What We have Learnt from PM6:Y6

Over the past three years, remarkable advancements in organic solar cells (OSCs) have emerged, propelled by the introduction of Y6—an innovative A‐DA'D‐A type small molecule non‐fullerene acceptor (NFA). This review provides a critical discussion of the current knowledge about the structural and physical properties of the PM6:Y6 material combination in relation to its photovoltaic performance. The design principles of PM6 and Y6 are discussed, covering charge transfer, transport, and recombination mechanisms. Then, the authors delve into blend morphology and degradation mechanisms before considering commercialization. The current state of the art is presented, while also discussing unresolved contentious issues, such as the blend energetics, the pathways of free charge generation, and the role of triplet states in recombination. As such, this review aims to provide a comprehensive understanding of the PM6:Y6 material combination and its potential for further development in the field of organic solar cells. By addressing both the successes and challenges associated with this system, this review contributes to the ongoing research efforts toward achieving more efficient and stable organic solar cells.

This Review, comprising both retrospective analysis and forward-looking perspective, delves into these inquiries, presenting state-of-the-art insight but also unresolved disputes in the PM6:Y6 blend.We start with describing the design principles behind PM6 and Y6, where we outline what properties render these materials particularly suited for bulk heterojunction (BHJ) OSCs.We discuss how the molecular building blocks of PM6 foster polymer chain aggregation and why the unique A-DA'D-A structure of Y6 leads to extended near-infrared absorption as well as strong intermolecular interactions.We then introduce the excited states and processes involved in the photovoltaic action of the PM6:Y6 blend.Here, we highlight some unique properties of Y6, such as its long exciton lifetime but also comment on the ongoing controversy about the D/A energy offset or the mechanisms of free charge generation.In addition, we delve into the examination of charge recombination within photovoltaic devices, underscoring the criticality of extended carrier lifetime and the significance of reducing non-radiative decay to maximize the fill factor (FF) and open-circuit voltage (V OC ).Furthermore, we highlight the significance of optimal doping concentration in enhancing charge generation, as well as the influence of dopants in modifying material morphology.Moreover, our discussion encompasses V OC , with a particular focus on the impact of hot carriers and en-ergetic disorder.Then we provide a detailed picture of the blend morphology on length scales ranging from several micrometers to sub-nanometer.Here, we aim at a physical understanding of how certain solvents, post-processing treatments, and additives alter the morphology and subsequently the device efficiency.We finally address the remaining challenges towards commercialization, where we discuss important issues such as device stability, materials batch-to-batch variation, and module efficiency.

Design Principles Behind PM6 and Y6
Currently, state-of-the-art organic solar cells are usually equipped with the BHJ active layer of small molecule acceptors (SMAs) and polymer donors, with single junction binary PCE exceeding 19%, which is an amazing leap from the first BHJ device with 2.9% PCE in 1995. [6]Conspicuously, this great improvement cannot be achieved without continuous optimization and collaborative promotion of donor and acceptor materials.Before 2015, more structure regulation focused on polymer donors due to the limitation on structural modification of fullerene acceptors (FAs).During this period, D-A (D denotes a donor unit and A denotes an acceptor unit) copolymers have gradually become a research mainstream because of the excellent photoelectric properties given by their multiple D-A push-pull interactions.Later, when donors were optimized to a certain extent, it was realized that the FAs with limited tunability of the absorption range and the frontier orbital energy to match the appropriate donor was the bottleneck limiting the further improvement of OSCs.Here, the emerging narrow bandgap SMAs broke the logjam.
[53] The BDT unit in PBDB-T consists of two parts, the benzodithiophene core and the thiophene side chains.The conjugated benzodithiophene ensures the planarity of the D-A backbone.The side thiophene unit [54] cannot only enhance interchain - interactions, but chemical modifications also allow for fine-tuning morphology, energetics, and finally the photoelectric properties.For example, PM6 was designed by attaching two fluorine atoms to the conjugated thiophene side groups of the BDT unit to lower the energy level of the highest occupied molecular orbital (HOMO), which is beneficial for increasing the V OC of the device.The BDD unit, on the other hand, as a well-known A unit, has a perfect symmetric skeleton with good planarity, which ensures the delocalization of charge and tight intermolecular packing of the segment.In addition, the carbonyl group on the BDD unit with strong electronwithdrawing ability can effectively lower the HOMO level.At the same time, there is a non-covalent interaction between the oxygen atoms on the carbonyl group and sulfur atoms on the nearby thiophene, [55] enhancing the planarity as well as light absorption.
The extended conjugations and enhanced rigidity of the BDT and BDD units guarantee strong - stacking and aggregation of the polymer both in solution and in the solid film. [56,57]As shown in Figure 1c, solution-processed BHJ films of PBDB-T congeners exhibit a fibrous morphology.Hereby, a nanoscale phase-separated domain structure is formed with various acceptors, with a typical domain size ˜20 nm that is suitable for exciton dissociation. [50,58]It was further proven that PBDB-T congeners retain their ordered aggregation and face-on orientation in the blend (Figure 1d), which facilitates efficient charge extraction.As we discuss in greater detail in Section 5.1, the packing of PM6 in BHJ blends with Y6 is rather insensitive to changes in the cast solvents.Finally, when matching PBDB-T derivatives with complementary absorption narrow-bandgap SMAs, such as A-DA′D-A type SMAs, high short circuit current (J SC ) can be achieved. [59]hese characteristics contribute to the perfect match of PM6 and Y6.
Hitherto, A-DA′D-A type SMAs (SMAs), represented by Y6, have boosted the PCE of OSCs to over 19% in recent years which brings the research area to a new stage.The universally prominent device performance based on the A-DA′D-A structured acceptors has driven us to explore the underlying molecular design principles behind Y6 (Figure 2b), and provide guidance for the next-generation acceptors.
We will now elaborate on the history and development of the A-DA′D-A type SMAs in order to illustrate the step-by-step optimization.Over the past decade, commercially available benzotriazole (BTA, Figure 2a) and benzothiadiazole (BT, Figure 2a), as heteroaromatic units with strong electron-withdrawing ability derived from the two imine bonds in the triazole or the thiadiazole ring, were very popular in the organic semiconductor field.[62][63][64][65] Particularly, there is an additional site on at the sp(3) 3 hybridized nitrogen atom in the BTA unit, which is conducive to the adjustment of solubility, structural, and electronic properties.The BTA unit was first introduced into polymer donors by Zou et al. [66] to synthesize a series of D-A copolymers, [62,67] realizing reasonable modulation of bandgap and energy levels, gradual improvement of mobility and photovoltaic properties as well  A).c) J-V curves of the OSCs based on PM6:Y6 under illumination with AM1.5G, 100 mW cm −2 .d) EQE spectra of the corresponding OSCs.e) Absorption spectra of thin films of PM6 and Y6.Reproduced with permission. [19]Copyright 2019, Elsevier Inc.
as exhibiting good thermal and air stability.68][69][70] Meanwhile, advancements in the field have seen the integration of electron donor materials possessing excellent photoelectronic properties with the SMAs. [12,71,72][78] Subsequently, the efficiency of the A-D-A SMDs-based OSCs was gradually improved [79,80] by optimizing the D unit from the thiophene/oligomer units [81,82] to the benzodithiophene fusedring units. [83,84]92][93] Inspired by this approach, researchers explored the implementation of fused-ring D units derived from polymer donors, such as indacenodithiophene (IDT) [72] in the synthesis of numerous narrow-band-gap A-D-A SMAs. [12,48]However, the utilization of the BTA electron acceptor component to achieve high-performance OSCs remained largely unexplored.Considering the early research experience with BTA and its derivative monomers, as well as insights from previous literature, [94,95] the synthesis of a ladder-type DA'D fused dithieno [3,2-b]pyrrolobenzotriazole (BZTP, Figure 2a With BZTP as the central five-membered-fused-ring central backbone, and 2-(3-oxo-2,3-dihydroinden-1-ylidene) malononitrile (INIC, Figure 2a) [96] as the A units at both ends, the A-DA′D-A type NFA named BZIC (Figure 2b) was synthesized and reported by Zou et al. in 2017. [97]This unique combination induced a planar orthogonal configuration in the DA'D framework instead of a twisted structure, eliminating the steric hindrance caused by sp(3)-carbon side chains and facilitating efficient electron delocalization and charge transport. [98]Consequently, the optical absorption, energy levels, and intermolecular interactions could be well-tuned.Notably, BZIC exhibits a significant red-shifted near-infrared absorption and higher HOMO energy compared to other SMAs such as ITIC.These attributes were advantageous for increasing the photocurrent and decreasing the device energy loss of the OSCs.These improvements can be attributed to the strong electron-donating ability of pyrrole units and the multiple D/A interactions, which help to upshift the energy levels and to enhance the intramolecular charge transfer (ICT).
Given the remarkable characteristics of A-DA′D-A type SMAs, extensive chemical modifications have been applied to improve the properties of BZIC-type NFAs, particularly by optimizing the A′ unit.Regarding the success in modulating the absorption, electron affinity, mobility, and other merits in the benzothiadiazole (BT) -based polymer donors, [60,61,95] the BT unit was introduced into the DA'D backbone to replace the BTA unit of BZIC.This marked the emergence of BT-core-based A-DA′D-A type SMAs, including Y6.However, compared to the electron-deficient BTA central core, the BT-based A-DA′D-A type SMAs possess relatively poorer solubility due to the lack of aliphatic substituents on the BT unit.Therefore, side chain modifications on the pyrrole rings and the terminal thiophene units of the DA'D backbone were considered to tune the molecular solubility, crystallinity, and packing motifs.For example, Zou et al. introduced -alkyl substituted thiophene [3,2b] thiophene (C11-TT) units into A-DA′D-A type SMAs, which coincides with the idea of Tang. [99]This way, the NFA Y6 (Figure 2b 2a) was designed and synthesized.Y6 shows good solubility, a stable, planar unified molecular conformation, and extended absorption, covering the VIS-NIR wavelength range down to 900 nm (Figure 2e).When matching with absorption-complementary polymer donor PM6 (Figure 2e), the Y6-based devices exhibit an outstanding PCE of 15.7% with a J SC of 25.3 mA cm −2 , V OC of 0.83 V, and an FF of 74.8%, achieving the highest efficiency at that time.Even when the film thickness reached 300 nm, the efficiency still remained above 13% (Figure 2c,d). [19]In the following, numerous chemical modifications based on Y6 were applied, resulting in the development of a series of high-performance A-DA′D-A type SMAs such as Y11, [100] Y18, [101] N3, [102] BTP-eC9, [103] L8-BO, [104,21] etc. (Figure 3).These advancements propelled the PCE to soar above 19%.
Simultaneously, many researchers have devoted much effort to investigating the intrinsic mechanism of Y6-based devices (see the next section for a detailed discussion of these processes).For example, Zhang et al. [105] demonstrated an intra-moiety excited(i-EX) state formed in the neat Y6 domains, which enables the dissociation of excitons into free holes and electrons.Neher et al. [106] showed efficient photocurrent generation in PM6:Y6 under a small driving energy for the dissociation of CT states.Yi et al. [107] also mentioned a barrierless charge separation in the Y6-based OSCs.This was ascribed to an increase in the polarization energies of hole and electron during their separation, which allowed to overcome the Coulomb attraction of the interfacial CT state.These large polarization energies could be attributed to the fluorinated end groups and electron-deficient A′ core in Y6. [108] Almost all these results indicate that the Y6-based OSCs could realize efficient charge separation despite a small energy level offset between the donor and acceptor materials, accounting for the high photovoltaic performance, especially the low voltage loss.Combined with the single crystal structure of the A-DA′D-A molecules, [109] it can be concluded that all these features derive from the unique molecular structure of Y6.Therefore, we summarize the following characteristics of the A-DA′D-A type SMAs below.
i) The A-DA′D-A molecular configuration is an excellent combination of a ladder-type electron-deficient-core-based central fused-ring backbone (DA′D) and two electronwithdrawing end groups (A), keeping two sp(2)-hybridized nitrogen atoms in the pyrrole motif of the fused-ring in a C2 symmetric manner. [23]In the DA′D fused-ring backbone, the strong electron-donating ability of pyrrole units as well as the extendable -conjugate system and the multiple D/A interactions are the essential reason for the enhanced ICT effect and the strong and extended near-infrared absorption.Besides, this distinctive core can tune the energy level to match high-performance donor materials with a small energy offset, which helps reduce energy loss.ii) The alkyl side chains attached to the N-atoms of the pyrrole motif can influence the planarity of the DA'D framework and the stacking of adjacent molecules, as they adopt an orthogonal conformation to the main plane (Figure 4a). [102,98]The periphery of the DA'D framework should be sterically hindered by side chains to prevent over-aggregation, and the fused-ring backbone should ensure an efficient intramolecular interaction under the multiple D/A interactions.iii) Adding -alkyl on the terminal thiophene of the fused-ring backbone is beneficial to achieving proper solubility, a highly unified conformation, and thus reduced the density of states (DOS) width and disorder.The relatively narrow DOS both in the electron affinity and ionization energy (Figure 4b) could lead to trap-free ambipolar transport. [110]v) Due to the exclusive curve shape type molecular geometry of the A-DA′D-A SMAs, dimerization will occur, forming a unit with a large quadrupole moment [106] (Figure 4e).Thereby, a unique - stacking mode forms between end groups (A and A) as well as the DA′D unit and the end groups (DA'D and A), resulting in a twisted transport channel and effective 3D ambipolar transport network [109] (Figure 4d,e).At the same time, such a high external quantum efficiency (EQE) above 80% between 400 and 900 nm (Figure 2d) in the A-DA'D-A-type SMAs, coupled with high electroluminescence quantum efficiency, makes it possible to deliver high J SC , high FF, and low voltage losses.[23] Thus far, significant efforts have been dedicated to maximizing the PCE of OSCs through strategies such as molecular structure modification and device optimization.The highest PCE of singlejunction OSCs based on A-DA′D-A-type SMAs is close to 20%, validating the success of the A-DA′D-A strategy.However, high cost is still one of the main restrictions for the further commercial application in OCSs.For example, the total synthesis route of PM6 consists of 17 steps, as for Y6, it has 15 steps.These individual steps often entail harsh reaction conditions, laborious postprocessing, and yield limitations.Moreover, the multi-fused-ring structure, which necessitates multiple ring-closing and coupling steps, inevitably escalates both the cost and complexity of synthesis. Thereore, it is imperative to focus on developing new materials with non-fused-ring structure or new synthesis methods with high yield, simple processing, and environmentally sound trait to reduce the cost while maintaining excellent PCE.
[120] However despite these advancements, further improvements in device efficiency are still required.Therefore, the device physics, morphology control, and the detailed intrinsic mechanism behind the photoelectric conversion process will be explored in the rest of this Review.This exploration will provide valuable insights for the development of next-generation donor and acceptor materials for OSCs.

From Photoexcited Excitons to Free Charges
This section is devoted to the details of free charge generation from photoexcited excitons in PM6:Y6 BHJ devices.We start by presenting initial evidence for activationless free-charge formation in this system.We then discuss in greater detail the nature and energetics of involved species and the processes of charge generation and dissociation before turning to the actual discussion on the reasons for the efficient free charge generation in this blend.

Local Excitons, CT States, and Free Charges
It is generally believed that free charge formation in organic BHJ devices is a multistep process, involving several excited states  [19] Copyright 2019, Elsevier Inc. b) Calculated DOS for electrons (EA) and holes (IE) in a model crystal of Y6.Reproduced with permission. [106]Copyright 2020, Wiley-VCH.c) Isosurfaces of the electrostatic potential of Y6, together with the ellipsoid of the quadrupolar tensor.Reproduced with permission. [106]Copyright 2020, Wiley-VCH.Molecular pairs in the Y6 single crystal.d) Top and e) side views of the extended-crystal structure (the blue column is the stack of end groups in the b direction, the pink column is the stack of end groups in the c direction.Reproduced with permission. [109]Copyright 2020, Springer Nature. (see Figure 5 for the schematic representation of these states and the transition rates). [121,122]First, an absorbed photon creates a local exciton (LE), in general a singlet state, on either the donor or acceptor domain.This eventually dissociates at the donor-Figure 5. State diagram describing the generation and recombination of free charges in a DA organic blend (see the text for the definition of the rates): S 1 , CT, and CS stand for the local singlet exciton on a neat component, for the charge transfer state and the charge-separated state, respectively.Each of these states exhibit energetic disorder as indicated by the Gaussian envelopes. [129]Copyright 2023, Wiley-VCH.
acceptor interface to form a charge transfer state (CT state).To generate a current in an outside circuit, this CT state must split into a pair of free (independent) charges to be collected at the electrodes through drift and diffusion.In this picture, it is common to consider two parameters: one is the energy offset between the frontier orbitals of the bulk heterojunction components, either the LUMOs or the HOMOs, ΔE LUMO or ΔE HOMO .This is considered as the energy gain when one of the photogenerated carriers crosses the DA heterojunction.This property, however, ignores the binding energy of the LE and of the CT state.A more appropriate property is, therefore, the difference between the energy of the local singlet exciton, E LE , and that of the charge separated state, E CS : ΔE CG = E LE − E CS .This is the total gain in energy when the exciton dissociates into free charges.For PM6:Y6, the LE with the lowest excitation energy is that of Y6, and its energy can be determined via the crossing point of the absorption and photoluminescence (PL) spectra of the blend, yielding E LE ≈ 1.41 eV. [123,124]The situation is more complicated when considering the energy of the charge-separated state, represented as the fundamental bandgap of the blend, E CS = E Y6 LUMO − E PM6 HOMO , simply because different ways of measuring the LUMO energy of the acceptor and the HOMO energy of the donor have yielded rather different numbers.We will come back to this point later.In the original work on PM6:Y6, cyclic voltammetry (CV) of films of the neat materials was used to determine the energy level offsets at the heterojunction interface, yielding ΔE LUMO = 0.6 eV .Bias has no effect on the charge generation efficiency.b) Internal quantum efficiency of photocurrent generation (IQE) and internal efficiency of free charge generation (IGE) as a function of photon energy, overlaid with the EQE spectrum.IQE and IGE are independent of photon energy even when exciting below the photovoltaic bandgap.c) Photogenerated charge as a function of bias and temperature for two photon energies.d) EQE spectra measured down to cryogenic temperatures.Except the very low-temperature range where transport issues become important, temperature has little effect on the EQE. [106]Copyright 2020, Wiley-VCH.e) Temperature dependence of the logarithm of the normalized internal quantum efficiency of PM6:Y6 (red diamond) compared to the blend of PM6 with the Y-series NFA BPT-eC9 (blue squares) and with the NFA ITIC (green circle).Also shown are the corresponding data for the blend of the non-fluorinated version of PM6, PBDB-T, with the NFA EH:IDTBR (yellow triangles).Solid lines show fits to a kinetic model which considers the competition between the splitting and the decay of the interfacial CT state. [130]Reproduced from Ref. [130] with permission from the Royal Society of Chemistry.
and ΔE HOMO = 0.09 eV. [19]This translates into a small if negligible driving force for dissociating the LE on Y6 into a CT state and eventually into a free electron-hole pair.Therefore, one would expect that the exciton dissociation has to be assisted by thermal energy, a strong electric field, or excess photon energy.We note at this point that all states involved in photoinduced charge generation experience inhomogeneous (and homogenous) broadening due to variations in the molecular con-formation and intermolecular interactions, [125][126][127][128] as indicated by the Gaussian envelopes in Figure 5.This needs to be kept in mind when assigning single numbers to E LE , E CT and E CS .
To study the role of temperature, electric field, and photon energy, Perdigón Toro et al. measured the efficiency of free charge generation using the time-delayed collection field (TDCF) method, with the main results displayed in Figure 6. [106]In a TDCF measurement, the device is initially held in the dark at a certain prebias V pre .Then, the sample is excited with a short laser pulse of low fluence and shortly after, the voltage is changed to a large reverse collection bias.This way, non-geminate recombination is largely suppressed.Measurements at room temperature revealed no effect of the prebias on the external efficiency of charge generation (EGE), pointing to only a low, if any, barrier to split the exciton and the CT state into free charges.Importantly, this observation remained valid even when exciting the blend at 1.29 eV, which is well below the band edge of the blend.This rules out any significance of the excess photon energy in the charge formation and separation processes, consistent with most fullerenebased systems studied.Finally, cooling the blend down to 230 K had very little effect on the bias-dependence of the EGE.To study this further, EQE spectra were recorded over a wide temperature range.Analyzing the results with the Arrhenius-equation yielded a very small activation energy for free charge formation of 6 meV.It was concluded that free charge generation is barrierless, despite the reported small HOMO offset. [106]A very low influence of temperature on the charge generation efficiency was independently confirmed by Ardalan Armin and coworkers. [130]hese authors employed temperature-dependent ultra-sensitive EQE on PM6:Y6 to prove a near unity charge generation yield.Only very small effect of temperature on the free charge generation efficiency was also seen in blends of PM6 with another Y-series NFA, BPT-eC9 (see Figure 3).In contrast, blends with traditional NFAs such as ITIC or EH:IDTBR yielded a more pronounced temperature dependence of the charge generation yield.The data were analyzed with a kinetic model based on the Braun-Onsager theory, [131] with the best fits shown by solid lines.Interestingly, this analysis yielded a significant activation energy for the CT dissociation rate k d in PM6:Y6 of ≈100 meV, suggesting a non-negligible barrier for charge generation.We will come back to this important point in Section 3.2.

The Local Exciton (LE)
The absorption of light by the PM6:Y6 blend has been discussed in detail in a recent paper by Köhler et al. [132] To identify the different contributions to the blend absorption spectrum, they first investigated the neat components in a MeTHF solution as a function of temperature, down to 140 K.For Y6, the high-temperature absorption in solution is characterized by a typical vibronic structure with the 0-0 transition at 1.75 eV which they assigned to non-aggregated molecules.Upon cooling, a new strong absorption appeared at 1.54 eV.This was attributed to the transition from the ground state to the first excited state in Y6 aggregates.Decreasing the temperature further revealed a second aggregate with a peak at 1.65 eV.The observed strong red-shift of the absorption onset when going from a single Y6 molecule to the aggregated state is in line with recent theoretical predictions by Andrienko and coworkers. [133]These assignments were then used to explain the absorption of the blend (Figure 7a).It turned out that the blend spectrum consists of contributions from all three species described above, though slightly shifted in energy.Notably, the spectral decomposition needed to take into account two additional higher energy features, centered at 1.85 and 2.0 eV, which were due to weak transitions to the upper levels of the Davydov-split excited state of the two aggregates.The important message was that the absorption onset of the Y6 is due to aggregates, but also that another aggregate as well as non-aggregated molecules contribute to the absorption.Also, the PM6 absorption in the blend was mainly due to aggregates, with a smaller contribution by non-aggregated chains centered at 2.1 eV.This is due to the strong tendency of PM6 to readily aggregate in solution. [18]owever, as discussed below, there is rapid energy and charge transfer from PM6 to Y6 in the blend; therefore, the nature and energy of the LE on PM6 is of less importance for the device properties and will not be discussed further.
These experimental findings are supported by structural studies and simulation work.For example, Bredas and coworkers showed that Y6 can adopt a large number of coplanar dimers (Figure 7b). [109,134]This was related to the specific molecular geometry of the molecule which enables different configurations (core/core, terminal-terminal but also core-terminal).In contrast, other NFAs such as ITIC-type acceptors mostly form TT-dimers.As a consequence, even amorphous Y6 exhibits a large number of interconnected transport pathways, in contrast to ITIC.More importantly, the special arrangement of the curved Y6 molecules causes significant intermolecular excitonic coupling, which is of the order of 50 meV. [109]From that, an exciton hopping rate of less than 1 ps was concluded based on Marcus theory.
There is conclusive experimental and theoretical evidence that the lowest energy excitations in Y6 aggregates exhibit significant CT character.For example, the electroabsorption (EA) spectrum of neat Y6 films followed the second derivative of the linear absorption spectrum, which is characteristic of a change of the dipole moment upon excitation. [126,133]In agreement to this, abinitio calculations of the aggregate absorption only reproduced the measured absorption edge when taking into account excited states with CT character. [133]A recent publication provided a detailed view on the excited properties of Y6 and other Y-series NFAs in dissolved state and in an aggregated film, in comparison to NFAs from the ITIC family. [109]For example, Figure 7c displays the transition orbitals of the hole (upper scheme) and the electron (lower scheme) of the CT-CT Y6-dimer, which is the dimer with the lowest excitation energy.There is a significant electronhole separation due to the stacking of the terminal acceptor unit with the core donor moiety.As a consequence, the electron-hole overlap is only 0.4, compared to 0.55 for the lowest energy singlet state of the non-aggregated molecule.This in turn was predicted to reduce the radiative decay rate, as proven experimentally.At the same time, these excited dimers exhibited a rather small nonadiabatic coupling (NAC).It was concluded that the decrease in NAC compensates for the effect of the lowering of the excitation energy upon aggregation, which would otherwise speed up nonradiative recombination according to the energy gap law. [135]As a result, a fairly small non-radiative decay rate k nr of 0.6 × 10 −9 s −1 was reported for all Y-series NFA studied in the solid state.As a consequence, neat films of Y6 (and other Y-series NFAs) exhibit fairly high photoluminescence quantum yield (PLQY), ranging typically between 0.6% and 7%, despite their rather small optical bandgap. [109,123,136]Indeed, organic light-emitting diodes based on a Y6-derivative showed record NIR EL efficiencies. [137]Note that PLQY measurements on films can be challenging because of waveguiding in the film and in the substrate, and the exact value will depend on the outcoupling and reabsorption efficiencies. [138]igure 7. Excitations in the PM6:Y6 blend: a) Spectral decomposition of the absorption spectra of the PM6:Y6 blend films into a disordered phase of the Y6 (yellow) and two acceptor aggregated phases (red, blue) with their low energy (solid line) and high energy (dashed line) absorption components.Reproduced with permission. [132]Copyright 2022 Wiley-VCH.The absorption contributed by PM6 is drawn with a turquoise line.b) Schematic dimer configurations with interactions between the molecular core (C) and terminal (T) in different orientations on the basis of MD simulations.Reproduced with permission. [134]Copyright 2021, Elsevier.c) Transition orbitals of the hole (upper scheme) and the electron (lower scheme) in the CT-CT Y6-dimer, which also has the lowest excited state energy, as derived from quantum-chemical calculations. [109]Due to specific face-to-face packing of the acceptortype end-group with the donor-type core moiety in this dimer, the excited state exhibits a significant CT character with reduced electron-hole overlap.d) Measured photoluminescence lifetimes of different Y6 and ITIC derivatives in neat acceptor films.Despite a smaller energy of the lowest excited state, most of the Y-series NFAs exhibit a longer PL lifetime e) The measured exciton diffusion length in neat NFA layers versus the reorganization energy from DFT calculations.The solid line shows the prediction of Marcus theory for a fixed intermolecular coupling.Reproduced with permission. [100]Copyright 2020, Springer Nature.Among the studied NFAs, Y6 stands out by a small reorganization energy.f) Natural transition orbitals of the interfacial CT states by using the TD-B97XD/6-31G(d,p) method coupled with the PCM model for molecular clusters of one PBDB-T-2F donor fragment with one (upper) or three (lower) Y6 molecules.Reproduced with permission. [109]Copyright 2020, Springer Nature.Due to the delocalization of the electron wavefunction, the estimated distance between the hole and electron in the CT state increases to 51 Å for clusters of three Y6 molecules.
Related to the small k nr is an exceptional long PL lifetime of Y6 and other Y-series NFAs in the solid state, significantly longer than of ITIC-based NFAs, see, e.g., Figure 7d.][141] Whether this is due to different morphologies or impurities is the subject of current studies.A consequence of the large Y6 exciton lifetime and the strong excitonic coupling is a long exciton diffusion length L D .Firdaus et al. applied two independent methods, namely steady state exciton quenching at a hole-transporting layer and transient exciton-exciton annihilation in neat films of Y6 and other acceptors. [100]For Y6, this yielded L D = 37 ± 1.1 nm.Exciton diffusion in this system was shown to proceed mainly via resonant Förster transfer due to the significant spectral overlap of the Y6 exciton absorption and emission.This is due to a small reorganization energy of 250 meV (Figure 7e). [100]Also, temperature dependent measurements on PM6:Y6 [142] revealed a small effect of temperature on the exciton diffusion coefficient, pointing to low energetic disorder due to the high structural order of these molecular layers.Similar conclusions were reached from earlier studies on the fused ring electron acceptor IDIC. [143]

The Charge Transfer State (CT state)
The nature and, especially, the energy of the CT state in the PM6:Y6 blend are intensively debated.In the community, an often-used approach to determine E CT is to fit the low energy part of the EQE PV spectrum and the high energy part of the EQE EL spectrum to Gaussian functions. [144]However, this turns out to be difficult for the PM6:Y6 blend because of the strong absorption and emission of the Y6 LE.Alternatively, the low energy shoulder in the blend EL has been assigned to the CT emission which, when fitted to a Gaussian function, [106,145] would yield E CT of (1.3 ± 0.1) eV, with an only 0.12 eV offset to the Y6 LE.A third approach is to compare the EQE PV and EQE EL of the blend and the neat films to identify the contribution of the CT state to the optical spectra.However, these spectra are prone to microcavity effects, which affect the shape and eventually the position of the optical features. [146,147]Changing the film composition from the neat components to the blend will unavoidably change the optical constants of the layer, and with this the properties of the microcavity.In addition, as pointed out above, the low energy Y6 absorption originates from aggregates, and it is likely that the optical properties of these aggregates change when going from the neat film to the blend.
Fortunately, it turned out that the PL of the blend in the actual device is entirely dominated by the emission from the Y6 LE. [123] This was proven by the fact that the position, strength, and spectral shape of the device PL was the same when measured at V OC or under short circuit conditions.If there is an appreciable contribution of the CT state to the PL, this contribution would be reduced once efficient CT dissociation and charge collection are realized at short circuit.This was not the case.Then, the application of the optical reciprocity to the PL spectrum gave the absorption spectrum of the Y6 LE in the actual device (including microcavity effects).Surprisingly, this absorption agreed exactly with the EQE PV spectrum of the device, except the tail at energies below 1.1 eV which was assigned to traps.It was concluded that the EQE PV is dominated by the Y6 exciton, with no or very little contribution by CT state absorption.Notably, the shoulder at 1.18 eV in the EQE PV spectrum, easily being misinterpreted as the CT state absorption, was consistently assigned to the 1-0 transition-the transition from the thermally excited first vibronic state of the electronic ground state to the vibronic ground state of the Y6 LE.Evidence for the existence of CT states in the blend then came from the comparison of the EL and PL spectra, both measured on the same spot on the actual device.Subtracting the PL from the EL revealed a broad emission centered at around 1.15 eV that was assigned to CT emission.One may be tempted to fit this emission with a Gaussian line shape to obtain the energy and energetic width of the CT state but microcavity effects are severe, especially in the region of low absorption.In fact, the exact position of this emission peak depended quite substantially on the device layout and active layer thickness, rendering it nearly impossible to determine the CT energy from the blend EL spectra.An alternative approach to determine E CT makes use of the fact that the CT state carries an electric dipole moment, which makes it accessible to electroabsorption spectroscopy.Wan et al. observed a sub-bandgap feature in the electroabsorption spectra of the PM6:Y6 blend, with a lineshape proportional to the second derivative of a Gaussian peak. [126]This indicates a strong charge transfer character of the excited state.From their analysis, E CT was determined to be 1.27 eV, ≈150 meV below Y6 LE.Interestingly, the extrapolation of the V OC toward 0 K, which is a common approach to determine the CT energy, suggested a low value of about 1.12 eV, which is again smaller by 150 me than the value from EA spectroscopy.It was concluded that the CT state exhibits significant energetic disorder and that as a consequence, only states in the tail of the CT state manifold probed by EA are occupied in the working device.
Recent DFT calculations dealt with the nature of the CT state.These calculations predicted that the delocalization of the electron over several Y6 molecules reduced the Coulomb attraction of the interfacial electron-hole pair from 160 to 70 meV.Figure 7f compares the orbitals of the interfacial CT states for a DA pair formed by one PM6 donor fragment with one or three Y6 molecules. [109]Delocalization increases the distance between the hole and electron in the interfacial CT state from 22 to 51 Å.In combination with other processes as discussed above, such delocalization effects are a likely cause of the observed activationless free charge formation.

The Charge Separated State (CS State)
In the CS, the electron and hole are independent.The nature and energy of the CS in PM6:Y6 is, therefore, determined by the properties of individual holes on PM6 and electrons on Y6.Temperature-dependent transport studies, as outlined below, showed that the carrier mobility of both the electron and hole increases with temperature and electric field, meaning that charge moves via hopping of localized polarons rather than by band transport.
The correct determination of the energies of the individual carriers on PM6 and Y6, and from these data of the energy offset at the heterojunction and the CS energy, is among the hottest debated topics of research on the PM6:Y6 blend.In the past, these energies were mostly derived from CV on neat layers.While there is quite a large scatter of the obtained values (see, e.g., table s1 in the supporting information of ref. [148], most studies reported a Y6 fundamental bandgap E Y6 G of 1.7 eV and a HOMO offset ΔE HOMO of 0.3 − 0.1 eV, yielding E CS ≈ 1.4 − 1.6 eV.There is a fundamental problem with these numbers.First, with E Y6 G = 1.7 eV and E Y6 LE = 1.42 eV, the exciton binding energy is ≈0.3 eV, larger than the average HOMO offset from CV.Thus, exciton dissociation at the heterojunction would require excess energy.Related to this, the average energy of the charge-separated state would be larger than E LE , rendering free charge formation from photoexcited excitons an endothermic process. Results from photoelectron spectroscopy (ultraviolet photoelectron spectroscopy, UPS, and inverse photoelectron spectroscopy, IPES) gave a different picture. [142,149]While values for E Y6 G were quite similar to those from CV, the comparison with the PM6 energy levels yielded a significantly larger Δ E HOMO = 0.5 − 0.7 eV and a smaller E CS ≈ 0.8 − 1.0 eV.This was mostly due to a smaller PM6 ionization energy of 5.1 eV compared to 5.65 V from CV.In a recent study, Baran and coworkers measured energy gaps and energy offsets for a wide range of material combinations, applying both CV, UPS, and IPES on neat layers. [150]The central results from this study are shown in Figure 8a.For almost all donor polymers, photoelectron spectroscopy gave a 0.2-0.6 eV smaller ionization energy than CV, while there was no such strong and systematic difference in the NFA energy levels.By comparing these data with the extrapolations of V OC towards zero temperature (a common procedure to estimate E CT ), the authors concluded that PES measurements provide more reliable results.Here, one needs to keep in mind that PES is very surface-sensitive.Therefore, the energy levels from PES are not necessarily representative of the bulk properties.Also, the results of PES are affected by electrical dipoles and quadrupole moments of the molecules, depending, e.g., on their orientation to the surface. [151,152]Related to this, the ionization energy of a specific molecule measured by PES will depend on its local environment, [153] which is different for a neat layer or a blend.Following this line of arguments, energy offsets as determined from the measurements on neat layers may not be representative of the blend.For example, recent UPS measurements gave the Y6 HOMO at −5.38 eV in the blend, with a 0.3 eV shift compared to the HOMO position in a neat Y6 film which was at −5.68 eV). [148]With a PM6 HOMO of −5.13 eV in the blend, ΔE HOMO was estimated to 0.26 eV, ca.0.3 eV smaller than the value determined from the UPS spectra of the neat components.To confirm this value independently, spectroelectrochemistry (SEC) was applied to the blend layer.In contrast to classical CV, SEC uses characteristic changes of the optical absorption spectrum to determine the onsets of oxidation and reduction.This measurement revealed a HOMO offset of 0.33 eV, quite similar to the value from UPS on the blend (see Figure 8b).In turn, the E CS was measured as ≈1.4 eV, which is quite close to the energy of the Y6 LE.In this simple picture (which neglects energetic disorder and other phenomena), free charge formation would not require excess energy, but also, there is no net driving force.Interestingly, the SEC study revealed rather similar values for the HOMO and LUMO energies in the neat layer and in the blend.SEC is different from UPS as its probes primarily the bulk properties.This study also investigated the influence of the poly-mer chain orientation on the energetics.Neat films and blends from o-xylene exhibit a preferential edge-on orientation of the PM6 chains in the bulk, which is in contrast to CF/CN-cast layers with a strong face-on orientation of PM6.This had, however, only a minor effect on the absolute orbital energies and with that on the fundamental bandgap.A more recent study focused on the effect of the orientation and packing of Y6.Here, ambient photoelectron spectroscopy gave a 0.26 eV high ionization energy for a neat Y6 film-coated from CB compared to CF (Figure 8c).It has been shown that CB-coated films exhibit stronger aggregation but also a more random molecular orientation, see Figure 15b and the related discussion.It was concluded that the more ordered eclipsed stacking of the Y6 molecules but also the more pronounced edge-on orientation in the CB-coated layer allows for a better superposition of the molecular quadrupole fields, causing the observed down-shift of the energy levels.As such, the bulk-HOMO offset is predicted to depend on the order and orientation of Y6, adding complexity to the interpretation of device data.These authors also measured the ionization onset of a 5 nm PM6 film on a neat Y6 layer.The result (≈−5.1 eV) was quite similar as for a neat PM6 layer on ITO but there was also a light difference of the HOMO onset for PM6 on top of Y6 CF and Y6 CB.This was attributed to slight band bending induced in the ultrathin PM6 layer due to the electrostatic potential created by the molecular quadrupole moments in Y6.In conclusion, this study revealed a rather large HOMO offset of the charge-separated state between 0.6 and 0.8 eV, depending on the Y6 packing and orientation.With the well-established HOMO-LUMO gap of 1.7 eV for Y6, E CS is only 1.0 ± 0.1 eV, which is rather small given the high V OC of the blend.Besides these sample-and method-related topics, a further difficulty arises from the fact that organic semiconductor films exhibit inhomogeneous energetic disorder.As such, there is not HOMO and LUMO energy.Instead, the energetics is described by a density of states (DOS) distribution g(E).This is shown in Figure 8e which displays the DOS of a neat PM6 film, a neat Y6 film and a PM6:Y6 blend, as inferred from energy-resolved electrochemical impedance spectroscopy (ER-EIS) measurements. [154]For all samples, the measurements revealed a significant tailing of the states towards the gap.The DOS of the frontier orbitals was fitted by Gaussian distributions, where the center of the distribution was taken as the HOMO respectively LUMO energy, see Figure 8f.This study showed only a small difference between the energies of the frontier orbitals in the neat layer and in the blend, in agreement to the results from SEC shown in Figure 8b.Finally, the analysis of the data in Figure 8f yielded a fundamental gap energy E CS = 1.44 eV and a HOMO offset ΔE HOMO ≅0.2 eV, rather similar to the energy scheme from SEC but different from the results from PES. ER-EIS and SEC have in common that the measurements are performed with the samples immersed into a liquid electrolyte.The presence of mobile ions at the surface and eventually in the organic layer may affect the electrostatics, e.g. by screening the dipole-and quadrupole-induced fields.Finally, when comparing numbers from these and other measurements, it's important to state whether energy values were taken from energy onsets (as for the data in Figure 8a,b,d) or from fits of the data to DOS distributions (as in Figure 8f).Copyright 2022, Wiley-VCH.All measurements were performed on neat films.While the results from PES and CV agreed rather well for the NFAs, UPS consistently reveal a smaller ionization energy of the polymer layers, compared to CV.As a consequence, PES predicts a smaller energy of the charged separated state.b) HOMO and LUMO energies from spectroelectrochemistry of neat films and the PM6:Y6 blend, coated either from chloroform-chloronaphthalene or chlorobenzene.Reproduced with permission. [148]Copyright 2022, Royal Society of Chemistry.The results show little difference between the energy levels in neat and blend films but also a small effect of the used solvent and with that of the molecular orientation

Mechanisms of Free Charge Generation
It is generally believed that free charge generation from excitons involves firstly the formation of a CT state followed by its dissociation in free charges.We will now discuss free charge generation in the framework of this picture before turning to alternative models.

Exciton Dissociation
Pump-probe transient absorption spectroscopy is a powerful tool to study exciton, CT, and free charge generation dynamics in NFA blends at ultrafast time scales.In case of PM6:Y6, several works have been dedicated to study the hole transfer process via selective excitation of Y6.This holds relevance due to the strong energy funneling from PM6 to Y6 upon direct PM6 excitation [142] so also ultrafast electron transfer from photoexcited PM6 to Y6 has been suggested. [156]Similar to other NFA blend systems, [157] the charge transfer in PM6:Y6 has been demonstrated to occur in sub-picosecond timescales. [105,156]Upon selective excitation of Y6, the excited-state absorption (ESA) band centered at 920 nm is observed instantaneously due to the Y6 excitons.The ESA feature is found to evolve rapidly within <5 ps into a polaron absorption band centered 980 nm due overlapping hole [156] and electron [136] absorption bands of PM6 and Y6, respectively.Even with subpicosecond charge transfer, the polaron absorption band undergoes an exponential rise ( ≈ 15 ps), taking up to ≈100 ps to reach complete evolution, [105,149,156] due to the longer times required for excitons to diffuse to the interface, consistent with long-range exciton diffusion in Y6 as discussed in the previous section on the local exciton.Notably, the growth of the PM6 ground-state bleach (GSB) upon direct Y6 excitation, being a proper measure of the Y6 exciton dissociation dynamics, was barely affected by temperature. [142]This agrees with the finding of a very small temperature dependence of exciton diffusion, as discussed earlier.

Mechanisms of Charge Separation
As outlined earlier, both TDCF and EQE measurements showed that free charge formation is independent of electric field and photon excitation energy and depends very little (if at all) on temperature.It was concluded that no barrier exists for CT dissociation, contrasting the view of a significant Coulomb barrier as expected from simple electrostatics considerations.Thus, another force must counteract the Coulomb attraction.Efficient charge separation, independent of electric field and photon energy, has also been observed for some polymer:fullerene systems, and different models have been proposed to explain this surprising phenomenon. [158]very popular model is that of a morphology-derived driving force, pioneered by Durrant et al. [159] This model can be applied to all blends that exhibit ordered (crystalline) domains of the neat compounds.It is assumed that the heterojunction interface area is less ordered.For almost all organic materials, the electron affinity (ionization energy) depends on the molecular conformation and it is generally larger (smaller) in the aggregated than in the non-aggregated phase.For example, Jamieson and Shoaee reported a ≈200 meV larger electron affinity for PCBM molecules in a neat film compared to a blend in polystyrene. [159,160]For PM6:Y6, this topic was recently addressed by TAS by the group of Ohkita. [124]To this end, the authors performed a careful decomposition of the PM6:Y6 TAS spectrum, in part based on the comparison with different blend compositions and other material combinations.The Y6 GSB exhibited a significant red shift from 820 to 850 nm after photoexcitation at 800 nm.At the same time, a photoinduced absorption (PIA) feature at 780-800 nm (assigned to Y6 anions on well-ordered Y6 domains) and at 680 nm (due to electroabsorption of the PM6 GSB) appeared.These characteristic spectral changes developed at the time scale of 10-100 ps.Importantly, the growth rate of these two features depended little on temperature.The authors concluded that free charge formation involves a down-hill process, where electrons move from Y6 molecules in the more disordered DA interfacial area to aggregates in neat Y6 domains Figure 9a.To estimate the driving force for this process, the HOMO energy of Y6 in solution (measured by CV) was compared to the HOMO energy of solid Y6 (determined by photoelectron yield spectroscopy, PYS).From these numbers, the LUMO energies were estimated by adding the optical bandgaps, yielding −3.74 eV and −4.35 eV for the electron affinity of Y6 in solution and in the neat film, which implies a very large morphological driving force of 600 meV.However as noted earlier, the direct comparison of data from CV and photoelectron spectroscopy must be considered with great care.
A particular property of Y6 is that it carries quite a large quadrupole moment.This may actually be the reason why UPS measurements on neat Y6 and the blend yield different HOMO energies as mentioned earlier.Regarding CT dissociation, Andrienko and coworkers predicted that for a corrugated (or intermixed) donor-acceptor heterojunction, such molecular quadrupole moments can cause a band bending, which pulls charges away from the interface towards the neat domains. [161]n exemplary energy scheme based on the data in Figure 8c,d is shown in Figure 9b.It was later confirmed that for many NFAs, this pull-out force is large enough to compensate the Coulomb attraction, rendering CT dissociation barrierless. [106,142]he role of the quadrupole moment in charge separation was then confirmed by comparing CT dissociation in blends of the donor polymer PTB7-Th with two NFAs, ITIC and h-ITIC. [162]ransient absorption spectroscopy (TAS) and TDCF revealed a significantly reduced CT recombination and a smaller field of the PM6 chains on the frontier orbital energies in the bulk of the neat and blend layer.c) The ambient photoelectron spectra (APS) of Y6 in neat films coated from CF and CB.A significantly higher ionization energy of the CB-coated film is explained by a larger impact of the molecular quadrupole moment on the bulk energy, due to better packing.Reproduced with permission. [155]Copyright 2023, Springer Nature.d) APS spectra of 5 nm PM6 layer on top of Y6 CF and Y6 CB compared to a neat PM6 layer on ITO/ZnO.Shown in the inset is the normalized density of states.e) The density of states (DOS) distribution for the HOMO and LUMO of a neat PM6, a neat Y6 and PM6:Y6 blend film, as deduced from energy-resolved electrochemical impedance spectroscopy (ER-EIS).Reproduced with permission. [154]Copyright 2023, Wiley-VCH.f) The complete DOS distribution for a PM6:Y6 blend film, together with Gaussian fits to the frontier orbitals.The fundamental energy gap was deduced from the centers of the HOMO and LUMO DOS, yielding a value of 1.44 eV.The vertical lines indicate the centers of the HOMO and LUMO DOSs of neat PM6 (yellow) and Y6 (red).Proposed mechanisms to explain activationless free charge generation in the PM6:Y6 blend.(a) A cascaded energy landscape is created by the lower-lying LUMO of Y6 molecules in ordered domains compared to the more disordered interface.Reproduced with permission. [124]Copyright 2022, The Royal Society of Chemistry.This drives electrons into the bulk of the Y6 domains.b) The quadrupole moment of Y6 molecules and dimers increases the ionization energy and electron affinity in the Y6 bulk relative to the interface.This creates a band bending towards the donor that increases the energy of the CT state relative to the CS, which counteracts the Coulomb attraction, and also suppresses recombination.Reproduced with permission. [155]opyright 2023, Nature Publishing Group.Because the polymer carries a much smaller quadrupole moment, there is only little band bending in the donor phase.c) Due to the larger energetic disorder for free charges compared to Y6 excitons, charges can equilibrate at energies well below the mean energy of the photogenerated excitons, providing a driving force for exciton dissociation into free charges (graph derived from data in [141] ).d) Free charges are generated by efficient exciton dissociation in neat Y6 domains, while the role of the donor is mainly to collect the photogenerated holes and reduce non-geminate recombination.Reproduced with permission. [136]Copyright 2022 Springer Nature.
dependence of free charge generation for the blend with ITIC which carries a significantly larger quadrupole moment as h-ITIC.
The view of barrierless free charge generation, however, contradicts the results from TAS measurements by Chow and coworkers. [142,163]Here, photoinduced absorption (PIA) in the 700-790 nm range was used to measure the charge separation dynamics.At room temperature, the CT state dissociation rate k d was ≈ 5.5 × 10 10 s −1 but decreased considerably when cooling the sample down to cryogenic temperatures.At the same time, the saturated PIA signal decreased, the PL decay slowed down and the PL intensity increased.It was concluded that CT dissociation is thermally assisted because charges have to overcome a significant Coulomb potential in order to separate.In this picture, CT state dissociation becomes less efficient at a lower temperature, thereby increasing the likelihood of LE reformation and emission.These findings contrast the results of TAS measurements by Ohkita and coworkers as discussed earlier, though both groups used similar excitation conditions (800 nm, ca.1.5 − 3.5 μJ/cm 2 ) and analyzed the signal dynamics in the same spectral region at 780 nm.Variation in blend morphology and pos-sible unknown details of the measurement conditions could account for this inconsistency.
In any case, a temperature dependence of the CT state dissociation rate does not necessary imply an energy barrier for charge separation.In the absence of disorder, the zero field CT dissociation rate can be approximated by: [164] where is the Langevin recombination rate.Here, B is a prefactor depending on the lattice and interface morphology, a is the CT state diameter, q is the elementary charge,  0 and  r is the vacuum and relative dielectric constant, and μ e and μ h is the electron and hole mobility, respectively.Because charge transport proceeds through thermally-assisted hopping, k d will be temperature dependent even in absence of a dissociation barrier.On the other hand, a temperature-dependent k d does not necessarily imply that the same is true for the efficiency of CT dissociation: Here, k f is the rate of CT recombination which is generally assumed to have little dependence on E and T. If k d (E,T) > >k f , (E, T) is nearly independent of field and temperature even if a barrier existed.In fact, the analysis of the temperature-dependent data in Figure 6e with Equations ( 1) and (3) yielded a quite considerable energy barrier for CT dissociation of 103 meV for the PM6:Y6 blend, despite the very small effect of temperature on the free charge generation efficiency.This was explained by fast though temperature-dependent dissociation, which competes efficiently with CT recombination. [130]Clearly, further work is needed to clarify the details of the CT state dissociation process in this blend.

The Role of Energetic Disorder
Despite the high crystallinity of Y6, the PM6:Y6 blend exhibits significant energetic disorder.Temperature-dependent spacecharge limited current measurements on PM6:Y6 blends were consistent with Gaussian-type disorder of width , where  is typically 55-60 meV and 70-80 meV for the LUMO of Y6 and HOMO of PM6, respectively. [141,165]Note that even lower disorder values have been reported for the blend of PM6 with other Y-series NFAs. [110]168] If carriers fully equilibrate during their lifetime, their mean energy will be  2 /k B T below the center of the DOS distribution. [169]his may provide a sufficient driving force for CT dissociation.The question whether carriers in PM6:Y6 make full use of this extra energy gain prior to recombination is, however, heavily debated, [170] see Section 4.5 for a detailed discussion of this topic.In a recent study, Perdigón Toro et al. measured the V OC and the charge carrier density as function of temperature for different light intensities. [141]The data could be consistently explained under assumption of full equilibration of the photogenerated carriers.Based on the measured disorder parameters, the authors concluded that the energy of the fully equilibrated CS is ca.0.3 eV below E Y6 LE , thereby stabilizing the free photogenerated charges Figure 9c).It was also shown that the steady-state density of photogenerated carriers increases with decreasing temperature, indicating that the CT-CS balance shifts towards free charges for lower T and again implies a down-hill process.It was concluded that energetic disorder is likely to contribute to free charge formation in the high-performance PM6:Y6 blend.Here, we point out that an approach frequently used in literature to determine the disorder is to analyze the slope of the tail of the EQE PV spectrum.However, for a Gaussian disorder, this slope will always be equal to the thermal energy, independent of the width of the DOS, . [171]Also, because the tail of the EQE PV spectrum is dominated by the Y6 LE, it does not provide any information on the disorder of the charge transporting states.In turn, the initial drop of the EQE PV at the band edge yielded information on the energetic disorder of the LE, which was determined to ca. 35 meV. [127]

Direct Free Charge Formation from Y6 Excitons
There is a current debate whether CT formation and separation are actually part of the free charge formation pathway.For example, Wei and coworkers employed the quantum mechanics/embedded charge method to calculate the energy of the Y6 exciton and of free charges on Y6 domains in the solid state.Their result was that the energy of the CS is actually 0.11 − 0.15 eV lower than that of the Y6 LE. [172] This theoretical prediction was confirmed by the increase of the steady-state PL intensity with increasing temperature, which was explained by an uphill process for the reformation of the LE from independent carriers.It was argued that the CS is strongly stabilized by polarization effects.On the other hand, their calculated solid-state E CS of (1.83 ± 0.1) eV is rather large and their E LE of (1.95 ± 0.1) eV is well above the experimental value.
Direct free charge generation in neat Y6 films was conclusively shown by a combination of PL and TAS spectroscopy and supported by quantum chemical calculations. [136]TAS yielded the growth of a polaron-related signal within 2.5 ps.To determine the efficiency of free charge generation, intensity-dependent steady state PL measurements were analyzed by a dynamic model (including reformation of Y6 excitons due to free charge recombination on neat Y6 films competing with trap-assisted recombination) and the Saha equation.This equation considers the dynamic equilibrium between excitons and free charges as schematically shown in Figure 9d.If this equilibrium exists, both have the same electrochemical potential, μ LE = μ CS resulting in. [123,173] Here, n LE and N LE is the number density and site density of local excitons and n CS and N CS are the corresponding parameters for the free charges.Then, If is expressed by the corresponding term for parabolic bands in a crystalline semiconductor, Equation 5 is also called the Saha equation.Notably, if the total excitation density goes to zero, n CS ≫ n LE .This is because of entropy.For an exciton, there are N LE possibilities to be placed but N 2 CS for a pair of independent charges.From the analysis of their data, the authors concluded an exciton binding energy of 270 meV.Note that this number was derived based on the original Saha equation, which may not necessarily hold for a crystalline Y6 domain.If one, instead, uses the molecular number density N Y6 for both N LE and N CS , the exciton binding energy deduced from their data becomes 360 meV.Within the uncertainties of the experiments and the models, these numbers agree well with the estimate of the binding energy outlined earlier.If excitons dissociate directly and efficiently into free charges, the main role of the donor is to take up photogenerated holes from the Y6 phase and to keep them separated from the electrons.Transient decay data showed rapid Langevintype nongeminate recombination of free charges on neat Y6 and that the recombination rate was largely reduced upon the addition of a small concentration of the donor polymer PTB7-Th. [136]A follow-up paper presented efficient quasihomojunction solar cells based on the PTB7-Th:Y6 blend, with only a small reduction of the PCE for a D:A ratio of 1:8 compared to the optimum concentration (1:1.2). [174] Clearly, spontaneous exciton dissociation in neat acceptor domains needs to be taken into account in future work on the understanding and optimization of NFA-based solar cells.Indeed, neat Y6 devices were successfully fabricated to achieve 4.5% PCE. [175]In Saglamkaya et al.'s research, the authors demonstrated efficient charge generation within the neat Y6 device.This process was suggested to involve both interface charge generation (small contribution only) and bulk generation facilitated by aggregation-induced energetic sinks.The study showed that whilst charge generation readily occurs in neat Y6 films, nonetheless recombination of the free carriers is very fast, and appropriate transport layers are needed in order to stabilize the separated charges and slow down the recombination.

Device Physics
Once a photogenerated charge carrier successfully separates from its geminate counter charge, the internal electric field in the device drives the free charges toward the electrodes.Holes drift to the anode while electrons drift to the cathode.The maximum photocurrent is achieved when all of these charges are collected at the electrodes.As a forward bias is applied, the driving force for charge extraction decreases and so too does the charge collection efficiency.On the other hand, as most OSCs comprise nanoscale phase-separated domains, holes and electrons frequently encounter at the interface during charge transport, resulting in non-geminate recombination (NGR).The competition between charge recombination and transport efficiency in a given system is reflected in the FF. [176]s an important photovoltaic parameter, the FF of OSCs stands for the effectiveness of charge generation and collection, which plays a critical role in power conversion efficiency. [177,178]nder idealized conditions, the current under illumination is expected to obey the Shockley diode equation, according to which. [179] Here, J G is the photogeneration current, J 0 is the associated dark saturation current, and n id is the diode ideality factor (n id = 1 when bimolecular recombination dominates, and 2 when trap-assisted recombination dominates).However, the FF of PM6:Y6 device is ≈75%, lacking well behind the prediction of the Shockley-Queisser theory for the given V OC value (0.87 V).The difference may arise from the fact that this model applies to conditions when surface recombination and trap-assisted recombination are negligible and only recombination of free carriers dominate.In reality, this is often not the case, perhaps providing an explanation for the smaller than expected FF.Another possibility for the lower FF may be unintentional doping of the active layer.As reported by Tokmoldin et al., unintentional doping, with doping concentrations close to 10 16 cm −3 was observed in thick junction PM6:Y6 devices. [180]Whilst such high doping levels can cause space charge effects, thereby reducing the FF, this does not explain the FF in thin junction devices, where such high doping concentrations are almost never reported.Rather, because of the low charge carrier mobilities, the competition between charge extraction and non-geminate recombination, results in FFs falling short of the ideal values.Using analytical assumptions, Neher et al. modified the Shockley equation (Equation 6a) to account for the competition between recombination of free carriers and charge extraction. [181]In this equation, the decisive parameter is the dimensionless figure of merit .
The fill factor, as per Equation 6a, is only restored for  < 1, resembling a Shockley-type cell.Conversely, in all other scenarios, the fill factor declines due to too slow motion motion, characterizing transport-limited cells (Figure 10a).Recently, Tokmoldin et al. expressed the charge collection losses in transport-limited solar cells in terms of drift and diffusion of the charges. [176]As shown in Figure 10b, the J SC is determined by the drift length l dr at short circuit conditions, as expected.On the other hand, the FF is determined by the diffusion length, l dif , relative to the active layer thickness, see Figure 10c.It was determined that under short circuit conditions, for almost all systems studied, including PM6:Y6 blends of different layer thickness (d between 100 and 550 nm), the drift length is adequately long to extract most carriers before their recombination occurs.This explains the high J SC .On the other hand, none of the systems had a sufficiently high diffusion length to enable charge extraction at small internal fields, causing a loss in FF compared to the ideal case.

Charge Recombination
The photoexcited charge-carrier lifetime is an important metric in a photovoltaic device, where longer carrier lifetimes have been shown to directly correlate with higher power conversion efficiency.The recombination of charge carriers in a semiconductor can be summarized through the following rate equation: where n is the electron charge carrier density (assuming electron carrier density equals hole carrier density), k sur is surface recombination (non-radiative), k SRH is the first-order Shockley-Read-Hall (SRH) trapping (non-radiative) rate constant via mid-gap states [182] or tail states, [183,184] k rec is the second-order recombination rate constant of free carriers (radiative and non-radiative).Non-radiative losses play a significant role in constraining the power-conversion efficiency of OSCs.The primary non-radiative loss within the bulk is inherent to the active layer.Nevertheless, beyond bulk non-radiative recombination, we must also address surface recombination at the contacts to enhance device performance, in particular the V OC .Surface recombination at the contacts, i.e., the extraction of the wrong carrier type (electrons at the anode, holes at the cathode) has been found to reduce FFs and increase non-radiative recombination losses of the open-circuit Merit  for various BHJ blends, as reported in ref. [181].Reproduced with permission. [181]Copyright 2016, Springer Nature.Solid lines are analytical predictions of the FF- relation for V OC increasing from 0.7 to 0.9 V. Open circles are FF- points from simulated J-V curves with balanced mobilities and V OC between 0.7 and 0.9 V. b) Correlation of the relative short circuit current J SC /J G versus the effective drift length and c) fill-factor versus the effective diffusion length at the 1 Sun-equivalent illumination.Reproduced with permission. [176]Copyright 2021 Wiley-VCH.Systems 1-7 are PM6:Y6 blends of different thicknesses and preparation conditions.
voltage. [185,186]Two studies by Le et al. [187] and Riley et al. [188] elegantly quantified bulk and interfacial recombination by using either photoinduced absorption spectroscopy or electromodulated photoluminescence.By obtaining the quasi-Fermi level splitting (QFLS) in a BHJ film on glass and in the complete device, and the radiative thermodynamic limit of the photovoltage,it was demonstrated that unintentional diffusion of carriers to the wrong contact-surface recombination-does not play a role in state of the art PM6:Y6 device.From the same work, it was also concluded that trap recombination is negligible in this system.Consistently, the work by Wu et al. also quantified trap and tail states to be exceptionally low. [149]In this picture the presence of tail states is a source of energetic loss, as charge carriers relax into these states, reducing the QFLS and therefore device V OC .
Nongeminate recombination of free carriers to the ground state is not a direct transition in organic donor-acceptor blends.As shown in Figure 5, this process is mediated by the charge transfer states.If the oppositely charged carriers are statistically independent of each other, then NGR is a random process and hence depends on charge carrier densities and the relative mobility μ (μ = μ e + μ h where μ e and μ h are electron and hole mobility, respectively).Thus, the recombination rate R can be described as follows: here, n and p represent charge carrier density for electrons and holes, respectively.The prefactor k rec is the recombination rate coefficient and is a function of the relative mobility μ.
We categorize NGR into two classes. [189]In one limit, the recombination of free charge carriers is described with the Langevin theory.When the mobility of localized charge carriers is relatively low, the mean free path of the carriers is less than the radius of capture of one carrier by the other.In this case, the recombination coefficient is proportional to the probability of opposite charges encountering one another, given by k L in Equation 2. Thus, the Langevin recombination rate is proportional to the mobility of the free carriers.This means that a higher mobility system would experience a faster NGR, resulting in a negligible or often negative impact on charge collection efficiency.In the alternative limit, the mean free path of the carrier is greater than the capture radius.
To understand the NGR, we relate the recombination constant k rec (the rate observed by an experimenter as an effective recombination rate constant of free charges to the ground state) to the Langevin recombination constant k L ; the encounter rate constant of the oppositely charged carriers in a homogeneous medium: k rec = k L .In an equilibrium picture, by detailed balance and assuming Langevin CT state formation, the dissociation probability must also depend on the relative concentrations of CT states and free charges, as well as on their mobility.When re-dissociation of the CT state is possible, compared to the Langevin theory, the recombination rate is reduced by : which is defined as a reduced recombination factor and in which , k BET is the rate constant of back electron transfer of triplet charge transfer ( 3 CT) states to form triplet excitons on either the donor or acceptor, k d is the dissociation rate constant of the CT states to free charges, and k f the CT decay rate constant to the ground state.For the non-Langevin systems, a clear correlation between k rec and ΔV nr can be seen.No such correlation is apparent for the Langevin systems, where the assumption of quasi-equilibrium between free carriers and CT does not hold. [129]Copyright 2023, Wiley-VCH.
In PM6:Y6, NGR coefficient values ranging from 10 −11 to 10 −13 cm 3 s −1 have been reported.The variation in k rec value stems from both morphology variations obtained from sample preparation in different laboratories (in Y6 giving rise to different crystallinity and thereby different device performance) and techniques used to determine the recombination coefficient.In particular, some techniques are sensitive to microcavity effects that arise due to thickness variation. [148]Whilst Hosseini et al. and Wu et al. measured a relatively high recombination coefficient value of 10 −11 cm 3 s −1 for PM6:Y6 (with an FF of 73%), Karki et al. measured a very low k rec (3 − 6) × 10 −13 cm 3 s −1 (FF 72%). [145,149,190]In 2021, Nyman et al. also studied the recombination of PM6:Y6 and found a recombination value of 10 −12 cm 3 s −1 (FF 73%). [191]When neglecting triplets, an important consequence of reduced Langevin recombination is that strongly suppressed recombination would be limited to systems with efficient and field-independent CT dissociation.In this regard, the measured recombination values by Karki et al. and Nyman et al. all  indicate efficient re-splitting of the interfacial CT state in this blend, in agreement with its efficient charge generation.Conversely, Zuo et al. used kinetic Monte Carlo simulations to understand the interplay between free charge motion and recombination in an energetically disordered phase-separated donoracceptor blend. [192]It was found that mobility is not the decisive parameter determining the NGR coefficient, rather CT reformation and resplitting involved mostly states near the transport energy.On that basis, the authors concluded that charge encounter is more affected by increased disorder than the resplitting of the CT state.On the other hand, a recent study by Gillett et al. demonstrated that 90% of recombination proceeds through the formation of triplet excitons. [156]As such, and as can be seen from Equation 9, the direct correlation between charge generation and recombination is altered and the faster recombination observed by Hosseini and Wu insinuates that indeed in addition to the decay of the CT state to the ground state (k f ), there is an additional recombination channel.
In the above picture for systems showing reduced recombination because of high CT state re-dissociation, an equilibrium between free carriers and CT states is established and the position of the equilibrium is given by the decay of the CT states, mainly through non-radiative pathways. [189]However, the underlying principle that dictates the non-radiative decay of the CT is not fully understood.Whilst the decay of the CT state has originally been considered using Marcus theory, the studies assume a single CT energy, whose spectral features are broadened only by the dynamic disorder. [193]However, there is a debate about the importance of static disorder as the weak cohesion between individual molecules through van der Waals interactions together with conformational irregularities also lead to a broadened distribution of CT states. [194]To answer this question, Hosseini et al. studied the device performance of various fullerene and non-fullerene systems and correlated the recombination of free carriers with the energetic disorder of free electrons and holes. [129]It was found that by reducing the (static) energetic disorder, the recombination of free charge carriers were suppressed (Figure 11a).Using the modified Marcus−Levich−Jortner model for describing the decay of the CT state to the ground state, [126] the authors postulated that the underlying mechanism is an interrelation between the decay rate of the interfacial CT state and the broadening of the DOS experienced by carriers in the CT manifold; a less disordered system results in a slower (non-radiative) decay of the CT state.Thus, suppression of free carrier recombination upon decreasing energetic disorder is a key parameter in reducing non-radiative recombination to gain in both V OC and FF (see Figure 11b for the relation between the non-radiative voltage loss and the recombination coefficient).This trend is consistent with the work of Wu et al. who observed exceptionally small characteristic energy E ch in PM6:Y6; where this exponential tail was associated with the energetic disorder.The authors correlate a smaller E ch with a lower degree of trap-assisted recombination and a higher mobility.It should be noted that whilst the characteristic energy estimate is on the order ok k B T at room temperature, and its validity with regard to energetic disorder has been questioned, however the trend in values reported correlates with recombination and device performance for the systems studied.a-c).In PM6:Y6 devices, the chemical potential of the Y6 singlet exciton, μ S1 , is equal to the quasi-Fermi-level splitting in the bulk; thus, singlet excitons are in dynamic equilibrium with free carriers in the CS state and with the CT state population .Most of the photon emission of the excited blend originates from the Y6 exciton.However, most non-geminate recombination occurs through a very weakly emitting state, different from the Y6 singlet.We can relate the electroluminescence quantum efficiency (ELQY) of the singlet excitons in the device to the PLQY of the PS:Y6 film and conclude that <0.6% of injected charges are reformed into excitons.The low yield of reformation can be explained by the barrier between the singlet energy and the effective transport gap (CS state).Adapted with permission. [123]Copyright 2021 American Chemical Society.(e) State diagram of an organic solar cell with the low energy offset, indicating various transitions between the ground state singlet S 0 , singlet exciton S 1 , charge-transfer (CT), and charge-separated (CS) states: photon absorption under illumination (h), carrier injection under external bias (j inj ), exciton decay (k f ,S 1 ), exciton dissociation to CT (k d,S 1 ), CT decay (k f,CT ), CT dissociation into free carriers (k d,CT ), free carrier encounter to form CT (k rec ), and reformation of the singlet exciton (k ex,ref ).Reproduced with permission. [195]Copyright 2023, American Chemical Society.
Upon reducing the HOMO-HOMO energetic offset, an additional loss pathway may also occur from exciton reformation, with a subsequent decay to the ground state.In PM6:Y6 BHJ the offset is found to be 0.3 eV, [148] resulting in an electroluminescence spectrum of the blend that signifies Y6 excitons only (giving no information of the CT state manifold).To understand this contribution, Perdigón Toro et al. performed a comprehensive study of the absorption and emission from the blend of the donor polymer PM6 with Y6.It was found that photon emission from the blend is almost entirely determined by the re-occupation of the Y6 singlet due to free charge recombination. [123]Despite this, only less than 1% of the recombination proceeds through the S 1 state upon reformation from the CT state and 99% decay via the CT state.Whilst this is only 1%, the recombination of free carriers through singlet exciton reformation now adds a new loss channel for free carriers (Figure 12e) such that the equation describing recombination of free carriers is modified to: (11)  where k f,CT is the CT-state decay rate, k f,S1 is the exciton decay rate,  CT,diss is the probability of charge generation,  ex,ref is the probability of exciton reformation and n S 1 , n CT , and n CS are the densities of singlet excitons, CT states, and free charges, respectively. [195]s thoroughly elucidated in the study conducted by Sandberg et al., [173] it is crucial to incorporate the charge generation efficiency into the equation mentioned above.Another consideration is molecular orientation.As mentioned earlier, Fu et al. studied the effect of Y6 molecule orientation and its consequence on energetics and charge dynamics using differential capacitance and transient photovoltage analyses by altering the orientation of Y6 from a face-on configuration to a more edgeon configuration through the use various processing solvents, a notable shift in energy is obtained (as discussed above).The authors reported that this shift in energy significantly influences the rate of non-geminate recombination in both bilayer and BHJ devices, by 50 times.This phenomenon was assigned to the molecular quadrupole moment and band bending as noted earlier. [155]

Charge Carrier Mobility
Amongst many factors, the photo-induced optoelectronic properties of a material and its device performance are directly influenced by the carrier transport.The charge carrier mobility is a key performance criterion for FF (since extraction competes with recombination).The current paradigm for mobility of photogenerated charge carriers is that one must consider carrier density regimes and transient effects.Depending upon experimental techniques employed, the obtained carrier mobility in organic semiconductors varies widely.Often space charge limited current (SCLC) technique is employed to measure mobility.SCLC experiments performed on single carrier diodes having architectures quite different from that of a complete solar cell reveal scattering values of charge mobility in PM6:Y6 varying from 10 −5 to 10 −3 cm 2 V −1 s −1 . [106,145,19,196,197]Challenges in fabricating single carrier devices that exhibit space charge limited current response are one reason of the scattering charge mobility results.Other factors such as thickness, solvent, substrate-sensitive orientation, contacts, and better injection currents also will affect the extracted mobility values.Due to the strong aggregation/crystallinity of Y6, PM6:Y6 blend has been observed to have a thickness dependent mobility.Typically, free electrons and holes are delocalized within the conjugated segments (one or two repeat units).The separate segments are bound by weak van der Waals forces, influenced by dynamic and static disorder.In a study by Hosseini et al., [190] using GIWAXS measurements it was observed that by increasing the thickness of the active layer from 100 nm to 400 nm, Y6's crystallinity enhanced (at the cost of PM6) while reducing its energetic disorder from 59 to 54 meV.The more ordered Y6 resulted in increased electron mobility (by an order of magnitude).However, since this increase came at the expense of reduced PM6 mobility, the overall effect was not beneficial to device performance (for example due to space charge effects limiting FF).However, by regulating the energetic disorder of PM6 through active layer engineering with additive solvents (65 to 61 meV), the mobility of PM6 in the thick junction blend was also enhanced, which then resulted in higher performance of the thick device.

Molecular Doping
In an effort to control and ameliorate the transport properties, molecular doping of the BHJ layer has been intentionally considered. [198]In 2020 Anthopoulos and co-workers studied the effect of the n-type dopant benzyl viologen (BV) in the active layer PM6:Y6. [199]The authors demonstrated that a small amount of dopant (0.004 wt%) improved J SC (25.1 to 26 mA cm −2 ) and FF (73% to 74%) of PM6:Y6 device.This improvement was assigned to trap density reduction which was reflected in mobility and recombination studies as well.In another study by Xie et al.N-DMBI in PM6:Y6 was considered. [200]In terms of device performance, the champion device was obtained with 0.005 wt% doping-delivering an efficiency of 15.34%-which reflected an increase in J SC (26.41 mA cm −2 ) and V oc (0.86 V).The enhanced performance was assigned to reduction in trap assisted recombination which also improved transport and charge pathways (reduced trap density due to increased free carrier concentration).This is consistent with the author's findings that upon doping crystallization of active layer is facilitated and the crystal coherence length is elongated.However surprisingly, the FF was reduced with addition of the dopant at any concentration.Yet another successful n-doping study was reported by Li et el on using DMBI-BDZC. [201]When using 0.02 wt% dopant, the device efficiency improved from 17.17% to 18.33%.From light intensity dependent measurements, the authors suggest that trap-assisted recombination had suppressed in the doped sample.In particular, it was proposed that DMBI-BDZC dopes Y6 in the BHJ blend, leading to enhanced and balanced charge carrier mobilities (from a μ e /μ h 1.14 to 1.07) and slightly longer carrier lifetime.Similar observations in terms of role of trap and morphology were made by Fu et al. [202] when employing DCIB as an additive/ dopant.Another common theme between all of the reported work is their observation of enhanced charge generation with the optimum doping concentration.This may be due to the modified morphology where the dopant can act as a microstructure modifier.

Open Circuit Voltage
In an idealized solar cell, the radiative voltage, V rad oc , gives an upper limit for the open-circuit voltage, when recombination occurs from the lowest energy excited state to the ground state.As these transitions are directly related to the inverse process of absorption, the presence of radiative voltage losses is inevitable.However, there is always significant non-radiative recombination taking place simultaneously, reducing the V oc .A complete picture of the total voltage losses in OSCs contains both recombination losses as well as electron-transfer losses due to the conversion of strongly bound excitons in the neat material into CT states.The open circuit voltage can be described according to: where G is the charge carrier generation rate, E CT the energy difference between the ground state and CT state, k f, rad the radiative  [203] and b) dispersive non-geminate recombination [205] in an organic layer with inhomogeneously broadened DOSs.Reprinted with permission. [203]Copyright 2017, Wiley-VCH.Reprinted (adapted) with permission. [205]Copyright 2019, American Chemical Society.The energetic relaxation of carriers within the DOS slows down carrier transport but also NRG. c) J-V characteristics of a 115 nm thick PM6:Y6 blend measured at different temperatures (lines).The symbols display the best fits with a kinetic Monte Carlo (kMC) code.d) Experimental V OC as a function of temperature from the data in (c) (solid squares).These data can be well reproduced with kMC simulations which include hot carrier effects (red lines and symbols) while drift-diffusion yields a too-small V OC (blue lines and symbols). [170]Copyright 2021 American Chemical Society.Electro-optical simulation of the fill factor (e) and the PCE (f) for PM6:Y6 devices as a function of active layer thickness (red lines).The blue line and symbols in Figure 13f are for a blend of PM6 with the Y-series NFA BTP-eC9.The same set of input parameters (mobility, bandgap, NGR coefficient) was used for all layer thicknesses, indicating that charges equilibrated before being extracted or recombining. [130] state decay rate, k f the total CT decay rate, N CT the density of DA interfaces.The V oc value reported for PM6:Y6 is typically 0.85 V, which means that the QFLS is significantly smaller than the energy of the photovoltaic bandgap (1.38 V). [187] This system does not exhibit a distinct, sub-gap CT band even when measured with ultra-sensitive EQE (Figure 12c), [123] explaining the significant variations in the reported values for the CT energy in the literature.[142,145,98] The V rad oc has been calculated to be 1.08 V (from convoluting EQE with the blackbody photon flux), ca.0.30 eV below the photovoltaic bandgap.[123] The remaining voltage loss, ΔV oc, nrad originates from the non-radiative recombination.Using electroluminescence quantum yield measurements, ΔV oc, nrad of this system has been reported to be 0.27 V; sitting on the lower end of the non-radiative voltage loss spectrum (ranging from 0.5 to 0.15 V) for organic solar cells.We note that the low ΔV oc, nrad value in PM6:Y6 is a result of the relatively small CT-S 1 energy offset, resulting in observation of the S 1 state entirely (due to the much higher oscillator strength of Y6 S 1 state compared to the CT state).However, only 1% of recombination of free carriers proceeds via the S 1 state, and still 99% of recombination is from a very weakly emitting CT state as discussed earlier.As such, the presence of the singlet exciton reduces the non-radiative voltage loss, but not because it is more emissive but because its stronger absorption sets an upper limit to the V rad oc .

The Role of Energetic Disorder and Hot Carriers
A heavily discussed issue in relation to inhomogeneous energetic disorder is the role of hot carriers and dispersive phenomena.Dispersive phenomena occur when photogenerated charge carriers initially populate high-energy sites in the DOS distribution and then undergo thermalization.There are two important consequences.The first is that charges exhibit a time-dependent mobility, meaning that their drift mobility slows down with time while charges occupy states deeper and deeper in the DOS. [169]his is depicted schematically in Figure 13a.It has indeed been shown that photogenerated charges exit the active layer at the contacts faster than they thermalize, even at maximum power point conditions. [203]As such, mobility values determined at longer timescale or by steady-state experiments such as SCLC may not be relevant in solar cell working condition.The other phenomenon of relevance for solar cell performance is dispersive NGR.As stated above, NG recombination is driven by charge encounter.Charges deep in the DOS will need to be detrapped to transport levels to be able to meet the opposite carrier.Indeed, theory and experiments revealed a significant slow-down of the recombination rate with time, as schematically shown in Figure 13b. [192,204]This raises the question whether or not PM6:Y6 devices benefit from dispersive phenomena.For example, Kemerink and coworkers could successfully reproduce the measured J-V characteristics of a 115 nm thick PM6:Y6 blend at different temperatures with kinetic Monte-Carlo simulations (kMC), which explicitly take into account hot carrier motion and recombination, see Figure 13c. [170]More importantly, their kMC simulations predicted that the V OC of PM6:Y6 benefits substantially (by 0.13 V) from the slow thermalization of photogenerated (hot) carriers as shown in Figure 13d.This was attributed to the fact that such hot carriers recombine (or exit the active layer through the electrodes) long before they acquire thermal equilibrium.On the other hand, simple drift-diffusion simulations that only considers equilibrated carriers with a constant mobility and recombination rate yielded a too-low V OC .There is, however, other work that supports the model that equilibrated charges determine the steady-state device properties.For example, Armin and coworkers investigated PM6:Y6 devices with different active layer thickness. [130]The optical transfer matrix approach was used to simulate the optical field distribution within the active layer, which was then applied as the charge generation profile in DD simulations.Surprisingly, the photovoltaic parameters of all devices could be fitted with the same values of the electron and hole mobilities, the bandgap, and the recombination coefficient, for all thicknesses.This is shown for the FF and the PCE in Figure 13e,f, respectively.Since the FF is very sensitive to the extraction-recombination balance, dispersive effects would have shown up by a stronger effect dependence of the FF on the active layer thickness, simply because a thin active layer would have benefited more from hot carrier extraction than a thick device.This was apparently not the case.Also, Perdigón Toro et al. succeeded in reproducing the temperature dependence of the V OC for different illumination intensities through an analytical model as discussed earlier.
Here, the experimentally determined values for the energetic disorder from SCLC and the temperature-dependent carrier density from photoinduced absorption spectroscopy served as input, with the only fit parameter being the HOMO-LUMO separation.The fit to the experimental V OC (T) yielded E CS = 1.42 eV, in very good agreement with the results from spectroelectro-chemistry and energy-resolved electrochemical impedance spectroscopy shown in Figure 8b,f.This, again, supports the view that equilibrated rather than hot carriers govern the device performance. [141]

Morphology and How to Tailor It Toward Optimum Device Performance
Structure-processing-property relationships of BHJ-based OSC structure have been difficult to achieve and predict, and PM6:Y6 BHJ is not an exception. [206]The morphology of a PM6:Y6 BHJ involves D:A interactions, molecular packing, domain size and purity, film crystallinity, and 3D D/A networks, whose length scales range from several micrometers to sub-nanometer. [33]herefore, a set of morphological techniques is often required to reveal a complete characteristic picture of this BHJ, including atomic force microscopy (AFM), photoconductive (pc-) AFM, transmission electron microscopy (TEM), grazing incidence wide-angle x-ray scattering (GIWAXS), resonant soft xray scattering (RSoXS), grazing incidence small-angle X-ray scattering (GISAXS), and solid-state (ss) NMR (ss-NMR).A summary of frequently used characterization techniques and their length scale is illustrated in Figure 14.

Morphology of PM6:Y6-Based OPV under Common Film Processing Control Methods
The original report by Yuan et al. on PM6:Y6 OPV demonstrated that excellent PCEs of >15% were obtained when using chloroform (CF) as a processing solvent. [19]2D GIWAXS patterns showed that the optimized PM6:Y6 blend film displayed a strong diffraction peak in the out-of-plane (OOP) direction associated with the - stacking of Y6, while a scattering peak in the inplane (IP) direction assigned to the lamellar stacking of either PM6 or Y6.The backbone ordering of Y6 is maintained in the blend film, and the active layer exhibits nano-fibrillar structure with an acceptor domain size of ≈44 nm derived from GISAXS profiles. [19]However, similar to other BHJ-based OPV systems, PM6:Y6 film morphology and device performance strongly depend on processing conditions. [19,207]Extensive works on binary PM6:Y6 OPVs have indicated that the PCE of devices in the Figure 14.Length scale of different characterization techniques used to understand the OSC BHJ morphology: solid-state NMR, grazing incidence wide-angle scattering, transmission electron microscopy, resonant soft X-ray scattering, and atomic force microscopy.Reproduced with permission. [33]opyright 2021, Wiley-VCH.
The morphological properties of the BHJ are influenced heavily by the processing solvent due to differences in solubility and solvent evaporation process, which result in very different optoelectrical behaviors at the macroscale. [214]For example, Liu et al. showed that Y6 films processed from chlorobenzene (CB) consisted of polycrystalline domains with no preferred domain orientation, while CF-processed Y6 films displayed face-on, polymerlike extended crystal transport channels (Figure 15a). [98]These morphological differences were associated with the differences in macroscale photovoltaic performances, with CB-processed devices reaching only 12% PCE and CF devices achieving up to 16.9%, depending on the thermal annealing conditions used. [98]s outlined in Sections 3.1 and 4.1, the different orientation and packing of Y6 molecules in CB and CF-processed layers have a significant effect on the blend energetics, and consequently on the device parameters.Further molecular-level understanding of BHJs could be achieved by employing a short-range technique such as ssNMR.Luginbuhl et al. embarked upon an investigation to meticulously analyze and discern the intricate differences in morphology and performance exhibited by PM6:Y6 BHJs processed by CF, CB, and o-xylene (o-XY).CF casted devices yielded the best performance of 15% followed by CB and o-XY casted devices (10.45% and 9.66%). 19F NMR reveals that PM6 films are insensitive to change in the casted solvent whereas Y6 films show different solid-state 19 F signals depending on which solvent the film was cast from (Figure 15b).The combined results from scanning probe microscopy, ssNMR, X-ray crystallography, DFT calculations, and molecular dynamics (MD) simulations reveal that the choice of solvent has a large impact on the resulting solidstate interactions between the Y6 end groups and the aliphatic sidechains, both within the same molecule and from neighboring Y6 molecules. [206]The relative orientations of the sidechains and end groups of Y6 molecules to their fused-ring cores dictate the resulting morphology and overall performance of the solar cells. [206]t is worth noting that in many cases, photovoltaic performances decline significantly when moving from chlorinated solvents (such as CF and CB) to less volatile and environmentally friendly solvents (such as o-XY and toluene).To mitigate this effect, hot-solution casting is a widely-used strategy.High temperature improves the material solubility and interplay between solvent evaporation (kinetics) and materials miscibility (thermodynamics), leading to the attainment of desired morphology and charge transport. [209,214,215]As a showcase to the utilization of non-chlorinated solvents in high-performance OPV, aromatic hydrocarbon solvent 1,2,4-trimethylbenzene (TMB) was employed to process PM6:Y-series BHJs (including Y6). [215]PM6:Y6 active layer casted from hot TMB (at 120 °C) exhibited an optimum balance between the strong phase separation (represents by the length of the aggregated acceptor domain, Rg) and strong  [98] Copyright 2020, Wiley-VCH.Solid-state 1D 19 F NMR spectra of c-e) PM6, f-h) Y6, and i-k) PM6:Y6 BHJ films processed from CF, CB, and o-XY solvents. 19F signals correspond to PM6 and Y6 moieties as indicated.Figures (c-k) adapted with permission. [206]Copyright 2022, Wiley-VCH.
Figure 16.Choice of solvent, morphology, charge mobility, and device performation: a) Relationship between the choice of solvent, PCE, Rg/ ratio, and charge mobilities of PM6:Y6 BHJ processed from different solvents.Reproduced with permission. [215]Copyright 2021, Wiley-VCH.b) d-spacing, peak area, and crystal coherence length of (11-1) diffraction planes of 2D GIWAXS scattering profiles of PM6:Y6 blends for different processing solvents and thermal annealing temperatures.Reproduced with permission. [98]Copyright 2020, Wiley-VCH.c) Schematic drawings of the D/A interfaces PM6:Y6 blend films when NA, FN, CN, and BN were used as an additive, respectively.Reproduced with permission. [216]Copyright 2021, Royal Society of Chemistry.
self-aggregation (represented by the length of the intermixed phase, ), giving an intermediate desirable Rg/ ratio of ≈0.4.Charge mobilities in TMB are comparable among BHJs processed from CF and CB (Figure 16a).As a result, the high PCE of ≈15% is preserved. [215]ost-processing thermal annealing is among the most common protocols for fine-control film formation, inducing better contact/interface, as well as removing residual solvents. [217]lthough various annealing temperatures and durations have been reported, the optimized thermal annealing conditions tend to stay close to 100 °C for a period of 10 min without much deviation. [34]Figure 16b suggests that the intermolecular distance and crystal coherence length (CCL) of a PM6:Y6 BHJ (CFprocessed) were improved after thermal annealing. [98]Specifically, the molecular packing in the (11-1) and (020) directions exhibited noticeable improvements in their peak areas and crystal coherence lengths after thermal annealing, with the maximum improvement achieved at 80 °C.The packing distances also reached their minimum value under this condition.This enhancement leads to better charge transport, which is wellcorrelated with device performance.The PCEs of CF-processed OPVs under different conditions of as-cast, annealed-60 °C, annealed-80 °C, and annealed-100 °C were 15.68%, 16.19%, 16.54%, and 16.21%, respectively. [98]n contrast to post-processing thermal annealing, solvent additives have been used to control the kinetics and thermodynamics during film formation.The most commonly used solvent additive for fine-tuning the morphology of PM6:Y6 OPV is 1-chloronaphthalene (CN), which has been shown to improve efficiency to 15.7% compared to the non-additive (NA) version, which has an efficiency of 15%. [19,216]In a prominent work by Lv et al., the impact of naphthalene derivative sol-vent additives (including 1-fluoronaphthalene (FN), CN, and 1bromonaphthalene (BN)) on the morphology and photovoltaic performance of PM6:Y6 OPVs were explored. [216]Different solvents exhibited a progressive increase in the miscibility of PM6 and Y6 (assessed by the Flory-Huggins interaction parameter, ) as follows: CN < FN < BN < NA.By employing several morphological characterization methods (AFM, TEM, x-ray techniques, and optical spectroscopy), morphological differences in these four blend systems were revealed and are illustrated in Figure 16c: i) lowest crystallinity and highest miscibility of PM6 and Y6 in the NA-based film led to well-mixed but fewer aggregation stacks; ii) BHJ cast with FN-additive has medium D and A miscibility and improved crystallinity, which facilitates efficient exciton dissociation and charge transport due to well-suited D/A interface and continuous pure phase; iii) low miscibility and excessive pure domains/phase separation are found in CN processed PM6:Y6, which reduces exciton dissociation efficiency; and iv) BN-induced high miscibility of D and A causes overmixing and insufficient phase segregation, which also lessens the exciton dissociation efficiency.As a result, 17.5% PCE achieved by the FN-induced miscibility control strategy stands as the champion PCE value among these systems. [216]Besides thermal annealing, solvent vapor annealing has been shown to improve the PM6:Y6 performance.Ge et al. showed that PM6:Y6 film exposed to carbon disulfide (CS 2 ) has a higher degree of crystallinity and better charge transport than as-cast PM6:Y6 film, resulting in even better PCE (18.01%versus 16.57%). [218]inally, it is worth noting that several of the most successful binary PM6:Y6-based OPVs have attained high performance only after a considerable effort to achieve an optimal morphology.Typically, an iterative optimization process is required, and very frequently consists of a combination of two or more optimization protocols.Some common optimization parameters are blend ratios, coating techniques, film thicknesses, thermal annealing times, solvents and times for solvent vapor annealing, additive concentrations, etc. [219,220] Shortly after the first demonstration of high-performance binary PM6:Y6 OPVs, several reports employed ternary and quaternary strategies and elevated the PCEs of PM6:Y6-based OSC to over 18%. [221,26,34]The added components varied from small molecule donor, [222][223][224][225][226][227][228][229] polymer donor, [230][231][232][233][234][235][236] fullerenebased acceptor, [109,[237][238][239][240] NFA,  and polymer acceptor [263] (Figure 17). Adding tird and fourth components offer several benefits such as improved absorption in combination with host materials (PM6 and Y6) to increase the J SC value, adjustment of energy levels to enhance the V OC , and optimization of active layer morphology for favorable performance in dissociation, charge transport, and collection, as well as recombination suppression, ultimately leading to better J SC and FF values.[264,27,34] In the following, we outline a few works with well-established structure-processing-property relationships.
Adding a small molecule donor as a third component in ternary OSCs is advantageous for achieving a higher V OC and optimized morphology of the active layer, as it leads to a deeper HOMO energy level and appropriate film crystallinity. [264,34]ased on this idea, Li et al. employed a highly crystalline small molecule donor DRTB-T-C4 into the host PM6:Y6. [222]Using the parameter wetting coefficient, the authors predicted that DRTB-T-C4 is distributed at the interface between PM6 and Y6, and thus forms a cascade-type junction that promotes exciton dissociation and charge transfer.The high crystallinity of DRTB-T-C4 with an appropriate coherence length of - stacking enabled improved charge transport, increased the hole and electron mobilities of the optimized ternary blends, balanced the charge carrier density ratio, and ultimately improved FF and PCE. [222]In contrast, compared to small molecule donors, third-component polymer donors show less tendency to be crystalline, and their corresponding ternary OPVs are seen more often to operate as alloy and parallel-like models.As an example, An et al. incorporated a polymer donor S3 that has a very similar chemical structure to PM6, leading to an alloy-like state of two donors in the BHJ film. [231]The ternary blend film exhibits a distinct fiber-like structure and phase separation as observed in both TEM images and AFM phase images, while the x-ray scattering profile of the ternary and PM6:Y6 are relatively unchanged.These structural features are favorable for exciton dissociation, as well as efficient charge transport and collection, resulting in a highly improved FF value. [231]dding an acceptor as a third component, on the other hand, results in different morphological characteristics with associated photovoltaic enhancement.PC 61 BM and PC 71 BM are fullerene-based acceptors that have been used in the first ternary blends to show improved performance.][239][240] Fullerenebased acceptors, whose sizes are relatively small in comparison to PM6 and Y6, are usually distributed uniformly in the active layer and only gently affect the packing of PM6 and Y6. [109]Nevertheless, the incorporation of fullerene-based acceptors facilitates the electron mobility and balances charge mobility ratio, reduces recombination losses, and sometimes, helps to achieve complementary absorption with the host material.
NFAs, with the flexibility of absorption spectra and energy levels design, became the most popular choice as the third component of PM6:Y6-based ternary OSCs.Interestingly, while the NFA pool is large, many of the successful ternary OSC (using NFA as a third component, with PCE > 17%) share the common characteristic of Y6, causing the third-component NFAs forming an alloy-like acceptor phase in the optimized films. [244,245,247,254,257,259]n example can be seen in Figure 18a,b, where An et al. introduced the MF1 acceptor to blend with PM6:Y6.[257] Experimental results obtained from spectroscopy, contact angle measurements, cyclic voltammetry, and morphology analysis indicated that the small molecular acceptors Y6 and MF1 exhibit strong compatibility.In particular, the Raman mapping of ternary blend films demonstrates that MF1 preferred to form an alloy-like state with Y6, as evidenced by the majority of yellow-green spots (representing MF1) being embedded in the red zone (representing Y6) (Figure 18a).In addition, the apparent diffraction peak at ≈0.4 Å −1 in the in-plane direction originating from MF1 is observed in the 2D GIWAXS profile of PM6:MF1 blend films but is absent in ternary film profiles, which suggests MF1 should prefer to mix well with Y6, rather than form some individual domains (Figure 18b).[257] Finally, pseudo-layer-by-layer PM6:Y6 ternary OPV with suitable vertical phase separation has been demonstrated, showing excellent PCE > 18%.[265] The number of added components is not limited to one and obviously can be extended to benefit more diverse morphological/optoelectrical enhancements brought by different types of molecules.Despite the promising nature of these control techniques and their demonstrated ability to achieve very high device performance, [221,266] their implementation demands extensive effort through a trial-and-error approach. Repreentative work in this category was carried out by Zhang et al., where PM7 and PC 71 BM were added to form a quaternary BHJ.While PM6 and PM7 mixed intimately and formed fibrillar networks for hole transport, both PM7 and PC 71 BM slightly meddled with the packing of Y6 and improved the crystallization (Figure 18c).The CCL and peak area for both the lamellar and - stacking peaks were found to be the largest in the quaternary blend in comparison to these of binary PM6:Y6 and ternary PM6:Y6:PC 71 BM blends, which implies improved crystallinity and crystal quality in the quaternary blend (Figure 18d).These morphology improvements contribute toward the excellent PCE of > 18% and enhanced photostability and storage stability.[221] Arguably, after molecular design, D/A band alignment, and active layer optimization, the next most important aspect is interfacial and electrode engineering.By modifying the interfacial contact with the active layer, PCEs of several binary PM6:Y6 OPVs have been demonstrated to surpass 17%.[218,[267][268][269][270][271][272][273][274][275][276] Interestingly, among these works with impressive performances (PCEs ranging from 17.0% to 18.01%), the conventional device structure using ITO/PEDOT:PSS or its derivation as a hole transport layer is used.[218,268,[270][271][272][273]275,276] On the other hand, PFNbased, perylene diimide (PDI)-based, and naphthalene diimide (NDI)-based electron transport layers are the materials of choice (Figure 17c,d).To the best of our knowledge, inverted device structure (ITO/ZnO/PM6:Y6/MoO x /Ag) has been reported with the best PCE of 17.1%.[274]

Toward the Commercialization
As the PCE of PM6:Y6 OSC now approaches commercial viability, research that paves the way for commercialization is increasingly invested.In this section, we highlight the remaining challenges i) device stability and storage/operation life-time, ii) materials batch-to-batch variation, and iii) module efficiency of PM6:Y6.
or light-induced reactions of the photoactive layer [278,279] whereas morphological degradation occurs due to the diffusion of the donor and acceptor overtime. [277]It is generally observed that the binary PM6:Y6 OPVs do not maintain stable morphologies, leading to performance drops during thermal stress/photoillumination testing, and sometimes even under storage conditions (in the dark and inside a glovebox).Specifically, the stability of PM6:Y6 OPV in the literature has been assessed under a few test protocols.It is noted that these test protocols are also used for all PVs, and below we provide brief descriptions of the test setup and test IDs assigned by consensus stability testing protocols. [280]Shelf storage test: In the dark, room temperature, ambient/inert air, open circuit load (ID: ISOS-D-1).The intrinsic morphological stability under storage conditions (shelf-or high-temperature storage) of PM6:Y6 blend and other polymer:NFA systems were methodologically examined by Ghasemi et al. (Figure 19). [277]This work elucidated why certain combinations of polymers and NFA exhibited excellent resistance to heat stress and long shelf life in device stability.However, Y6 and its derivatives did not yield stable morphologies in binary blends.In detail, a layer of Y6 atop the PM6 films was fabricated using the water-transfer method, and then annealed at different temperatures/durations.By analyzing the composition profile of PM6 and Y6 using time-of-flight secondary ion mass spectrometry (SIMS), the diffusion coefficient D(T) of Y6 on PM6 in a bilayer structure ≈4.1 × 10 −17 cm 2 s −1 at 90 °C was evaluated. [277]This value indicates that a Y6 acceptor molecule can diffuse 20 nm on a time scale of ≈1 day.Thermal  [277] Copyright 2021, Springer Nature.
stabilities of Y6 and other BTP-based BHJs have been correlated with proxy metrics such as the thermal transition temperatures (T g ) and elastic modulus of the donor polymer. [277,281]In particular, a higher T g component is desirable to achieve higher thermal stability.
Extensive research has been conducted on the photostability of PM6:Y6, revealing intriguingly diverse conclusions regarding the underlying causes of instability.Zhao et al. found that the UV portion in the solar spectrum contributed majorly to the active layer degradation. [282]J SC was found to be very stable, while PCE degradation was ascribed to decreases in V OC and FF.By a series of control experiments with donor-exposed-, acceptor-exposed-, and the blend-exposed-devices, it is inferred that the photo-degradation of PM6 dominates the photo-degradation process of the PM6:Y6based device. [282]A recent report by Wang et al. identified a light-induced backbone-twisting in a shared structural motif of PM6 and D18 as a dominant degradation mechanism, [283] while Y6 was found to have excellent photostability.All device performance parameters (V OC , J SC , and FF) decreased significantly, which contributed to the ≈90% drop in PCE after just 12 h of illumination in ambient air. [283]In contrast, some other research groups found that the photochemical decomposition of the Yseries is responsible for performance degradation. [284]Liu et al. found that the vinylene linkages between DA'D core and A moieties are susceptible to cleavage, based on mass spectra of aged Y6 and its derivative NFAs.Excellent stability of donor PM6 was observed in this case.From the existing literature, it is clear that the performance-photostability conundrum of PM6:Y6 is still not well understood.
Besides the chemical degradation and morphological instability, [277,281] the deterioration of electrodes and interlayers represents another pathway for degradation.In conventional structures, while PEDOT:PSS and other oxide-based HTLs (e.g., NiO x and MoO x ) commonly used in literature are relatively stable during thermal and photostability tests, ETLs that enable high device performance (including PFN-based, PDI-based, NDI-based ETLs, Figure 17c,d) are more prone to degrade under thermal stress and photo illumination. [26,34]The choice of the top cathode electrode (e.g., Ag, Al, or Cu) was also demonstrated to affect the stability of the PM6:Y6 OPV. [268]On the other hand, in inverted structures, ZnO ETL is generally considered the bottleneck of photostability due to its light-soaking effects. [285,286]295] For quality control purposes, it is important to highlight the batch-to-batch variations of polymer donors and their influence on the binary polymer:NFA film morphology and solar cell performances.298][299][300][301][302][303][304][305] Karki et al. found a strong correlation between low-M w -fractions (LMWF) in different batches of PM6, BHJ film morphology, and PCEs.In particular, BHJ films using a PM6 batch with 52% LMWF showed significantly lower nanoscale photocurrent (by pc-AFM), poorer molecular ordering (by GIWAXS), and >2 times broader long-period (by RSoXs), and PCE of ∼5% compared to 1% LMWF BHJ devices which have the PCE ≈ 15%. [156]Advanced ssNMR analyses provide insight into the interaction of D and A molecules at sub-nm length scales (Figure 20).Aliphatic and aromatic 1 H and 13 13 C-1 H heteronuclear correlation NMR spectrum of 52% LMWF PM6:Y6 blend.Reproduced with permission. [156]Copyright 2020, Royal Society of Chemistry.
aromatic groups in the 1% LMWF blend but not in the 52% LMWF blend. [156]PM6:Y6 blend with 1% LMWF shows ultrafast hole transfer and has efficient charge generation, charge transport, and charge extraction leading to superior performance as compared to PM6:Y6 blend with 52% LMWF.The effect of M w and polydispersity index (PDI) of PM6 on BHJ film morphology and macroscopic device performance were studied by Liu et al. [274] The study found that the domain sizes were 39.2, 18.6, and 16.4 nm, with the relative domains were 0.24, 0.61, and 1, for blend films based on 41, 74, and 114 kDa PM6, respectively.Better face-on ordering and molecular packing, and more ideal domain sizes were generally found with high M w BHJ film, however, a stronger tendency of donor aggregation was also observed.As a result, a medium M W of ≈74 kDa was found to provide the ideal size and purity of the domains for the optimal PCE of 17.1%. [274]Moving forward, it is crucial to control the M w of the donor polymers or searching for polymers where device performance is resilient toward the M w . [306]p-scaling while preserving the cell efficiency is another challenge toward commercialization.Numerous efforts have already been made to tackle the technological challenges associated with scaling up efficient cells to achieve efficient modules, however, cell-to-module efficiency loss still remains.These losses are primarily attributed to the transition from spin-coating to scalable printing methods for processing, along with the challenges posed by inhomogeneities and variations in film quality when scaling up to larger areas. [307,308]Sun et al. demonstrated a layer-by-layer PM6:Y6 OPV module of 11.52 cm 2 with PCE of 11.86%, by subsequently blade-coating PEDOT:PSS, PM6, Y6, and PNDIT-F3N-Br. [307]In another work, Zhang et al. employed forward/reverse blade-coating to realize a 36 cm 2 photovoltaic module with PCE of 13.47%. [309]Notably, a recent work by Zhao et al. developed a low-molecular-weight PM6-doping approach to overcome the inherent fabrication limitation of the PM6:Y6 active layer in terms of high-speed coating preparation, successfully achieved above 15% PCE for ≈7.5 cm 2 module. [310]It is important to understand the loss mechanisms in OPV modules, which are partly engineering and partly related to physics of scaling up disordered semiconductors.This highlights the importance of characterizing [311] and understanding NGR in order to suppress recombination (as highlighted in Section 4.2).

Summary and Perspective
Since 2019, extensive research has been conducted on PM6:Y6 solar cells with the aim of understanding the reasons behind their high power conversion efficiency.We have outlined in this Review that there is consensus on some of the physical processes and properties, while others are heavily disputed.In the following, we summarize the most important findings and conclusions: PM6 comprises two highly symmetric and planar building blocks.As a consequence, PM6 is already aggregated in the solution state and forms well-aggregated domains in the solid state, as neat material and in the blend with various acceptors; Y6 and related Y-series NFAs exhibit a unique ADA'DA chemical structure and a curved molecular shape.This allows for multiple intermolecular interactions in the solid state, among which dimers with a strong spatial proximity between the donor core and the acceptor terminal are of paramount importance.By virtue of these interactions, the energy of the first excited state in neat Y6 film is significantly red-shifted compared to that of isolated single Y6 molecule but also carries a strong CT character.This in turn causes relatively slow non-radiative decay rates, a long fluorescence lifetime and in combination with small energetic disorder a long exciton diffusion length.The exciton energetic disorder is small, a characteristic attributed to the substantial order within Y6 domains.However, it is crucial to acknowledge the potential role of exciton delocalization in this scenario.
Expanding our understanding of Y6's performance requires comprehensive information on the energy and energetic broadening of the CT state; a challenge as its absorption and emission signals are overshadowed by the more prominent transitions between the ground state and the Y6 singlet state.This asks for the advancement of techniques and models to unravel the intricacies PM6:Y6 puzzle and to understand why it performs so well.Electroabsorption measurements suggest a CT energy of 1.27 eV, ≈150 meV below Y6 LE, although further confirmation is warranted.Quantum chemical calculations offer insight into a rather strong delocalization of the CT state, which would be beneficial for its separation into free charge.
Moving to the energy levels, despite intense studies, there is no consensus about the HOMO-HOMO offset, ΔE HOMO , and the energy of the charge-separated state.While initial cyclovoltammetry on neat layers suggested a small HOMO-HOMO offset of ca.0.1 eV, the results from photoelectron spectroscopy yielded much larger values (0.5 eV and above).The value of ΔE HOMO is crucial for the development of models to understand the device properties.If ΔE HOMO is too small, it may not provide enough driving force for free charge generation while a too large ΔE HOMO results in small energy of the charge-separated state that may be inconsistent with the large V OC of the blend.Recent work suggests a HOMO-HOMO offset of ≈0.3 eV, which is just enough to overcome the exciton binding on Y6.
There is consensus that free charge generation in PM6:Y6 is efficient and depends little on temperature.The physical mechanisms behind this high efficiency is yet not clear, be it a cascaded energy landscape, the presence of quadrupole moments or energetic disorder.More recently, spontaneous exciton dissociation has been demonstrated in neat Y6 layers, which may indeed contribute to free charge generation in Y6-based blends.
Notwithstanding the sufficient carrier drift length at short circuit condition to guarantee extraction of almost all charges, resulting in high J SC , the fill factor is limited by a too small diffusion length of ca.half of the active layer thickness.This in turn is related to rather fast non-geminate recombination, probably involving electron back transfer to the Y6 local triplet state.PM6:Y6 system demonstrates a V OC of 0.85 V, with a radiative voltage loss V OC,rad of 0.30 eV below the photovoltaic bandgap.Non-radiative voltage losses V OC,nr account for 0.27 V, primarily attributed to recombination from the weakly emitting CT state.The presence of the singlet exciton in PM6:Y6 helps in mitigat-ing non-radiative voltage losses, primarily due to its strong absorption setting an upper limit on V OC,rad .The improvements observed from n-doping of PM6:Y6 have been attributed to the reduction in trap density and improved charge carrier mobility and recombination due to facilitated crystallization and elongated crystal coherence length.Regarding energetic disorder, the free carriers have been measured to have relatively small energy broadening, amongst the lowest values reported for organic systems.Related to this, most studies suggest that the device performance does not benefit from hot carriers.Rather than that, photogenerated charges thermalize before they recombine or get extracted.
Related to the importance of Y6 interactions is the large sensitivity of the PCE of PM6:Y6 on the processing solvent, postprocessing treatment, additives, etc. Structural studies revealed that the Y6 local morphology is much more sensitive to solvent than the morphology of PM6.This offers a way to fine-tune the layer composition and preparation conditions toward optimum performance.In fact, third or fourth components have been used to tune the blend morphology to improve the PCE with great success.As for most BHJs, stability is an important issue.Y6 has a rather high molecular diffusion coefficient in PM6, meaning that the blend morphology might be intrinsically unstable.In addition, photoinduced degradation processes were studied.While there are still conflicting results, there is growing evidence that PM6 and related donor polymers suffer from light-induced backbone-twisting while Y6 was found to have excellent photostability.This is a very promising result, in particular in view of the application of Y6 in semitransparent devices or tandem cells.Finally, reproducibility is an important issue, given the strong dependence of the device performance not only on processing conditions but also on the PM6 molecular weight.
To conclude, the knowledge about the structural and function properties of the PM6:Y6 blend has developed rapidly during the past years.Yet, some of the underlying physical processes are still heavily debated.Here, we anticipate gaining further insight through the future use of novel characterization methods and advancement of quantum-chemical simulation.Enhancing the device performance through chemical modification will heavily depend on these advancements, which are expected to provide further understanding.In addition, to push the commercialization of OSCs, it is crucial to reduce the material cost by simplifying the material synthesis, to control the molecular weight of the donor polymers, or search for polymers where device performance is resilient from MW, to process OSCs from environmentally friendly solvents, and to understand the material and device degradation mechanisms and loss processes of OSC modules.In a broader context, the points discussed in this Review may be used to provide a plan for advancing the optimization and enhancements of NFA-based organic solar cells.These points serve as a map, guiding us on how to approach further improvements in this technology.

Figure 1 .
Figure 1.Classic PBDB-T Series D-A Copolymer Donors with Outstanding Aggregation.a) The structure and the building blocks of PM6 and its parent-PBDB-T.b) UV−vis spectra of PBDB-T as film and in o-DCB solution for different temperatures (inset: the color of the solution for different temperatures.Reproduced with permission.[46]Copyright 2012, American Chemical Society.c) Diagram for the formation of fibrous aggregation, leading to an excellent domain purity in PBDB-T-based BHJ layers.Reproduced with permission.[50]Copyright 2019, Elsevier Ltd. d) GIXD profiles of neat PM6 and blend films of PM6:PC 71 BM (1:1.2, w/w), showing that the neat polymer film and blend exhibit face-on dominated molecular orientation with respect to the substrate.Reproduced with permission.[47]Copyright 2015, Wiley-VCH.

Figure 2 .
Figure 2. Single-junction organic solar cells with over 15% efficiency using PM6 and Y6.a) The building blocks mentioned in the main text.b) The development of Y6 and the A-DA′D-A molecular design concept: The Y6 molecule with a ladder-type electron-deficient core-based central fused backbone (DA′D), two electron-withdrawing end-groups (A), and two sp 2 -hybridized nitrogen atoms in the pyrrole motif (A-DA′D-A).c) J-V curves of the OSCs based on PM6:Y6 under illumination with AM1.5G, 100 mW cm −2 .d) EQE spectra of the corresponding OSCs.e) Absorption spectra of thin films of PM6 and Y6.Reproduced with permission.[19]Copyright 2019, Elsevier Inc.
) backbone unit was established.Subsequently, Zou et al. first introduced a DA'D structure where the BZTP fused-ring replaced the central D-fused-ring in the A-D-A structure to obtain the A-DA'D-A structured SMAs, which further reduced the bandgap of SMAs.

Figure 3 .
Figure 3.Some classic examples of high-performance A-DA'D-A type SMAs.

Figure 4 .
Figure 4. Molecular properties and packing of Y6: a) Side view of the optimized geometry of Y6 computed with -B97xD/6-31+G(d,p), indicating a clear twist in the backbone of Y6 Reproduced with permission.[19]Copyright 2019, Elsevier Inc. b) Calculated DOS for electrons (EA) and holes (IE) in a model crystal of Y6.Reproduced with permission.[106]Copyright 2020, Wiley-VCH.c) Isosurfaces of the electrostatic potential of Y6, together with the ellipsoid of the quadrupolar tensor.Reproduced with permission.[106]Copyright 2020, Wiley-VCH.Molecular pairs in the Y6 single crystal.d) Top and e) side views of the extended-crystal structure (the blue column is the stack of end groups in the b direction, the pink column is the stack of end groups in the c direction.Reproduced with permission.[109]Copyright 2020, Springer Nature.

Figure 6 .
Figure 6.Activationless free charge generation in PM6:Y6 devices: a) photogenerated free charge as function of bias voltage as measured by time delayed collection field (see inset).Bias has no effect on the charge generation efficiency.b) Internal quantum efficiency of photocurrent generation (IQE) and internal efficiency of free charge generation (IGE) as a function of photon energy, overlaid with the EQE spectrum.IQE and IGE are independent of photon energy even when exciting below the photovoltaic bandgap.c) Photogenerated charge as a function of bias and temperature for two photon energies.d)EQE spectra measured down to cryogenic temperatures.Except the very low-temperature range where transport issues become important, temperature has little effect on the EQE.[106]Copyright 2020, Wiley-VCH.e) Temperature dependence of the logarithm of the normalized internal quantum efficiency of PM6:Y6 (red diamond) compared to the blend of PM6 with the Y-series NFA BPT-eC9 (blue squares) and with the NFA ITIC (green circle).Also shown are the corresponding data for the blend of the non-fluorinated version of PM6, PBDB-T, with the NFA EH:IDTBR (yellow triangles).Solid lines show fits to a kinetic model which considers the competition between the splitting and the decay of the interfacial CT state.[130]Reproduced from Ref.[130] with permission from the Royal Society of Chemistry.

Figure 8 .
Figure 8. Energies of the charge-separated state: a) Energy levels of different donor polymers and NFAs from UPS, LE-IPES, and CV.Reproduced with permission.[150]Copyright 2022, Wiley-VCH.All measurements were performed on neat films.While the results from PES and CV agreed rather well for the NFAs, UPS consistently reveal a smaller ionization energy of the polymer layers, compared to CV.As a consequence, PES predicts a smaller energy of the charged separated state.b) HOMO and LUMO energies from spectroelectrochemistry of neat films and the PM6:Y6 blend, coated either from chloroform-chloronaphthalene or chlorobenzene.Reproduced with permission.[148]Copyright 2022, Royal Society of Chemistry.The results show little difference between the energy levels in neat and blend films but also a small effect of the used solvent and with that of the molecular orientation

Figure 9 .
Figure 9.Proposed mechanisms to explain activationless free charge generation in the PM6:Y6 blend.(a) A cascaded energy landscape is created by the lower-lying LUMO of Y6 molecules in ordered domains compared to the more disordered interface.Reproduced with permission.[124]Copyright 2022, The Royal Society of Chemistry.This drives electrons into the bulk of the Y6 domains.b) The quadrupole moment of Y6 molecules and dimers increases the ionization energy and electron affinity in the Y6 bulk relative to the interface.This creates a band bending towards the donor that increases the energy of the CT state relative to the CS, which counteracts the Coulomb attraction, and also suppresses recombination.Reproduced with permission.[155]Copyright 2023, Nature Publishing Group.Because the polymer carries a much smaller quadrupole moment, there is only little band bending in the donor phase.c) Due to the larger energetic disorder for free charges compared to Y6 excitons, charges can equilibrate at energies well below the mean energy of the photogenerated excitons, providing a driving force for exciton dissociation into free charges (graph derived from data in[141] ).d) Free charges are generated by efficient exciton dissociation in neat Y6 domains, while the role of the donor is mainly to collect the photogenerated holes and reduce non-geminate recombination.Reproduced with permission.[136]Copyright 2022 Springer Nature.

Figure 10 .
Figure 10.The fill factor of organic solar cells.a) FF versus the Figure ofMerit  for various BHJ blends, as reported in ref.[181].Reproduced with permission.[181]Copyright 2016, Springer Nature.Solid lines are analytical predictions of the FF- relation for V OC increasing from 0.7 to 0.9 V. Open circles are FF- points from simulated J-V curves with balanced mobilities and V OC between 0.7 and 0.9 V. b) Correlation of the relative short circuit current J SC /J G versus the effective drift length and c) fill-factor versus the effective diffusion length at the 1 Sun-equivalent illumination.Reproduced with permission.[176]Copyright 2021 Wiley-VCH.Systems 1-7 are PM6:Y6 blends of different thicknesses and preparation conditions.

Figure 11 .
Figure11.Non-germinate recombination in the presence of energetic disorder: a) Recombination coefficient k rec for a carrier density corresponding to solar illumination as a function of the combined disorder of free electrons and holes.b) Recombination coefficient as a function of the non-radiative voltage loss determined from the measured EQE EL .For the non-Langevin systems, a clear correlation between k rec and ΔV nr can be seen.No such correlation is apparent for the Langevin systems, where the assumption of quasi-equilibrium between free carriers and CT does not hold.[129]Copyright 2023, Wiley-VCH.

Figure 12 .
Figure 12.Excitons and CT states in absorption and emission: a) Normalized PL spectra of thin films of neat Y6 and blends of polystyrene (PS):Y6 and PM6:Y6 on glass, showing the red-shift of emission peak for neat Y6 and the blend PS:Y6 with respect to the PM6:Y6 film.b) Normalized PL and EL spectra measured on full devices.The subtraction EL -PL SC reveals a broad emission with a maximum at 1.15 eV which is interpreted as the emission from CT states (dark red line).c) Sensitive external quantum efficiency of PM6:Y6 at two different temperatures.d) Energy scheme summarizing the main findings from the data in panels (a-c).In PM6:Y6 devices, the chemical potential of the Y6 singlet exciton, μ S1 , is equal to the quasi-Fermi-level splitting in the bulk; thus, singlet excitons are in dynamic equilibrium with free carriers in the CS state and with the CT state population .Most of the photon emission of the excited blend originates from the Y6 exciton.However, most non-geminate recombination occurs through a very weakly emitting state, different from the Y6 singlet.We can relate the electroluminescence quantum efficiency (ELQY) of the singlet excitons in the device to the PLQY of the PS:Y6 film and conclude that <0.6% of injected charges are reformed into excitons.The low yield of reformation can be explained by the barrier between the singlet energy and the effective transport gap (CS state).Adapted with permission.[123]Copyright 2021 American Chemical Society.(e) State diagram of an organic solar cell with the low energy offset, indicating various transitions between the ground state singlet S 0 , singlet exciton S 1 , charge-transfer (CT), and charge-separated (CS) states: photon absorption under illumination (h), carrier injection under external bias (j inj ), exciton decay (k f ,S 1 ), exciton dissociation to CT (k d,S 1 ), CT decay (k f,CT ), CT dissociation into free carriers (k d,CT ), free carrier encounter to form CT (k rec ), and reformation of the singlet exciton (k ex,ref ).Reproduced with permission.[195]Copyright 2023, American Chemical Society.

Figure 13 .
Figure13.Dispersive or non-dispersive transport and recombination: Scheme of a) dispersive transport[203] and b) dispersive non-geminate recombination[205] in an organic layer with inhomogeneously broadened DOSs.Reprinted with permission.[203]Copyright 2017, Wiley-VCH.Reprinted (adapted) with permission.[205]Copyright 2019, American Chemical Society.The energetic relaxation of carriers within the DOS slows down carrier transport but also NRG. c) J-V characteristics of a 115 nm thick PM6:Y6 blend measured at different temperatures (lines).The symbols display the best fits with a kinetic Monte Carlo (kMC) code.d) Experimental V OC as a function of temperature from the data in (c) (solid squares).These data can be well reproduced with kMC simulations which include hot carrier effects (red lines and symbols) while drift-diffusion yields a too-small V OC (blue lines and symbols).[170]Copyright 2021 American Chemical Society.Electro-optical simulation of the fill factor (e) and the PCE (f) for PM6:Y6 devices as a function of active layer thickness (red lines).The blue line and symbols in Figure13fare for a blend of PM6 with the Y-series NFA BTP-eC9.The same set of input parameters (mobility, bandgap, NGR coefficient) was used for all layer thicknesses, indicating that charges equilibrated before being extracted or recombining.[130]

Figure 15 .
Figure 15.The morphology of PM6:Y6 BHJ blends casted from different solvents: a) 2D GIWAXS scattering profiles for pure films of PM6 and Y6 (left) and b) corresponding arrangement sketch map of molecules in films processed with CB and CF.Panels (a) and (b) adapted with permission.[98]Copyright 2020, Wiley-VCH.Solid-state 1D19 F NMR spectra of c-e) PM6, f-h) Y6, and i-k) PM6:Y6 BHJ films processed from CF, CB, and o-XY solvents.19F signals correspond to PM6 and Y6 moieties as indicated.Figures(c-k) adapted with permission.[206]Copyright 2022, Wiley-VCH.

Figure 17 .
Figure 17.Chemical structures of added components to PM6:Y6 blend and interlayers: Chemical structures of the third/fourth components serving as a) a donor or b) an acceptor in PM6:Y6-based ternary/quarternary OPVs with PCE > 17%.c) HTL and d) ETL that have been used for binary PM6:Y6 OPVs with PCE > 17%.

Figure 19 .
Figure 19.Molecular diffusion and thermal stability of PM6:Y6: (a) SIMS diffusion profiles of reference and annealed PM6 (top)/Y6 (bottom) bilayer.Zero of the abscissa represents nominally the vacuum/PM6 interface.(b) Temperature-dependent diffusion coefficient D(T) of different polymer:NFA systems (including PM6:Y6 system) fitted by an Arrhenius relation.The horizontal dashed box denotes the diffusion coefficient for an acceptor molecule to diffuse 20 nm on a time scale of 1 to 10 years.Reproduced with permission.[277]Copyright 2021, Springer Nature.
C signals corresponding to PM6 and Y6 molecules are indicated in colored dots in the NMR spectra (bottom) and their respective chemical structures (top).The 2D 1 H-13 C heteronuclear correlation (HETCOR) NMR spectra of the two PM6:Y6 BHJs show correlations between H and C atoms present in the blends (contours within the plots in Figure 20b,d, indicating the closeness between the PM6 sidechains and Y6