asymmetrical isomer MeIC1 based on symmetrical NFA MeIC by replacing two thieno [3,2-b] thiophenes with thiophene and dithieno [3,2-b:20, 30-d] thiophene in the ladder-type core. Asymmetrical MeIC1 reveals stronger (010) peak than MeIC in pure films, suggested by smaller D-spacing (3.49 vs. 3.55 Å −1 ) and larger CL value (28.7 vs. 23.7 Å). This strong π-π stacking can be also fund in PBDB-T:MeIC1 blend film. Furthermore, interpenetrating network and large phase separation are available in PBDB-T:MeIC1 blend film. With such improved morphology, PBDB-T:MeIC1 blend presents a higher PCE of 12.58% than the one of PBDB-T:MeIC blend (12.03%). 50 In 2020, Sun et al. reported three NFAs Z1-aa, Z1-ab, and Z1-bb with isomeric core structures which synthesized through different the position of the cyclization reaction. It is apparent to observe the strong π-π stacking peak in pure Z1-bb film with face-on orientation rather than in pure Z1-aa and Z1-ab film. The CL values of Z1-aa, Z1-ab, and Z1-bb were calculated to be 1.0, 0.84, and 2.7 nm, respectively. In the blend films, PM6:Z1-bb also presents strong π-π stacking (CL = 2.7 nm) and face-on orientation among three blends. With favorable packing features, PM6:Z1bb blend device exhibits the highest PCE of 12.66% due to a significantly enhanced J SC of 18.52 mA cm −2 and FF of 70.05% compared to the other two isomer based devices. 51

| END GROUP
The electron-withdrawing end groups can effectively tune the optical bandgap, energy level and molecular packing for typical NFAs. It has been proved that the hidden long-range structure ordering along backbone direction by end-group π-π interaction is a valid packing formed in ITIC-based NFAs. 52 This end-group stacking extends the charge transport pathway, thus enhancing electron mobility. 53,54 Therefore, the endgroup modification could significantly improve stacking mode and further photovoltaic performance, which is specifically summarized in the following part.

| Halogenation
Introducing halogen atom into end-capped group is one of the most effective methods for tuning NFAs' properties. In particular, fluorination has been widely studied with remarkable effects on bathochromic absorption, electronegativity and the intermolecular interaction by noncovalent F-S and F-H bonds. 38 show both enhanced lamellar and π-π stacking. Difluorinated INIC3 presents the most crystalline scattering features among this series, suggesting stronger intermolecular interaction is formed after introducing two F atoms. In FTAZ:INIC3 blend films, the packing features of INIC3 remain, and FTAZ:INIC3 exhibits not only the strongest crystallinity but also the best PEC of 11.5%. 38 A similar fluorinated-induced ordering enhancement can be found in fluorinated ITIC-Th1 as well. The CL value of π-π stacking improves from 3.7 to 4.0 nm after introducing fluorine atom for ITIC-Th1-based blend film. As for the phase separation of blend films, the domain size decreases from 29 to 15 nm approaching to the exciton diffusion length. Such morphological improvements result in dramatic enhancement of PCE from 8.9% to 12.1% due to favorable J SC (19.33 mA cm −2 ) and FF (73.73%). Furthermore, both HOMO and LUMO levels of ITIC-Th1 shift down leading to a prominent lower V OC of 0.849 V. 57 The development of Y-series acceptors is a good illustration of fluorination effect as well. In 2019, Zou et al. reported a new NFA named Y5 with fused benzothiadiazole (BT) core and 1,1-dicyanomethylene-3indanone (INCN) end groups. This structure exhibits favorable photoelectric respond and temporally remarkable PCE of 14.1%, when blended with PBDB-T. 58 Based on the structure of Y5, they further introduced two F atoms on each INCN end group and synthesized one of the most famous NFAs, that is, Y6. The π-π stacking of Y6 is significantly increased after fluorinated, resulting in improved PCE over 15%. 19 The successful design of Y6 drew the attention, and the community of OSCs started to being committed to figure out the reason why such structure performs well. Several studies demonstrated that the perfect properties of Y6 originate from its polymer-like packing network, which is distinct from ITIC-based NFAs. 59 The Y6 molecules form this network through intermolecular interaction between fluorinated end groups and BT cores. Such extended conjugated structures greatly facilitate the charge transport in Y6 thin film. When blended with polymer donor PM6, the strong ordered packing of Y6 can be largely retained and form overlapped crystalline packing, which can improve the vertical charge transport. 60,61 Therefore, the fluorination in Y6 enables superior spatial charge transport and balanced charge carrier mobility, resulting in favorable device performance.
Besides fluorination, chlorination is another effective method to modify the FREAs photovoltaic properties, which is easy synthesizing and low-cost than fluorination counterpart. Hou et al. adopted chlorine-substituted strategy on a selenopheno[3,2-b]thiophene (SeT) based NFA (SeTIC), reporting an efficient acceptor SeTIC4Cl. The chlorination shows strong influence on lowering HOMO level from −3.90 to −4.08 eV for SeTIC4Cl. In thin film, neat SeTIC4Cl is relatively crystalline compared to the unchlorinated SeTIC. After blended with PM6, the original packing features retain in the as-cast films. However, PM6:SeTIC4Cl blend shows highly ordered packing after the addition of 0.5% DIO, which carried out the best performance. The results indicate the addition of DIO is beneficial to molecular packing, inducing better charge transportation and less recombination. 62 Then Chen et al. discussed multichlorination modification on their A-D-A type acceptor F-0Cl with the chemical structure shown in Figure 3A. Similar energy levels downshift can be observed as more chlorine atoms introduced. According to Figure 3B, the GIWAXS 2D patterns of F-nCl (n = 0, 1, 2), F-0Cl presents weak scattering peaks indicating less ordered packing. The introduction of Cl tremendously enhances the crystallinity of F-1Cl and F-2Cl which both exhibit sharp in-plane (100) and out-of-plane (010) scattering peaks. The π-π stacking distances obtained from (010) peak locations of F-0Cl, F-1Cl, and F-2Cl are 3.55, 3.45, and 3.41 Å, respectively, and the (010) CL values of F-nCl are 0.59, 5.67, and 6.43 nm, respectively. These data demonstrate chlorine atom can lead not only compact molecular packing but also long-range organization. In the DRCN5T based blend films, the phase separation is prominently increased as more chlorine introduced, and an interpenetrating network can be observed in the chlorinated films in terms of atomic force microscope (AFM) height images ( Figure 3C). With increased crystallinity and appropriate phase separation, devices based on chlorinated blend show notable PCE (from 5.49% to 9.89%) and external quantum efficiency (EQE) improvement shown in Figure 3D and 3E. 63 To find out the characteristics of chlorination compared with fluorination, Hou et al. investigated fluorinated IT-4F and chlorinated IT-4Cl simultaneously on multiple aspects. In general, chlorine has the stronger ability to descent the energy level of NFAs due to the unoccupied 3D orbital, which is available to the π electrons, as well as higher Cl-C bond dipole moment than fluorine resulting in larger intramolecular interaction. Compared to the energy level of IT-4F and IT-4Cl, chlorination presents stronger effect on down-shifting HOMO and LUMO levels of IT-4Cl. The GIWAXS results show apparent (010) peak of IT-4Cl in the out-of-plane direction, while the (010) peak intensity of IT-4F is relative weak. Meanwhile, they also compared F-substituted polymer PBDB-T-2F and Cl-substituted polymer PBDB-T-2Cl. Analogously, PBDB-T-2F shows stronger packing, which is resulted from large size of Cl atom. The different effect of fluorination and chlorination on molecular packing can be explained as different substituted position of the halogen atoms. 17,64 In addition, Chen et al. reported a series of halogenated acceptors (F-H, F-F, F-Cl, and F-Br) to further study the influence of halogenation. The GIWAXS results revealed that nonhalogenated F-H presents amorphous features, while halogenated F-F and F-Cl show visible (010) scattering peaks. Nevertheless, F-Br show lower crystallinity with respect to F-F and F-Cl, which can be accounted for the overlarge size and weak electronegativity of bromine atom. 56 After Y6 became a famous acceptor, Hou et al. also employed chlorination into Y6 molecule, and replaced the four F atoms with Cl atoms synthesizing BTP-4Cl. The chlorination shows similar effect on Y6-based NFAs, that is, the molecular packing of BTP-4Cl is slightly stronger than Y6 in both pure and blend films. The AFM and TEM show slightly improved phase separation of BTP-4Cl blend. Moreover, BTP-4Cl blend presents higher electroluminescence quantum efficiency and lower nonradiative energy loss, resulting in enhanced V OC of 0.867 V and higher PCE of 16.5%. 65 After that, Zhan et al. studied the effect of chlorination number, where nonchlorinated and single chlorinated acceptors were compared with BTP-4Cl. With two chlorine atoms on each end group, the largest down-shifting can be obtained for BTP-4Cl compared to BTP-2Cl-δ. Clearly, the crystallinity and the degree of phase separation are enhanced as more Cl atoms were introduced into NFAs. With optimized morphology, BTP-2Cl-δ performed better in OSC devices among different number of chlorination. 66 With these examples, we can conclude that introducing halogen atoms into the end-group dramatically downshifts the energy level and enhances the

| Methyl and methoxyl substitution
Owing to the weak electron-donating ability as well as moderately low steric hindrance, methyl and methoxyl are effective functional moieties to modify the energy level and intermolecular interaction of FREAs. Hou et al. reported a methyl-substituted strategy, which linked methyl and methoxyl to the end group of ITIC obtaining IT-M and IT-DM ( Figure 4A). The GIWAXS results of pure IT-M and IT-DM illustrated enhanced π-π stacking with respect to ITIC ( Figure 4B). Furthermore, IT-DM presents slightly stronger scattering peaks than IT-M, though their π-π stacking distances are similar  Figure 4C). However, PBDB-T presents slightly stronger lamellar stacking when blended with IT-M. Meanwhile, the relative domain purity of PBDB-T:IT-M is higher than the one of PBDB-T:IT-DM, which is evidenced by resonant soft X-ray scattering (RSoXS) measurements in Figure 4D. The improved molecular packing and domain purity of PBDB-T:IT-M is conductive to charge transport, leading to a higher FF. Consequently, PBDB-T:IT-M based OSCs show a slightly higher PCE (12.05%) compared to PBDB-T:IT-DM counterpart (11.29%) shown in Figure 4E. 67 Yang et al. focused on the number of methyl introduced into the end-group, and synthesized BTTIC-0M, BTTIC-2M, and BTTIC-4M with 0, 2, and 4 symmetric methyl, respectively. It is interesting to see the blue-shifted absorption and energy level upshifting for BTTIC-2M and BTTIC-4M. The π-π stacking CL values of three neat films were calculated to be 21.81, 23.79, and 25.53 Å, which suggests more methyl groups obviously improve the molecular packing. But in the blend films, the π-π stacking CL value of acceptor in PBDB-T:BTTIC-4M blend is significantly smaller than the one of PBDB-T:BTTIC-2M (36.30 vs. 41.59 Å). This implies the ordered packing of BTTIC-4M is broken by PBDB-T polymer due to the excessive mixing of donor and acceptor molecules. The phase separation was characterized by RSoXS, for which the domain sizes were calculated to be 18, 18, and 20 nm for BTTIC-0M, BTTIC-2M and BTTIC-4M blends, respectively. Owing to the smaller domain size and stronger π-π stacking, PBDB-T:BTTIC-2M based devices present the best PCE of 13.15% with outstanding FF of 75.3%. 69 Furthermore, Hou et al. carefully studied the four methoxyl modified isomers with different substituted position, which showed interesting changes on not only molecular packing but also photovoltaic property. 70 They first conducted quantum chemistry calculation indicating that 5-and 6-methoxyl substituted acceptors (IT-OM-2 and IT-OM-3) have excellent planarity with only 0.6°d ihedral angle. As for the 4-methoxyl substituted acceptors, IT-OM-1 has significantly larger dihedral angle of 18.9°, which is due to the large steric hindrance of 4-methoxyl substitution. It is also interesting to note that IT-OM-4 presents great distortion between methoxyl and phenyl. Then GIWAXS measurements delivered that IT-OM-1 displays weak (100) and (010) scattering peaks, while the other three isomers show obvious crystalline features. Meanwhile, IT-OM-2 shows the most compact π-π stacking distance (3.9 Å) as well as the largest CL value (2.93 nm). Consequently, PBDB-T:IT-OM-2 blend devices show the best PCE up to 11.9% with much higher J SC (17.53 mA cm −2 ) and FF (73%) than other blend OCSs.
The methyl substitution strategy is applied in Y6based NFAs as well. In 2020, Chen et al. introduced monomethyl to substitute two F atoms of the Y6 end group reporting BTP-M. As expected, BTP-M shows higher LUMO and HOMO level than Y6. The electrondonating methyl groups apparently decrease the intermolecular interaction, indicated by the thin films GIWAXS characterization. Only weak π-π stacking can be found in pure BTP-M film along out-of-plane direction. Although BTP-M presents less ordered molecular packing, it shows good miscibility with Y6 and forms acceptor alloy due to their similar chemical structures. This can also be confirmed by median energy level between Y6 and BTP-M. Therefore, BTP-M can be introduced as a third component (20% in weight) into PM6:Y6 blend, which optimizes the phase separation in the ternary blend. Owing to the optimized phase separation and component distribution, the corresponding devices show 17.03% PCE, obviously better than the one of PM6:Y6 control group (15.61%). We note that BTP-M with such molecular packing features is incapable of efficient charge transportation and exciton quenching in PM6-based binary blend. Thus, the binary blends only yield 4.26% PCE. Hence, these methylintroduced NFAs based on Y6 are suitable to be a third component for adjusting phase separation in PM6-based ternary system. 71,72 To conclude the effect of methyl and methoxyl substitution, it shows moderate end-group interaction and upshifting energy level compared to the halogenation, but a favorable miscibility with acceptors. Therefore, introducing methyl-/methoxyl-substituted NFAs can optimize the phase separation in ternary blend.

| Aromatic ring fusion
To further enhance the molecular packing along the π direction, Hou et al. extend the π-conjugated area by fusing a phenyl onto the terminal dicycanovinylindan-1one (DCI) group of IDTI and synthesized a new A-D-A acceptor IDTN ( Figure 5A). With phenyl extension, IDTN presents a higher HOMO level but deeper LUMO level than IDTI, resulting in a smaller band gap. As shown in Figure 5B, it is clear that IDTN shows remarkable in-plane (100) peak and out-of-plane (010) peak, while IDTI shows much weaker (100) and (010) scattering rings. The out-of-plane π-π stacking distance of IDTN (3.53 Å) is larger than the one of IDTI (3.55 Å). Meanwhile, IDTN presents distinct face-on orientation, while IDTI tends to adopt edge-on orientation. These results indicate extending the conjugated area with phenyl can not only facilitate π-π stacking but also induce a face-on texture, which is beneficial to the charge transportation. The strong ordering of IDTN also benefits the π-π stacking in PBDB-TF:IDTN blend, leading to a dramatically larger coherence length (9.4 nm) with respect to IDTI-based blend (3.1 nm). The domain spacing of IDTN-based blend film is confirmed by RSoXS and AFM to be 65.5 nm. In spite of a slightly larger phase separation size, the domain purity IDTN blend film is higher than that of PBDB-TF:IDTI blend, which can effectively decrease bimolecular recombination. As a result, the OSCs based on PBDB-TF:IDTN blend show 12.2% PCE ( Figure 5C), which is much higher than PBDB-TF:IDTI blend due to the prominent enhancement of FF (from 57% to 78%). 73 In 2019, Zhan et al. explored the terminal fusion effect on ternary OSCs with the same acceptor materials IDTI and IDTN. Introduction to IDTN brings stronger π-π stacking and smaller phase separation for ternary blend, which is similar to in the binary blend, and this kind of morphology is beneficial for obtaining a higher fill factor as well as efficiency of devices. 75 Thiophene fusion is another method to extend the end group. Based on the famous FREA ITIC, ITCC was synthesized with a thienyl-fused indanone as the terminal group. As sulfur atom is more easily polarized than carbon, the thienyl-fused indanone end-groups have stronger intermolecular interaction compared to ITIC. ITCC shows upshifted energy level, especially LUMO level shifting from −3.76 to 3.90 eV. The GIWAXS results of two pure films show that both ITIC and ITCC present moderate face-on stacking features, while ITCC has a more condensed π-π stacking distance than ITIC (3.7 vs. 3.8 Å), which agrees well with theoretical calculation using density functional theory (DFT). For blend films, PBDB-T:ITCC shows 3.83 Å stacking spacing, also smaller than the one of ITIC-based blend (3.93 Å). On the other hand, ITCCbased blend shows higher domain purity and larger domain size than that of ITIC-based blend. These characteristics facilitate the charge transport and suppress the bimolecular recombination. Consequently, the device based on PBDB-T:ITCC blend exhibits higher FF (71%), lower J SC (15.9 mA cm −2 ) and higher V OC (1.01 V), finally better PCE (11.4%). 76  As end-group extension obtained positive results on IDT-based NFAs, Marks et al. applied this strategy on BTP-based acceptors. Basing on the skeleton of Y5, they designed and synthesized two acceptors with linear and bent phenyl extension on their end groups, named BT-LIC and BT-BIC, respectively ( Figure 5D). Linear extended BT-LIC presents red-shifted absorption and higher lying LUMO level compared to Y5. Moreover, it is necessary to note the denser packing network and smaller end-group stacking distance of BT-LIC than the one of Y5 in terms of the single crystal analysis ( Figure 5E). Such packing features of BT-LIC result in higher electron mobility. However, bent extended BT-BIC shows significantly blue-shifted absorption and higher lying LUMO level, leading to very poor PCE of 7% for PBDB-T:BT-BIC blend. They further applied fluorination onto linear extended end group acceptor obtaining BT-L4F and BT-BO-L4F, which yielded favorable packing features as well as device performance for BT-BO-L4F based devices ( Figure 5G) Accordingly, aromatic rings fusion on end group can obviously improve the molecular stacking due to the π-π interaction between end-group rings. Nevertheless, rational ring type and extension direction should be taken into account to form an appropriate morphology for higher performance.

| SIDE CHAIN
The side chains of A-D-A type NFAs play a significant role in increasing the solubility and affecting the molecular stacking. These effects of side chains engineering can be attributed to size, conformation or distribution. In this section, we will illustrate some showcases from alkyl size, alkoxy/aromatic rings introduction and isomerization aspect.

| Alkyl size
The alkyl chain is the most commonly introduced moiety in sidechain engineering, which can alter the ordering of conjugated molecules and improve the solubility of the materials. With higher solubility, the phase separation can be optimized during solution processing. Zhan et al. modified the alkyl side chain size of 2-ethylhexyl IEIC synthesizing IEIC1, IEIC2, and IEIC3 which contain hydrogen, n-butyl and n-hexyl side chains, respectively ( Figure 6A). The 2D GIWAXS patterns of pure IEIC series films were shown in Figure 6C. It is pretty obvious to observe the in-plane (100) peak and out-of-plane (010) peak in IEIC1 film, which implies the face-on preferred orientation of IEIC1. As the alkyl side chain size increases, IEIC2, IEIC3 and IEIC show more mixed orientation and stronger lamellar stacking. When blended with PTB7-Th, the IEIC series blend films show similar scattering features as those in the pure films. The domain sizes of PTB7-Th:IEIC1, PTB7-Th:IEIC2, PTB7-Th:IEIC3 and PTB7-Th:IEIC blend films were characterized to be 31, 54, 18, and 29 nm, respectively, by RSoXS measurements ( Figure 6D). Moreover, the RSoXS profile of PTB7-Th:IEIC3 displays a sharp scattering peak and the largest scattering intensity integration, which suggest strong phase separation and large domain purity. As for the device performance ( Figure 6E), PTB7-Th:IEIC3 blend presents the best PCE of 6.9% among IEIC series OSCs due to the ordered packing and ideal phase separation. 78 Moreover, Heeney et al. replaced the 4hexylphenyl of ITIC with octyl and synthesized C8-ITIC. The GIWAXS results revealed that C8-ITIC presents strong scattering spots, indicating highly ordered molecular packing. As for ITIC, it shows much weaker crystallinity due to the large steric hindrance of phenyl group. In the blend films, apparent lamellar stacking peaks of both polymer donor and C8-ITIC can be observed in both PBDB-T:C8-ITIC and PFBDB-T:C8-ITIC blends, while polymer:ITIC blends only show (100) peaks of polymer donor. This suggests moderate phase separation of C8-ITIC blends rather than well mixed phases in ITIC blends. With enhanced ordered packing of C8-ITIC, the corresponding devices show higher J SC (19.6 mA cm −2 ) as well as PCE (13.2%) than ITIC based devices (J SC = 18.5 mA cm −2 , PCE = 11.71%) when blended with PFBDB-T. 79 A research of Sun et al. studied the influence of sidechain size on fused-ring backbone packing mode. They designed and synthesized three NFAs with the same backbone but different alkyl size, which are regarded as IDTT-C6-TIC, IDTT-C8-TIC, and IDTT-C10-TIC. Single crystal analysis illustrates the different packing modes for these three NFAs. IDTT-C6-TIC forms a 3D network with slip-stack backbone connecting by normally standing side chain, while IDTT-C8-TIC forms a crankshaft structure with octyl side chain interdigitating. For IDTT-C10-TIC, both harpoon and crankshaft type of side chain extension can be found with large distance π-π stacking. The different crystalline structures result in different molecular packing in thin films. Pure IDTT-C6-TIC film presents strong (001), (010)/(100), and (1-10) peak with random orientation, but relatively weak π-π stacking peak. IDTT-C8-TIC presents obvious in-plane (01-1), outof-plane (021), and (121) peak, meanwhile the π-π stacking peak is stronger. Pure IDTT-C10-TIC film presents a strong (112) and (224) peak corresponding to lamellar and π-π stacking, and it shows large π-π stacking distance (7.39 Å), which is due to the large alkyl chain spacing between adjacent molecules. In blended films, a similar trend can be observed that large sidechain size leads to stronger π-π stacking but weaker lamellar stacking. The authors also noted that IDTT-C10-TIC can form intermixed packing mode in thin film. As for the phase separation, the moderate domain size of 90 nm was detected for PBT1-C:IDTT-C8-TIC blend by RSoXS measurements. The proper size of alkyl chain in IDTT-C8-TIC enables fibril network and appropriate phase separation in the blend film. Therefore, this system show the best PCE (13.7%) with optimized J SC (20.3 mA cm −2 ) and FF (74.6%). 80 According to the work of Yan et al., they changed the Y6 branching position on the pyrrole motif to the third and fourth position, then synthesized two Y6-based NFAs N3 and N4. It is interesting to note that N3 shows an increased π-π stacking compared to Y6, evidenced by an increase of (010) CL from 24.0 to 27.5 Å. On the other hand, the N4 shows weak π-π stacking peak and enhanced lamellar stacking peaks. The difference can be attributed to the varied steric hindrance of branched side chains in different positions. For the phase separation of blended films, PM6:N3 and PM6:Y6 blends show similar domain size (22 nm) and domain purity, while PM6:N4 presents much larger domain size (56.1 nm) but lower domain purity. This can be attributed to excessive miscibility between PM6 and N4. As a result, PM6:N3 device show more favorable PCE of 15.98% compared to PM6:N4 OSCs (14.31%). 81 Accordingly, the alkyl size modification generally determines the steric hindrance of molecular which affects the complexity of ordered packing.

| Introduction of alkoxy/aromatic rings
Apart from alkyl chain, alkoxy and aromatic rings are in side-chain modification. As early as 2016, Hou et al. introduced alkoxy on IEIC and designed IEICO. This acceptor presents much red-shifted absorption in pure film and obvious phase separation in blend film, indicating stronger aggregation of IEICO than IEIC. As a result, IEICO-based OSCs obtained a much higher PCE (8.4%) due to significant enhancement of J SC and FF. 82 Zhan et al. reported an effective side-chain engineering strategy that uses oxygen atoms to control NFA properties, and substitutes hexyl of IOIC2 with hexyloxy yielding IOIC3. Compared with IOIC2, pure IOIC3 film was characterized to be more ordered packing with higher CL value. In the PTB7-Th:IOIC3 blend film, the π-π stacking CL is calculated to be 6.0 nm and the domain size is 9.1 nm, which are all larger than the one of PTB7-Th:IOIC2. The corresponding device exhibits an apparently higher J SC of 22.9 mA cm −2 , FF of 74.9% but lower V OC of 0.762 V, resulting in a better PCE of 13.1%. 83 Zhu and coworkers synthesized m-hexylphenyl side-chain ATT-6 and m-hexyloxyphenyl side-chain ATT-7. Interestingly, ATT-7 presents slightly blue-shifted absorption and increased crystallinity in blend film (larger π-π stacking CL of 5.68 nm after annealing), while the π-π stacking CL of ATT-6 blend is 5.09 nm. This may attribute to the difference between hexylphenyland and hexyloxyphenyl group. As for device performance, PBDB-T:ATT-7 blend shows higher PCE of 10.3% than ATT-6 counterpart (8.22%). 84 Considering the electron-donating and conformational locking effect of alkoxy groups, Yan et al. applied alkoxy substitution on the β position of thienothiephene unit of Y6 backbone, synthesized both asymmetric (Y6-1O) and symmetric (Y6-2O) molecules ( Figure 7A). As shown in Figure 7B, Y6-1O presents similar 2D scattering pattern to Y6, suggesting comparable molecular ordered packing (both π-π stacking CL values are 30 Å). Due to the strong conformational locking effect, Y6-2O shows extremely strong molecular ordering with high ordered lamellar stacking peak. In the blend films, PM6:Y6-2O exhibits higher degree of ordering than the one of PM6:Y6-1O and PM6:Y6 blends, which result in the largest phase separation. It should be noted that excessively strong ordering and phase separation are detrimental to the device's performance. Consequently, the device performance of PM6:Y6-2O blend is only 6.6%, while the PM6:Y6-1O blend device presents favorable PCE of 16.1%, even higher than the PM6:Y6 blend ( Figure 7C). 85 Furthermore, Y6-1O was reported to reveal good miscibility with Y6 yielding approaching 18% PCE in ternary blend, suggesting the effect of alkoxy modification. 86  The introduction of aromatic rings is also a common method for side-chain modification. In 2016, Zhan designed and synthesized a popular acceptor ITIC-Th, which is substituted the phenyl side chains of ITIC with 2-thienyl side chains. The thienyl side chains promote the intermolecular interaction as well as π-π stacking. Hence, the GIWAXS of ITIC-Th film shows more compact and longer range stacking, which were confirmed by the smaller d-spacing (ITIC-Th = 3.5 Å, ITIC = 3.8 Å) and larger (010) CL values (ITIC-Th = 5.5 nm; ITIC = 2.4 nm). The favorable packing features of ITIC-Th are beneficial to the charge transport. Thus, pairing with polymer donors (PTB7-Th and PDBT-T1), the blends show better photovoltaic performance with higher J SC as well as FF. 87 To extend the study of molecular packing of ITIC and ITIC-Th, Lu et al. reported a hidden long-range structure for ITICbased NFAs. As shown in Figure 8, ITIC-Th enhances the end-group stacking along backbone direction in PBDB-T:ITIC:ITIC-Th ternary blend, which can be hardly observed in PBDB-T:ITIC binary blend. Benefiting from this improved molecular ordering, the charge transport in the blended film is effectively facilitated resulting in better device performance. 52 In addition, Zhang et al. introduced m-BTP-PhC6 into PM6:Y6 blend, which is a phenyl-substituted NFA based on Y6. Clearly, m-BTP-PhC6 presents reduced π-π stacking in both pure and blend film. The lamellar stacking of ternary blend is apparently larger than PM6:Y6 and PM6:m-BTP-PhC6 blend, which results in a balanced electron and hole mobility. Consequently, an outstanding PCE of 18.12% for ternary device was fabricated with superior of FF 78.21%. 88 According to these representative works, it is obvious to find that the introduction of alkoxy has limited impact on molecular packing, which shows moderate regulation on absorption and charge transportation.

| Isomerization
Isomerization strategy can also be utilized in side-chain engineering to finely control the properties of FREAs. For the most well-known ITIC, the phenyl-alkyl side chains were initially introduced to prevent the excessive π-π aggregation. Li et al. designed an isomer m-ITIC, which is meta-position alkyl-phenyl, rather than paraposition for ITIC ( Figure 9). As shown in Figure 9B, the pure ITIC film shows weak lamellar and π-π stacking peaks, which result from the poor self-packing of paraposition substitution. As change to meta-position, m-ITIC shows an obvious scattering peaks for both (100) and (010). Compared to the CL value of two isomers, m-ITIC is significantly larger than ITIC as well (46.9 vs. 19.6 Å), indicating a higher crystallinity of meta-position substituted acceptor. 89  and out-of-plane (010) peak CL is 2.1 nm. The weak crystallinity of BTCN-M can be attributed to the large steric hindrance of twisted 3-and 5-substituted dioctylthiophene which prevents the π-π intermolecular interaction and highly ordered lamellar stacking. It is also interesting to be noticed that BTCN-O presents electron donor property, exhibiting 6.68% PCE when blended with PC 71 BM. However, BTCN-M works better as an acceptor when blended with PBDB-T showing PCE of 5.89%. This comparison suggests the dramatic influence of side chain modification on photovoltaic response. 90 Furthermore, the alkyl isomerization on the side chain is a reliable strategy to tune the FREAs. McCulloch et al. compare the liner side-chain (n-octyl) O-IDTBR with branched side-chain (2-ethylhexyl) EH-IDTBR. Both X-ray diffraction (XRD) and differential scanning calorimetry (DSC) measurements revealed that O-IDTBR is a crystalline small molecular acceptor. But EH-IDTBR is almost amorphous, which indicates its poor molecular packing due to branched side chain. When blendedwith P3HT, O-IDTBR show polycrystalline rings rather than diffraction spots, suggesting the presence of P3HT lower the ordering of O-IDTBR molecules. In P3HT:EH-IDTBR blend, a new peak from acceptor appeared in out-of-plane direction as P3HT changed the packing order of EH-IDTBR. Such structural features of IDTBR bring facile manipulation of morphology as indicated elsewhere. 31,32 As a result, P3HT:O-IDTBR based device shows slightly better PCE (6.3%) compared to P3HT:EH-IDTBR blend (6.0%) due to the higher J SC of 13.9 mA cm −2 . 91 For the isomerization in Y6-based acceptors, Sun et al. designed and synthesized branched alkyl chain L8-BO compared to linear-chain Y6. According to the crystal structure solving, L8-BO presents ellipse voids between packing molecules, while Y6 forms orthorhombic vacancies. This difference originated from the branched side chains of L8-BO interact with each other spatially, while the side chains of Y6 interact within the conjugated plane. Thus, more impact molecular arrangement can be observed in L8-BO film. The GIWAXS characterizations of pure thin films indicate stronger (021) scattering peaks along out-of-plane direction for both pure L8-BO and blend PM6:L8-BO thin films. The higher structural ordering of L8-BO brings significant performance improvement from 16.61% to 18.32% with remarkable FF up to 81.5%. Moreover, Yan et al. swapped the positions of nundecyl and 2-ethylhexyl alkyl chains of Y6 molecule, and synthesized a novel NFA N-C11. It exhibits poor solubility and overlarge phase separation, which is detriment to the device performance. 81 Therefore, rational isomerization design of NFA structure can effectively alter the molecular packing mode and optimize the phase separation for high efficiency OSCs.

| CONCLUSION AND OUTLOOK
In summary, the advent of nonfullerene small molecules greatly enriches the acceptor choice to combine with polymer/small molecular donors, which promotes the performance of blend film OSCs. To optimize the OSC devices, a proper morphology of active layer thin film is necessary after D/A selecting consideration of complementary absorption and matched energy level. The insight of NFAs' chemical structures can realize the difference in morphology as well as device performance. In terms of different functional moieties, the discussion is conducted in core, end group and side chain.
(1) The fused-ring central cores tend to form intermolecular interaction, in particular π-π interaction, due to large planar structures. With large size core, the long-range ordering and large scale aggregation are easy to observe in thin film. The over-large packing and aggregates adverse to exciton dissociation yielding lower FF. To prevent this case, appropriate core size design, heterocycle introduction and rational isomerization are useful methods. (2) For end-group engineering, halogenation is an effective way to enhance electron-withdrawal and intermolecular interaction. Especially, end-group stacking has been reported to be a universal form of π-π stacking in state-of-the-art NFAs. Fluorination and chlorination are widely utilized to broaden near-infrared (NIR) absorption and increase the molecular packing for higher charge mobility. At the meanwhile, end-group extension by aromatic fusion can extend conjugated area and moderately increase the intermolecular interaction. As for methyl substitution on end-group, it show reduced intermolecular interaction but can be usually used as a third component to improve the compatibility in ternary system, which will facilitate charge separation. (3) As alkyl side chain is introduced to increase the solubility of a molecule, it is important to control the alky size to promise a rational aggregation and phase separation. Alkoxy side chain is likely to form conformation lock or weak interaction, which helps to increase the conjugated aggregation. Aromatic side chain with large size increases the steric hindrance to prevent the molecular aggregation, while side-chain isomerization affects the molecular conformation as well. Owing to the flexibility of the side chain, the molecular packing and aggregation will be regulated then affect device performance.
As summarized above, clear dependences of NFA structures can be concluded to correlate with morphology and performance according to numbers of publications. Therefore, designing novel molecular structure of NFA can be regarded as an effective method to promote the development of OSCs as the advents of ITIC and Y6 did to this field. Currently, these state-of-the-art NFAs have NIR absorption range, high absorption coefficient and ordered molecular packing contributing to superior device performance. However, the large conjugated core size and various molecular modifications of these NFA molecules increases the difficulty of synthesis and cost of device fabrication. We carefully considered the developing trend denoting to realize the commercial application of NFA-based OSCs. The expectation of new NFAs is for easily synthesized and highly efficient. It will be an expected challenge to rationally design a simple core structure with sufficient light harvesting, appropriate aggregation and morphology in thin films. Meanwhile, the V OC loss of OSCs is still stronger than high efficiency inorganic or perovskite solar cells, though the J SC and FF are approaching to the competitors'. The band gap of the new NFA should be finely tuned with suitable D and A units then matched with polymer donor. In addition, the prospective NFA needs to have good processibility in green solvents and tolerance to ambient condition and film thickness fluctuation during large-area printing. Moreover, the device stability of efficient polymer: NFA blends need to be further improved under illumination and thermal stress, which may be realized in the new type of NFA. With this review, we believe the correlation will guide the material synthesis, morphology optimization and device fabrication towards higher efficiency OSCs in the near future.