The molecular structure of Y6 is shown in Fig.1a; it features a banana-shaped A-D-A'-D-A backbone configuration with alternating electron deficient (A, A') and rich (D) moieties. For the prototypical PM6:Y6 system, addition of 0.5vol% CN (Fig.1a) and thermal annealing at 80°C are common device fabrication conditions capable of delivering decent PCEs of ~ 16%6. To understand the specific role of the CN additive in the crystalline packing of Y6, we first measured the GIWAXS patterns of a pure Y6 film prepared with 0.5vol% CN. Intriguingly, as shown in Fig.1b, the GIWAXS pattern displays discrete and sharp Bragg peaks at (qr, qz) = (0.22, 0), (0.44, 0), (0.22, 0.48), (0.22, 0.62), (0.28, 0.18) and (0.46, 0.31) Å−1, as well as the p-p peak at (qr, qz) = (0, 1.79) Å−1, pointing to the formation of long-range ordered crystalline domains of Y6. On the other hand, thermal annealing at the optimized temperature of 80°C does not induce any notable changes in the crystalline packing motif of Y6, as evidenced from the nearly identical GIWAXS patterns with or without thermal annealing (Supplementary Fig.1–2). To confirm that the observed crystalline packing structure of Y6 is also present in the actual device active layer, we measured the GIWAXS patterns of PM6:Y6 blend films with 0.5vol% CN and 80°C annealing. Note here that, in this study, chloroform (CF) is used as the solvent for solution processing of all thin films, consistent with the reported fabrication condition of the high efficiency PM6:Y6 device active layer6. To consider the dilution effect of PM6 in the blend film, we gradually increased the mass ratio of PM6 towards the optimized D/A ratio of 1:1.2. The 2D GIWAXS patterns and the corresponding intensity profiles are presented in Fig.1c-f. At the D/A ratio of 1:4, the scattering peaks from the p-p and lamellar stackings of PM6 present at qz = 1.68 Å−1 and qr = 0.29 Å−1, respectively (Fig.1c-d and Supplementary Fig.3), which are expected to largely overlap with the scattering signals from Y6. Remarkably, however, the two feature scattering peaks of crystalline Y6 along qr can still be identified at qr = 0.21 and 0.42 Å−1, implying that the incorporation of PM6 does not significantly alter but slightly loosens the crystalline packing of Y6. When the PM6 concentration increases, stronger PM6 signals are observed, while Y6 signals are further weakened but can still be identified at the same positions, as indicated by the triple-peak fitting of intensity profiles in the small qr region in Fig.1d-f. Therefore, we infer that the crystalline packing motif observed in the pure Y6 film remains in the PM6:Y6 blend films fabricated under actual device conditions. To further demonstrate that CN plays here a critical role in the crystalline packing of Y6, GIWAXS patterns were measured for PM6:Y6 blend films with a fixed D/A ratio of 1:1.2 and a varying CN content. As shown in Supplementary Fig.4, it is clear that when the CN content gradually increases to 2vol%, the two feature scattering peaks of crystalline Y6 at (qr, qz) = (0.22, 0) Å−1 and (0.44, 0) Å−1 can be clearly identified in the scattering pattern of the blend film, indicating the direct impact of CN in the formation of Y6 crystallites in the BHJ film. To elucidate the crystalline packing structure of Y6 in the active layer, the GIWAXS pattern of the pure Y6 film with 0.5vol% CN was indexed in detail. As shown in Fig.2a, the scattering peaks (qr, qz) = (0.22, 0), (0.44, 0), (0.22, 0.48), (0.22, 0.62), (0.28, 0.18) and (0.46, 0.31) Å-1 can be indexed with a monoclinic unit cell and lattice parameters a = 29 Å-1, b = 16 Å-1, c = 29 Å-1, α = 90°, β = 90° and γ = 122° (Fig.2b). The square ac plane lies in the in-plane (IP) direction and the rectangular bc plane is 32° inclined with respect to the out-of-plane (OOP) direction. The corresponding Miller indices are labeled next to each peak. After resolving the unit-cell structure, we considered in detail the molecular arrangements within the individual unit cells. First, as presented in Fig.2a, the intense p-p stacking peak of Y6 is concentrated along the qz axis; therefore, the Y6 molecules within the unit cell should possess a predominant face-on orientation with the molecular backbone horizontally aligned with respect to the substrate. In addition, due to the averaged p-p distance of 3.41 Å (qz = 1.84 Å-1, Supplementary Fig.2d), the unit cell should fit four face-on oriented Y6 molecules along the b axis. The molecular architectures of a single Y6 and of typical Y6 dimers have been proposed in previous theoretical calculations20, which are summarized in Supplementary Fig.5. Based on the resolved lattice constants and the reported dimer configurations, we propose the following arrangement of the four Y6 molecules within the unit cell, as illustrated in Fig.2b-e: From bottom to top of the unit cell, the configurations of Y6 dimers are “arm-chair”, “zig-zag” and “arm-chair” with terminal-terminal (TT) overlap of the end groups. Between neighboring unit cells, the Y6 molecules are connected through “herringbone”-shaped core-core and terminal-terminal (CC-TT) overlaps, as well as TT dimer configurations. Periodic lattice structure can then be appropriately formed by extending the unit cell along the crystallographic primary axis directions. Monoclinic crystalline Y6 has also been reported in previous single-crystal studies34–35. Nguyen and co-workers made significant efforts to correlate the single-crystal structure of Y6 to its crystalline packing in thin film35; they compared the calculated atomic chemical shifts in the single crystals with those measured in Y6 thin films and used this comparison to evaluate how well the molecular packing in thin films is similar to that in the Y6 single crystal. Here, we directly observe the crystalline structure of Y6 in the active-layer thin film via GIWAXS measurements. The comparison of the crystalline packing structure of Y6 in the actual active layer proposed in this work and the single-crystal structure reported earlier34 is given in Supplementary Fig.6. We note that the details of the lattice parameters and molecular configurations are different, which is attributed to the crystallization kinetics impacted by solvent (CF) and additive (CN) during formation of the solution-processed active layer. As demonstrated above, it is clear that the solvent additive CN impacts the crystal packing of Y6 in the active layer, a feature that has not been reported to date. Although the influence of solvents on the single-crystal structure of Y6 has been highlighted before35, the crystallization mechanism of Y6 in a thin film in the presence of an additive is still unclear. Therefore, we choose to investigate at the molecular level how CN can influence the crystal structure of Y6 in a thin film, via atomistic molecular dynamics (MD) simulations. We note that the MD simulations primarily consider Y6 and the additives but do not include the solvent CF, which is due to the fact that while the solvent additives with high boiling point (~ 300°C at one atmosphere) can remain at least partially in the thin film after spin coating, the solvent CF with much lower boiling point (~ 65°C at one atmosphere) tends to evaporate quickly during film casting. Thus, in fact, our MD simulations try to understand the crystallization mechanism of Y6 during film casting and after the solvent volatilized. As shown in Fig.3a and Supplementary Fig.7, in the Y6/CN mixture, in addition to the typical p-p packings among the Y6 end groups, there also exist p-p interactions between CN and the Y6 end groups, which is confirmed by the radial distribution function (RDF) g(r) data (Fig.3b and Supplementary Fig.8–9); here, a higher g(r) peak implies a larger packing density at a given distance r. As presented in Supplementary Fig.8, at the local level, the g(r) intensity between Y6 and CN is largest at distance of ~ 0.4 nm at both 300 K and 353 K, which is in the range of p-p stacking distances. Further calculations reveal that it is the g(r) intensity between CN and the end group (A moiety) of Y6 that is strongest compared to the other moieties (Fig.3b and Supplementary Fig.9), implying the p-p interactions mainly originate from CN and the end groups of Y6. MD simulations were also conducted for pure Y6 and the mixture of Y6 with another additive 1,8-diiodoctane (DIO) for comparison (Supplementary Fig.10–12). There are no additional interactions in the Y6/DIO mixture, which is attributed to the alkyl nature of DIO instead of the aromatic nature of CN. As presented in Supplementary Fig.13, the calculated interaction energies (sum of the electrostatic and van der Waals forces) between CN and Y6 are also larger than between DIO and Y6. We note that the interaction energies between Y6 molecules in the existence of CN or DIO are rather similar, suggesting that the interactions between CN molecules and Y6 end groups are not detrimental to the self-assembly of Y6 molecules. We hypothesize that the additional p-p interactions induced by the conjugated nature of CN act as a bridge to gather Y6 molecules pre-aligned in the p-p stacking direction in solution and can thus induce the formation of a highly ordered crystalline packing structure during film formation, as seen from the GIWAXS measurements. A schematic illustration of the role of the CN additive in the crystallization of Y6 is given in Supplementary Fig.14. We also exploited RDF data to further characterize the interactions between the CN moieties and the end group of Y6; the CN molecule is considered as two moieties, one corresponding to the chlorine-substituted benzene ring moiety (CN1) and the other to the unsubstituted benzene ring moiety (CN2); the end group of Y6 is labeled as A1 (difluoro-benzene) and A2 (substituted five-membered ring) moieties (see Supplementary Fig.15). It is found that the A2-CN1 (or CN2) g(r) intensity is the largest among all moiety pairs (Supplementary Fig.16), indicating preferential interactions between A2 and CN, especially CN1. This result suggests that, besides the conjugated rings of CN and the end groups of Y6, the cyano groups in the terminal groups of Y6 and the halogen substituents in the conjugated additives may also contribute to the additional interactions between Y6 and CN. These results provide a good level of guidance for future molecular design of NFAs and additives, on the way towards the ideal molecular interactions and crystalline packing of the NFAs. As the configurations of the side chains can possibly also affect the packing of the NFA molecules via steric hindrance effects38, we evaluated the orientation distributions of the Y6 side chains with various additive conditions based on the above MD simulation results. As shown in Supplementary Fig.17, we define the normal vector of the end group as the main direction and then consider the angle distributions of the various side chains, evaluated over all Y6 molecules in the MD box (Supplementary Fig.18–19). In the case of pure Y6 without additive, the probabilities for all three types of side chains to appear on either side of the end group are rather similar. Interestingly, for the case of Y6 with CN or DIO, the linear side chains tend to appear on only one side of the end group, unlike the branched long chains or branched short chains, which have no preferential orientation around the end group. This specific distribution of the linear side chains may also facilitate an ordered packing of the Y6 molecules by reducing steric hindrance, which is consistent with the more efficient charge transport in devices based on PM6:Y6 with CN or DIO, as discussed below. Since both the crystalline and amorphous regions of Y6 in the thin film actually consist of various dimers, which have a significant impact not only on the formation of crystalline structure but also on the overall transport properties of the active layer20,22, we further investigated the influence of the additives on the dimer configurations by quantifying the probability distributions of various dimers. Figure3c presents the four typical configurations of dimers found in the MD simulations: (ⅰ) terminal-terminal (T-T) configurations; (ⅱ) core-terminal (C-T) configurations; (ⅲ) core-core (C-C) configurations; and (ⅳ) other configurations among which the core/core-terminal/terminal (CC-TT) overlapping configuration is the most representative. As shown in Fig.3d, the incorporation of solvent additives indeed impacts on the population distributions of the Y6 dimers. Although the proportion of T-T configuration is the largest (~ 50%) and similar for all six different cases we considered, the case of Y6 with DIO leads to an obvious decrease in the proportion of the C-C configuration. According to a previous report, the C-C interactions contribute significantly to large electronic coupling for exciton transport22; here, the decreased proportion of C-C configurations for Y6 with DIO can result in slower exciton diffusion rates, which is confirmed by the transient absorption data on PM6:Y6 blend films in the next section. Moreover, as displayed in Supplementary Tables1 and 2, it is found that for the cases of Y6 with the CN additive at 300 K and 353 K, the proportion ratio of CC-TT to T-T configurations (~ 1:3) is close to that of the molecular arrangement we proposed above for the unit cell of Y6, which confirms its validity. According to previous reports, smaller p-p stacking distances between moieties of neighboring molecules can result in larger electronic couplings within the dimers, which would contribute directly to superior transport properties in the active layer36. Based on the MD simulation results, we calculated the statistical p-p stacking distances for CT and TT dimers that account for the majority of Y6 dimer configurations. As presented in Supplementary Fig.20, at both 300 K and 353 K, Y6 with the CN additive has the smallest packing distance compared to pure Y6 and Y6 with DIO, which is also consistent with the trends in p-p peak positions observed by GIWAXS below (Supplementary Fig.1–2 and Supplementary Table3). This confirms once more the fundamental influence of the CN additive on the crystalline packing of Y6 and consequently on the device characteristics. In the following, we elucidate the impact that the Y6 crystal structure we have uncovered in the active layer thin film has on the device characteristics. We have considered two additional common device fabrication conditions: without additive and with 0.5vol% DIO. The GIWAXS patterns of the Y6 films without additive or with 0.5vol% DIO (Supplementary Fig.1–2) present qr profiles with similar peak positions at 0.21 Å−1, 0.28 Å−1 and 0.42 Å−1; this implies the existence of a similar crystalline packing motif for Y6 in these films despite of a much lower degree of ordering. The p-p peaks in these two films (Supplementary Fig.1d) appear at qz = 1.75 Å−1 (d-spacing = 3.59 Å), a smaller value compared with the p-p peak position of the CN-processed film (qz =1.79 Å−1, d-spacing = 3.51 Å); thus, CN appears to promote a tighter molecular packing and a better ordered crystalline packing structure for Y6. Furthermore, the p-p peak of the film without additive is the weakest and concentrated along the qz axis, while that of the DIO-processed film is more intense but spans a larger polar angle. This suggests that the film without additive has less crystallinity but favorable face-on orientation while the DIO-processed film has higher crystallinity but more random orientation; in contrast, CN can simultaneously improve both the long-range ordering and the orientation order of Y6. The GIWAXS patterns of the PM6:Y6 blend films fabricated with the same conditions (Supplementary Fig.21) show exactly the same trends as the neat films. GIWAXS of the pure PM6 films with various additives and the grazing-incidence transmission small-angle X-ray scattering (GTSAXS)40 of the PM6:Y6 blend films have been measured and displayed in Supplementary Fig.22–23; the data demonstrate that the crystalline packing in PM6 and the nanophase separation in PM6:Y6 are not altered notably under these three device fabrication conditions. Therefore, the device characteristics we describe below can be mainly attributed to the distinct crystalline packing of Y6 in the active layer. Figure4a and Table1 present the device characteristics of OSCs based on PM6:Y6. Compared to the device without additive, the Jsc and FF values are simultaneously enhanced for devices with 0.5vol% CN or DIO additives, which contributes to an increase in PCE. The improvement in Jsc and FF for the devices with CN are essentially due to the highly ordered crystal structure of the Y6 phase revealed above, which results in better optical absorption and charge transport efficiency. The solvent additive DIO, which has a high boiling point, can extend the evaporation time of the solvent chloroform during film casting; this contributes to increase the crystallization time of Y6 and improve its crystallinity in blend films, as evidenced by the higher intensity of the qr peak at 0.28 Å−1 in the GIWAXS pattern of Y6 with DIO compared with that of Y6 without additive (Supplementary Fig.1–2), which should also help enhance Jsc and FF33. Interestingly, compared with the device without additive, the Voc value of PM6:Y6 device with DIO decreases strongly from 0.849 V to 0.812 V, whereas Voc in the device with CN additive increases slightly from 0.849 V to 0.854 V. This opposite trend in Voc is possibly related to the distinct crystal orientations of Y6 induced by the various additives (Supplementary Fig.1–2 and 21), with the less ordered orientation of Y6 with DIO resulting in increased energetic disorder and energy loss within the BHJ active layer18 (the energy losses will be discussed in detail in the following section). The integrated Jsc values from the EQE spectra of the corresponding devices are highly consistent with the measured values from the J-V curves within less than 3% error (Fig.4b, Table1), suggesting the reliability of the J-V measurements of the devices. Compared with devices without additive or with CN, the device with DIO has a red-shifted absorption edge (Fig.4b), which can originate from a broadened distribution of the density of states caused by the orientational disorder in the Y6 crystallites33. The statistical analysis of the PCE values confirms that the PM6:Y6 binary devices fabricated with the CN additive possess superior performance (Fig.4c), which is attributed to the simultaneous improvements in Jsc, FF, and especially Voc, due to the distinctive long-range ordering and orientation order of Y6 with CN, indicating the direct influence of the Y6 crystalline morphology on the OSC device characteristics. Charge recombination in the various devices were evaluated via J-V tests under various light intensities (Plight). The relationship between Voc and Plight is described as Voc ∝ nkT/qln(Plight) in which q, k, and T denote the elementary charge, Boltzmann constant, and Kelvin temperature, respectively. As shown in Fig.4d, the slopes are fitted to be 1.07, 1.19, and 1.05 kT/q for the OSCs without additive, with DIO, and with CN additive, respectively. These results suggest that the addition of DIO dramatically increases monomolecular recombination while the CN additive tends to suppress those; we can thus expect difference in the energy loss mechanisms, as will be confirmed below. Figure4e presents Jsc vs Plight plots for the three cases; they are very similar with close-to-unity slopes (α), indicating negligible bimolecular recombination. The charge collection efficiency of the devices has been investigated via the dependence of the photocurrent density (Jph) on the effective internal voltage (Vint). The photocurrent density (Jph) is defined as Jph = JL – JD, where JL represents the current density under illumination and JD is the current density in the dark. The charge collection probability (PC) can be calculated according to PC = Jph/Jph,sat, where Jph,sat is the saturated photocurrent density at high Vint (defined as Vint = V0 – V, where V0 is the voltage when the net current is zero, and V is the applied voltage); ηdiss and ηcoll are defined as the PCs under short-circuit and maximum power-output points, respectively, and refer to the exciton-dissociation and charge-collection efficiencies. Figure4f shows the plots between Jph and Vint. The ηdiss and ηcoll values are simultaneously enhanced for OSCs with DIO or CN compared to the devices without additive, which should again originate from the improved crystallinity and contribute to the higher Jsc and FF values in the corresponding OSCs. We now turn to the impact of the intrinsic crystalline packing structure of Y6 on the fundamental photophysical properties of the OSCs during the photon to electron conversion processes. The hole-transfer kinetics from Y6 to PM6 within the BHJ active layer was studied via transient absorption (TA) spectroscopy, with the Y6 phase exclusively excited by a 750 nm laser excitation. As illustrated in Supplementary Fig.24, similar ground-state bleaching (GSB) signals at ~ 850 nm are observed for the Y6 pure films under various additive conditions. For the PM6:Y6 blend films, the corresponding bleaching peaks of Y6 also appear immediately after photoexcitation (Fig.5a-d, Supplementary Fig.25). With an increase in delay time, the bleaching signal of Y6 gradually decays, while a new bleaching peak arises at ~ 640 nm, which is attributed to the donor PM6 according to its absorption spectrum. The evolution of the two peaks is indicative of the hole transfer process from Y6 to PM6. The hole transfer kinetics of the different PM6:Y6 blend films were analyzed from the profiles of the PM6 bleaching peak intensity (I) versus delay time (t). Typically, the hole-transfer process consists of an ultrafast hole-transfer process and a diffusion-mediated process, which can be fitted by a biexponential function: $$I={A}_{1}\text{e}\text{x}\text{p}(-\frac{t}{{\tau }_{1}})+{A}_{2}\text{e}\text{x}\text{p}(-\frac{t}{{\tau }_{2}})$$ 1 with lifetimes τ1 and τ2 and pre-factors A1 and A2. The ultrafast charge-transfer process indicated by τ1 should be attributed to D/A interfacial charge-transfer processes while the NFA exciton diffusion mediated process described by τ2 is generally dominated by the BHJ morphologies41,42. As illustrated in Fig.5e-f, the PM6:Y6 blend film with DIO appears to present simultaneously the largest time constants τ1 and τ2. In contrast, PM6:Y6 blend film with CN exhibits the shortest time constants for both τ1 and τ2. The average hole-transfer times (τh) are calculated with τ1, τ2 and their respective proportions (A1, A2); as illustrated in Fig.5f, the τh value for the PM6:Y6 blend film with DIO is about 2 ~ 3 times longer than that of the blend film with CN. The remarkable hole transfer efficiency of the blend film with the CN additive is expected to be related to the molecular ordering of Y6 formed with the assistance of CN, as confirmed by the molecular dynamics simulations discussed above for the dimer distributions; this larger extent of molecular ordering allows faster exciton diffusion rates, while the much slower hole-transfer rate of PM6:Y6 with DIO is possibly due to the less ordered orientations of Y6 crystallites. Overall, the addition of CN promotes the molecular ordering of Y6, which enhances Jsc in the device via an improvement in the charge-transfer dynamics (rather than via a red-shift of the absorption edge), unlike the situation in the device with DIO. According to the results of earlier calculations, the distinct molecular packing of Y6 favors not only significant exciton couplings between adjacent molecules, which facilitates the exciton diffusion process, but also reduces exciton-vibration couplings and generally leads to slower non-radiative decay rates20. Now, we seek to further characterize the specific energy loss mechanisms in the OSCs. Supplementary Fig.26 and Supplementary Table4 collect the detailed energy loss analysis of the relevant devices. First, compared to the devices without additive, the addition of CN increases the bandgap (Eg) slightly from 1.400 eV to 1.405 eV. The addition of DIO decreases the bandgap from 1.400 eV to 1.385 eV, which should contribute to an improvement in Jsc but also a lowering of Voc for devices prepared with DIO. In addition, the radiative energy losses above (ΔE1) and below (ΔE2) bandgap are similar for the various devices. As for the non-radiative energy loss (ΔE3), compared to the devices without additive, while there is almost no change after addition of CN, the incorporation of DIO significantly increases the ΔE3; this can be attributed to the Y6 intrinsic dimer distributions in the presence of DIO, as shown by the MD simulations as well as to the lower orientational order. These results confirm the profound impact of molecular packing on the non-radiative voltage loss. The crystalline structure of Y6 in the active layer of an actual solar cell device has been resolved in detail via GIWAXS measurements, which reveal a highly ordered 3D molecular network. The use of 1-chloronaphthalene (CN) as a solvent additive is found to play a crucial role in the formation of a highly ordered crystalline structure of Y6 in thin films. Molecular dynamics simulations show that the intermolecular interactions between the CN molecules and the Y6 end groups are the underlying driving force that facilitates the ordered packing of Y6 and ultimately results in the distinct crystalline structure that Y6 adopts in thin films. The specific crystalline packing structure of Y6 we have uncovered is shown to lead to fast charge generation and efficient carrier transport while maintaining low energy losses, which accounts for the superior device performances. Table 1 Photovoltaic parameters of OSCs based on various PM6:Y6 blends. Average parameters with standard deviation in parentheses were obtained from 10 individual devices. Devices Voc (V) Jsc (mA cm− 2) FF (%) PCE (%) Jcala (mA cm− 2) w/o additive 0.849 (0.848 ± 0.001) 25.36 (24.86 ± 0.36) 74.62 (73.83.±0.82) 16.05 (15.56 ± 0.25) 24.73 with DIO 0.812 (0.807 ± 0.002) 27.77 (27.61 ± 0.15) 77.10 (76.33 ± 0.36) 17.41 (17.02 ± 0.17) 26.94 with CN 0.854 (0.849 ± 0.003) 27.23 (27.17 ± 0.29) 77.20 (76.27 ± 0.84) 17.87 (17.57 ± 0.20) 26.47Crystalline packing structure of Y6 in thin films
Molecular Dynamics Simulations
Correlation With Device Characteristics And Photophysical Properties
Synopsis
a Integrated current densities from EQE curves
See Also
Y6 and its derivatives: molecular design and physical mechanismNon-fullerene acceptors with branched side chains and improved molecular packing to exceed 18% efficiency in organic solar cells