Exploiting the donor‐acceptor‐additive interaction's morphological effect on the performance of organic solar cells

Organic solar cells (OSCs) have demonstrated over 19% power conversion efficiency (PCE) with the help of material innovation and device optimization. Co‐working with newly designed materials, traditional solvent additives, 1‐chloronaphthalene (CN), and 1,8‐diodooctane (DIO) are still powerful in morphology modulation towards satisfying efficiencies. Here, we chose recently reported high‐performance polymer donors (PM6 & D18‐Fu) and small molecular acceptors (Y6 & L8‐BO) as active layer materials and processed them by different conditions (CN or DIO or none). Based on corresponding 12 groups of device results, and their film morphology characterizations (both ex‐situ and in‐situ ones), the property‐performance relationships are revealed case by case. It is thereby supposed to be taken as a successful attempt to demonstrate the importance and complexity of donor‐acceptor‐additive interaction, since the device performance and physics analyses are also tightly combined with morphology variation. Furthermore, ternary blend construction for PCE improvement provides an approaching 19% level and showcases the potential of understanding‐guided‐optimization (UGO) in the future of OSCs.

and then understanding-guided-optimization (UGO) can be realized. [19]22][23][24][25][26][27][28][29][30] This is attributed to the selective interaction of additives on donor or acceptor materials, and slower evaporation process, that are crucial to form suitable sized domains with high purity without sacrificing the general well-intermixing donor-acceptor networks. [31,32]Two common solvent additives, 1-chloronaphthalene (CN: inducing H-aggregation) and 1,8-diodooctane (DIO: inducing J-aggregation), [33] are still assisting cutting-edge PCEs, although in recent years novelty in additives has been massively extended.Considering the broad acknowledgement, good reproducibility, and decent effectiveness of these two solvent additives, the mechanisms behind are supposed to be clearly understood.[36] In other words, the field currently has very limited knowledge on the donor-acceptor-additive interaction: a combined concept referring to all interactions exist in blend systems.
Herein, we select representative donor and acceptor materials: PM6, D18-Fu, Y6, and L8-BO, that are reported high PCEs, [37][38][39][40][41] to have a systematic study by cross-matching them with/without CN/DIO treatments.The main solvent is selected to be chloroform (CF) for its good solubility and fast evaporation, where the role of additive is significant and easy to be figured out. [42]Thereby, 12 groups of devices are fabricated and the active layers are characterized through in-situ; and ex-situ morphological techniques and device physics digging.Apart from their already effects of inducing H-and J-aggregation, CN and DIO's film morphology modulation pathways are found dependent on material-material interaction: less crystalline and pre-aggregated PM6 exhibits better miscibility with Y6 and L8-BO, while D18-Fu's stronger aggregation tendency and well-formed fibrillar network rarely depends on matching acceptors or processing additives.The PM6:Y6 blend can achieve ideal length-scale intermixing region and suitably sized pure domain morphology in film without using additive, while the use of CN and DIO leads to smaller pure domains but less ordered molecular stacking.The PM6:L8-BO pair with even better miscibility realizes large intermixing phase in the film from CF-only solvent, for which using CN and DIO both decrease the intermixing region's length scale, and keep the original pure domain size, where the crystalline order is not so much tuned.As for D18-Fu:Y6 system, CN extracts more small molecules from polymer matrix to form a pure phase than DIO does, but generates a larger intermixing phase length scale.Different scenario in D18-Fu:L8-BO combination tells that CN can induce significant pure domain growth, without very much enhanced intermixing region enlargement.The molecular packing order for D18-Fu-based films is not too much changed by additive treatment.These morphology results are consistent with device parameter variation and device physics data.Further vertical morphology inspection and insitu film formation observation confirms the conclusion and figure out the origin of different evolutions of the systems treated by additives.With such complete morphology char-acterizations, the miscibility analysis as the last puzzle is incorporated, and contributes to understanding the formation of each system's morphology, which further implies proper amount guest component addition will not destroy the already favorable morphology, thus possibly efficient ternary blend benefitted from energy transfer and energy level tuning.At last, the UGO enabled ternary matrix construction based on these materials produces excellent PCEs in both D18-Fu:L8-BO (CN) + PM6 and PM6:L8-BO (DIO) + D18-Fu-based solar cells.

RESULTS AND DISCUSSION
The chemical structures of active layer materials are drawn in Figure 1A, while their ultraviolet-visible (UV-vis) absorption and photoluminescence (PL) spectra are shown in Figure 1B,C.D18-Fu exhibits a blue-shift absorption but a red-shift PL spectrum compared with PM6 (larger Stokes shift thus greater energetic disorder), which means its wider bandgap might not offer desired open-circuit voltage (V OC ).
On the other hand, L8-BO demonstrates a blue-shift absorption spectrum, as well as more significant 0-1 shoulder peak than Y6, which implies further additive treatment on blend solution may result in different packing mode of them.Their PL spectra order similarly to absorption profiles, accordingly less differences energetically between them.Then the blend films of them, with or without additive treatments are also measured by these two tests, whose results are displayed in Figures S1 and S2.Simply blending results in significant blue-shift of acceptor(s) related absorption edge and peaks, representing their characteristic packing or aggregation shall be suppressed in active layers.In comparison, polymeric donor materials show less affected absorption peak positions, especially D18-Fu, whose absorption profile shapes are almost identical in neat and blend films.Therefore, typical molecular packing and aggregation of donor polymers are supposed to be better kept in active layers, for which in details: D18-Fu better than PM6.These results also indicate D18-Fu is less miscible with acceptors than PM6 is, which is probably attributed to its strong pre-aggregation and crystallinity.Additives (CN and DIO) redshift the PL spectra of blend films in most cases, implying tuned film crystallinity and phase separation also leads to enlarged energetic disorder.This gives the reason why in recent years, additive engineering becomes a hot topic of suppressing energetic disorder and voltage loss, for the bottleneck breakthrough of OSC performances, since traditional solvent additive cannot do so.The UV-vis and PL data can further combine with subsequent morphology analyses, and property-performance relationship discussion.
Before sharing the device performances, a brief summary of cutting-edge PCEs with the use of CN or DIO, is shown as Figure 1D.Corresponding detail information is listed in Table S1.These two traditional and widely used solvent additives are still capable to realize ideal morphology that demonstrate leading efficiencies for OSCs.Thus, there is sufficient motivation for us to explore the donor-acceptor-additive interaction based on them.
Then, a series of conventional structural devices (ITO/ PEDOT:PSS-TA/ active layer/ PNDIT-F3N/ Ag) [43] are fabricated based on these blend systems.The current density versus voltage (J-V) characteristics are plotted in Figure S3, with the photovoltaic parameters listed in Table S2.The efficiency variation is visualized as a heatmap in Figure 1E, as well.It shows complicated changes without a simple tendency, thus implying the strong impact of revealing the material-material-solvent interaction.PM6:Y6 systems requires no special additive treatment to achieve >17% efficiency, which is the same scenario for D18-Fu:L8-BO combination.Even more, the use of DIO is negative to the PCE.On the contrary, PM6:L8-BO's photovoltaic performance can be significantly enhanced by DIO, and D18-Fu:Y6 shows the same phenomenon.The V OC s are generally reduced by CN and DIO, as a result of tuned crystallization and aggregation, which has been widely noticed by the field. [44]The enhancement of J SC s due to additive treatment is significant as well, but for D18-Fu:L8-BO blend CN performs best while DIO is the better choice for others.The FF variation exhibits even more complicated distribution that PM6:Y6 additive leads to decreased values, PM6:L8-BO embraces significant promotion, and D18-Fu-based systems show very slightly enhancement.Overall, the parameters vary irregularly once all morphological and physical factors are not characterized, but these data provide a chance of better understanding the working mechanism for OSCs.
Next, related device physics analyses are implemented in order.The photocurrent versus effective voltage (J ph -V eff ) curves are presented in Figure S4, and the deduced saturated current density (J sat ), exciton dissociation and collection efficiencies (η diss & η coll ) are displayed in Table S3.The calculation details are elaborated in Supporting Information.
Here we found all (η diss & η coll ) values vary insignificantly, thereby a very close charge extraction process for all blends.This makes the comparison of J sat s more meaningful, which can directly correspond to material's aggregation motif and phase separation.In Figure 1F, the J sat distribution for each system is clear to claim that DIO's inducing redshift could contribute to best charge generation in most cases.Some exceptions such as D18-Fu:L8-BO could be the result of undesired morphology evolution.Meanwhile, the proper addition of CN can monotonously improve the J sat s for all photoactive systems.
The charge recombination is then another important factor to be correlated with FF.Accordingly, light intensity dependent V OC & J SC curves are tested, plotted (Figures S5  and S6), analyzed (Table S4), and visualized (Figure 1G).The results especially those of PM6:L8-BO perfectly follow the FF changing tendency and significance: suppressed trapassisted and bimolecular recombination.For other systems, the vague differences of their FFs corresponds to insignificant discrepancies of fitted S and n values (details of calculation method are discussed in Supporting Information).Therefore, as another tool to evaluate the recombination, mobilities of films that representing charge transport are assessed by space charge limited current (SCLC) method based on measuring J-V curves of hole-only and electron-only devices.The raw results are shown in Figures S7 and S8, and calculated results are in Table S5 and Figure 1H (method also demonstrated in Supporting Information in detail).Through corresponding results, more information and conclusions are obtained: DIO results in more unbalanced charge transport, thereby decreased FF for PM6:Y6; the hole/electron transport balancing is achieved in all other systems by CN or DIO, and the more balanced, the higher FFs for them; both additives increase the absolute mobility, but CN would make the electron mobility outperforming the hole mobility.
To have a more consistent and reliable charge generation and recombination dynamic study, femtosecond (fs) transient absorption spectroscopy (TAS), a reported powerful tool, becomes our choice.The representative spectra of neat films are given as Figure S9; 2D contour maps of blend films and signals probing the polarons are presented in Figure 2. The fitting results of charge generation lifetimes are shown in Table 1.After the charge transfer and dissociation of singlet excitons, there will be an increasing population of free charges that will lead to the rising kinetics of 520-560 nm (for D18) and 570-600 nm (for PM6), which are dominantly probing the hole polarons population at the corresponding donor molecules.It can be observed that the rise dynamics upon DIO treatment appear slower for all cases aside from D18-Fu:L8-BO.According to previous studies, the disorder-induced interfacial morphology will impart a cascade energy landscape at donor and acceptor interfaces, which will slow down the free charge generation but will assist in a more efficient process. [45]As a consequence, the energy cascade-driven charge separation contributes to the improved J sat .Likewise, since the DIOtreated D18-Fu:L8-BO exhibits the opposite trend relative to what is expected for the energy cascade-driven charge separation, its J sat has decreased.These suggest that donoracceptor-additive interactions can also regulate the properties of intermixing phases where free charges are being formed following singlets charge separation.By turning into the influence of CN treatment on PM6:L8-BO and D18-Fu:Y6, it can be observed that the rise kinetics are very similar with the blends without additive treatments, thereby only capable of slightly influencing the J sat when compared to DIO treatment.However, for the case of PM6:Y6 and D18-Fu:L8-BO, the interplay of pure domains nanomorphology can also be realized probably due to the more significant additive-assisted size evolution of such domains than the intermixing phase.More specifically, the PM6:Y6 exhibited smaller pure domains thereby the shorter exciton diffusion process prior to the charge separation will lead to faster free charge generation.The presence of smaller domains is known to increase the interfacial areas responsible for charge separation thereby is expected to also increase J sat , as observed herein.Through the same principle, as D18-Fu:L8-BO displays the opposite behavior of pure domain growth, its free charge generation tends to be slower.
After free charge generation, their recombination behavior as described by the polarons decay dynamics will be critical in modulating the device FFs.From the previous discussions, the CN treatment of PM6:Y6 and D18-Fu:L8-BO are identified to have more substantial additive-assisted domain size evolution, which is responsible for the density molecular interface area, the site where bimolecular recombination losses originate.Hence, tracking their sub-ns polarons recombination shows increasing rates for PM6:Y6 while it is the opposite for D18-Fu:L8-BO.When considering the DIOtreatment, the free charge recombination displays slower rates for all cases.It must be noted, however, that there are more significant changes with PM6:Y6 and D18-Fu:Y6.Interestingly, the dissociation rates are also found to be more evidently slower for these blend combinations, which is previously noted as the consequence of the interface energy cascade landscape.Likewise, since DIO-treatment in D18-Fu:L8-BO is previously found not to impart better energy F I G U R E 2 2D contour maps of transient absorption spectra of blend films excited at 800 nm, and corresponding polarons dynamics probed at 520-560 nm (D18 polarons) and 570-600 nm (PM6 polarons).
cascade configurations but only limited to slightly larger pure domains, it exhibits minor reduction recombination rates.It can then be speculated that the donor-acceptor-additive interactions influence the interface energy cascade, which will then impact not only the charge separation process and device Jsc but also critical for the free charge recombination in the sub-ns range (i.e., mostly bimolecular in nature) defining the device FF.However, these metrics are also highly dependent on the interplay of nanomorphology and molecular packing known to regulate the diffusion of singlet excitons and the long-range transport of polarons above a few ns.Hence, the discussion is combined with morphology data that will be elaborated in the following content.
Parallelly, the energy loss analysis is carried out upon all binary systems.The Fourier transform photocurrent spectroscopy (FTPS) external quantum efficiency (EQE), bandgap assessment (dEQE/dE), and electroluminescence (EL) EQE, are presented in Figure S10, with derived results in Table S6.Generally, non-radiative loss values are enlarged by additive(s) treatment, except PM6:L8-BO (DIO) system.This also indicates that additive treatment upon the blend systems could show different results in altering bandgap and energy loss values, as a result of crystalline behavior difference, determined by initial additive-free blend film morphology, which further suggest donor-acceptor-additive interaction should be taken as an unseparated studying target.The Urbach energy (E U ) values of all solar cells are evaluated by fitting FTPS-EQE tail states, from which no clear tendency could be recognized, suggesting the complexity of donor-acceptor-additive interaction, and the claim of solvent additive treatment's possibility in causing additional energetic disorder.
The morphology investigation begins with the grazing incidence wide-angle X-ray scattering (GIWAXS) experiments on neat films and processing-condition-varied blend films.The 2D patterns of PM6, D18-Fu, Y6, and L8-BO pure films are given in Figure S11, while those of blend films are demonstrated in Figure S12.Related in-plane (IP) and out-of-plane (OOP) line-cuts of all samples are depicted in Figure 3A.Via gaussian fitting, the lamellar and π-π stacking peaks could be located, as well as their coherence lengths (CLs).These results are summarized in Tables S7-S10, and for a more obvious comparison of crystallographic property, portrayed in Figure 3B.While PM6 exhibits distinct face-on orientation and moderate crystallinity, D18-Fu displays a much stronger crystallinity and more ordered molecular arrangement (dspacings 21.4 Å versus 20.7 Å & CLs 112.4 Å versus 19.7 Å for IP directional lamellar peak).Both Y6 and L8-BO contain two peaks in lamellar region alongside IP direction, locating at 0.29/0.41and 0.27/0.39Å −1 , respectively, which is consistent with literature reports.Considering the π-π stacking peak parameters of donor and acceptor materials: [46,47] PM6 has at most 2 (CL/d-spacing = 5.26/3.78)layer packing orderly; D18-Fu on the contrary exhibits 31.0/3.60 ≈ 9 ordered layer; Y6 shall contain 38.8/3.56 ≈ 11 layers of molecules stacking together; and L8-BO corresponds to 16.2/3.65≈ 4 layers.Thereby, the crystallinity and ordering issue of blend films based on PM6 donor polymer can be simplified as π-π stacking behaviors dominated by acceptors, while those based on D18-Fu are coherent results.
The PM6:Y6 blend film free from additive exhibits insignificant vice peak at lamellar region, which demonstrates their good miscibility (vanished independent crystallization of small molecule).After CN treatment, the π-π peak rises a lot, yet displays poorer stacking order, as well as slightly promoted characteristic vice lamellar peak at 0.42 Å −1 .More crystallites contribute to photon absorbing while less ordering explains the FF loss.As for DIO treated film, the intensity of π-π peak goes down very significantly, as well as a much more decreased CL, thereby not so high J sat even with more J-aggregates, and the lowest FF compared with other two counterparts.
For PM6:L8-BO system, additives cannot induce new vice peaks in lamellar region as clear as those of PM6:Y6.CN and DIO reduces the CLs of π-π stacking peaks from 30.5 Å to 29.8 Å and 19.5 Å, which seems to be negative of crystallite order and charge transport.However, pure L8-BO film only exhibits 16.2 Å CL for this peak.This means the higher CLs after blending is contributed by PM6 since it is also very miscible with L8-BO.By DIO's crystallization regulation effect, L8-BO can have better independent aggregation, thus demonstrating a lower CL, which means a purified domain should be presented.This explains why its best FF and excellent J sat .Turning the focus on D18-Fu:Y6 system, the characteristic peaks of Y6 are weakened, too.Considering the strong crystallinity and pre-aggregation of D18-Fu, this might be a combined result of donor-acceptor intermixing and polymer matrix trapping.The use of CN leads to slightly tighter and more ordered π-π stacking, and DIO makes lamellar vice peak show up and higher π-π peak CL.Both pathways lead to similar FF increase.The differences on V OC and J SC are mainly attributed to DIO's J-aggregation inducing effect.
Last system comes to D18-Fu:L8-BO blend.Similarly, no distinguishable vice peaks at lamellar region are observed.The π-π peak CL values vary not too much, thus there does not exist phase purification after additive's incorporation.The enhancement of CL values caused by CN and DIO can be then taken as a simple ordering improvement for a better charge transport.
Apart from <10-nm scale molecular packing analysis, the phase separation behaviors (>10-nm scale) of all active layers are also investigated, enabled by grazing incidence small angle X-ray scattering (GISAXS) experiments.The obtained 2D-patterns and IP intensity profiles are given as Figure S13 and Figure 3C.Using a fitting model reported elsewhere, [48] length scales of amorphous intermixing phase (X DAB ) and crystalline pure domain (2R g ) can be point out.The fitting lines and deduced length scale values are put in Figure 3C, too.PM6:Y6 exhibits very similar X DAB and 2R g , with c.a. 20 nm an ideal scenario for charge generation and transport.CN and DIO reduces pure domain length scale from 18 to 16 nm.Relatively smaller pure domains are in consistence with FF loss.Initial PM6:L8-BO exhibits too large X DAB of 53 nm versus the constant 19 nm 2R g .CN and DIO continuously reduce intermixing phase size, which makes the pure domain relatively larger, thus boosted FF.Additives uplift the X DAB /2R g values of D18-Fu:Y6, that in a rational range, is useful to support the enhancement of J SC .Same perspective explains the phase separation changes of D18-Fu:L8-BO and related J sat s.The smallest X DAB /2R g value of DIO treated one well support its higher FF and unsatisfied J sat .
Subsequently, the vertical phase segregation in active layers is compared by the film-depth-dependent light absorption spectroscopy (FLAS). [49]The raw spectra are displayed in Figure S14, based on which the donor/acceptor ratio curves are drawn in Figure 4A.Furthermore, Figure S15 offers 2D contour maps of calculated charge generation rates at different depth and wavelength, and integrated curves are plotted in Figure 4B.The results confirm that PM6 is well miscible with both Y6 and L8-BO, from the aspect of vertical distribution.Meanwhile, D18-Fu is assured less miscible with small molecular acceptors.Additive treatment is supported being useful to promote the phase separation as well.Fortunately, the calculated charge generation curves are consistent with the J sat variation.These ex-situ morphology characterization results can be evidenced by atomic force spectroscopy (AFM). [50]The captured phase and height images of neat and blend films are displayed from Figures S16-S20.The surface roughness of all films is controlled in a similar level, demonstrating a good contact between electron transport layer (ETL) and active layer.D18-Fu induces more significant fibrillar network than PM6 does in films when blending with acceptors.Incorporated solvent additives generate aggregates locating at the gap of these fibers, like stations of railways.
Co-solvent effect in film formation process tuning is then investigated by in-situ UV-vis absorption measurement. [51,52]he wavelength-time-absorbance-based 2D contour maps of the processes are drawn in Figure S21, from which 0-1 peak and 0-0 peak of donors and acceptors are found.Their posi-tions and intensities are plotted in Figure 5.According to the peak red-shifting and intensity decrease/increase, whole process can be divided to be four stages: (i) gray background refers to the fast removal of CF, where only the absorption intensity drops quickly, but peak position rarely changes; (ii) main solvent on substrate approaching to none (polymer donors finishing deposition, small molecular acceptors are still being removed), which is marked by blue background; (iii) both donor and acceptor materials show their aggregation, evidenced by either peak red shift or absorbance improvement, emphasized as orange region; (iv) film formation is mostly completed, only some DIO remains in film, but no peak shift or intensity change.Based on this concern, some analysis upon each blend can be implemented: additives shorten the duration of first three stages, especially stage (iii), thus smaller pure domain is gained; CN and DIO don't reduce the time of aggregation in PM6:L8-BO, therefore the dominant effect here is the phase purification, reflected as the decreased intermixing phase length scale; D18-Fubased systems possess faster film formation, probably due to donor polymer's stronger crystallization tendency, wherein both additives elongate the aggregation period (CN longer than DIO), well consistent with the enlarged length scales of pure domains.
After complete in-situ and ex-situ characterization of nanomorphology, the structure performance relationship is supposed to be revealed.Therefore, a miscibility analysis enabled by contact angle measurement is carried out. [53]he results are demonstrated in Figure 6 and Table S11, respectively: droplet photographs, contact angles, and derived interaction parameters.With the support of thermodynamic side, the claimed donor-acceptor-additive interaction and how it determines film morphology can be understood in a clearer way.The normal preaggregation of PM6 and its mediate miscibility against Y6 / poor miscibility against L8-BO lead to stable pure phase size.Because of strong interaction between L8-BO and additive, the intermixing region is significantly reduced.In contrast, D18-Fu provides a more complicated scenario due to its intrinsically strong pre-aggregation, and good miscibility with Y6 and L8-BO.These features correspond to the well-distributed length scale of pure and mixed phases.On the other hand, CN's strong interaction with both acceptors and D18-Fu enlarges these two indexes at the same time, while this effect of DIO is less significant.One step further, we understand that though the material-material interaction that mainly relies on thermodynamic issue (miscibility) can be easily changed by adding a third component, but those binary cells treated by additives are not the same.Their morphology is co-determined by miscibility and the interaction between additive and material, which is a closer-to-equilibrium state, and adding a guest component shall cause ignorable morphology change once its ratio is beneath the threshold.Having such understanding, the performance improvement strategies based on binary systems are foreseeable: PM6:L8-BO-DIO system properly introduced D18-Fu is able to maintain its original morphology, and vice versa for D18-Fu:L8-BO-CN system with minorly added PM6.Such ternary blend can benefit from energy transfer between two polymer donors, thus higher J SC .
In addition to clarified donor-acceptor-additive interaction case by case from CN and DIO processed systems, another typical solvent additive diphenyl ether is chosen to further support the existence and meaning of exploiting this phenomenon (0.5% vol in CF).The device results are presented in Figure S21 and Table S12.Similarly, more different cases in photovoltaic parameter distribution are presented here, endorsing our declaration of donor-acceptor-additive interaction's existence, complexity, and research meaning.
Based on the data collected in SZTU, complete and in-depth understanding has been achieved about how donoracceptor-additive interaction modulates film morphology and device performance.At the last step, two series of ternary cells as an understanding guided optimization (UGO) is fabricated in PolyU, that are D18-Fu as guest component in DIO treated PM6:L8-BO and PM6 incorporated in CN enabled D18-Fu:L8-BO host blend.The J-V characteristics are illustrated in Figure 7A,B.Corresponding device performance parameters are listed in Table 2 after extraction.Because of the existence of energy transfer between two donor polymers, J SC values of 20% guests incorporated systems are boosted compared with their binary counterparts. [54]Meanwhile, ternary blend construction only causes very marginal change for V OC and FF in both series of solar cells, thus the PCEs are enhanced.Interestingly, here we demonstrate an optimal-ratio-reversed case in ternary photovoltaic systems simply caused by varied additives.It can be naturally concerned as a typical UGO case for OSC performance  improvement and as well as a further phenomenon emphasizing the important role of donor-acceptor-additive interaction in determining device efficiency.Notably, such results are at state-of-the-art level for those OSCs with benzo-difuran corebased polymer donors, which is elaborated in Figure S22 , Table S13. [55,56]

CONCLUSION
In summary, two polymer donors (PM6 + D18-Fu), two small molecular acceptors (Y6 + L8-BO) and two traditional solvent additives (CN + DIO) are included to be our research targets, and corresponding 12 groups of binary OSCs are fabricated to compare the performance systematically.Simultaneously, physical features revealed by TAS and morphological parameters characterized by various measure-ments (GIWAXS, GISAXS, FLAS, AFM, in-situ UV-vis) are provided based on active layer thin films.The results showcase that changing either material or additive will alter the film morphology and thus device efficiencies, thus their combined concept donor-acceptor-additive interaction should be comprehensively considered in future OSC studies.In addition, we also suggest that donor-acceptor-additive interaction co-determined morphology can be maintained when proper ternary strategy is applied, which could be a new train of thought of pursuing OSC performance.At last, such UGO case is achieved, in which part approaching 19% PCE is gained.As a complementary study for recent rapid efficiency development of OSCs, this work could provide more insights to the field from the aspect of device optimization.

C O N F L I C T O F I N T E R E S T S TAT E M E N T
The authors declare no conflict of interest.

F
I G U R E 1 (A) Chemical structures of target donor and acceptor materials in this work.Normalized (B) ultraviolet-visible (UV-vis) absorption spectra and (C) photoluminescence (PL) profiles of these four materials in film state.(D) State-of-the-art efficiencies achieved by CN and DIO according to recent literatures.(E) The heat map contains efficiency comparison of 12 group of devices.(F) the variation of J sat values of different systems treated by typical ways.(G) The fitted S and n values for bimolecular recombination and trap-assisted recombination evaluation.(H) The calculated hole and electron mobilities.

F
I G U R E 3 (A) The line cuts main peaks locating at in-plane lamellar region and out-of-plane.(B) Calculated d-spacing values and coherence lengths.(C) Grazing incidence small angle X-ray scattering (GISAXS) in-plane line cuts and fitting lines/parameters of blend films, as well as inserted derived length scales of pure phase and intermixing phase.

F
I G U R E 4 (A) Deduced donor/acceptor ratio curves.(B) Calculated depth related charge generation rates.

F I G U R E 5
Time dependent position and intensity curves for all active layers.

F
I G U R E 6 (A) Water (upper row) and ethylene glycol (lower row) droplet pictures on pristine films.(B) Captured contact angle values.(C) Calculated interaction parameters for miscibility.(D) Visualized donor-acceptor intrinsic and interaction features.

A
C K N O W L E D G M E N T S G. L. thanks the support by the Research Grants Council of Hong Kong (project numbers: 15221320 and C5037-18G), the RGC Senior Research Fellowship Scheme (SRFS2223-5S01), the Shenzhen Science and Technology Innovation Commission (JCYJ20200109105003940), the Hong Kong Polytechnic University Internal Research Funds: Sir Sze-yuen Chung Endowed Professorship Fund (8-8480), RISE (Q-CDBK), G-SAC5, 1-YW4C and the Guangdong-Hong Kong-Macao Joint Laboratory for Photonic-Thermal-Electrical Energy Materials and Devices (GDSTC number: 2019B121205001).G. Z. thanks the support from the Guangdong Basic and Applied Basic Research Foundation (2022A1515010875), (2021A1515110017), Natural Science Foundation of Top Talent of SZTU (grant number: 20200205), and Project of Education Commission of Guangdong Province of China (2021KQNCX080).R. Ma is thankful for the support through the PolyU Distinguished Postdoc Fellowship (1-YW4C).
Fitted charge generation lifetime from TAS spectra.
TA B L E 1 The brackets contain averages and standard errors of PCEs based on 20 devices. Note: Investigation and formal analysis: Lu Chen.Conceptualization, supervision, project administration, investigation, formal analysis, methodology, writing-original draft, and writing-review and editing: Ruijie Ma.Formal analysis and conceptualization: Jicheng Yi.Investigation, formal analysis, and methodology: Top Archie Dela Peña.Investigation and methodology: Hongxiang Li.Investigation: Qi Wei.Resources: Cenqi Yan, Jiaying Wu, Pei Cheng, Mingjie Li, and He Yan.Resources, supervision, and funding acquisition: Guangye Zhang.Resources, supervision, and funding acquisition: Gang Li.