Exciton diffusion and dissociation in organic and quantum‐dot solar cells

For the process of photovoltaic conversion in organic solar cells (OSCs) and quantum‐dot solar cells (QDSCs), three of four steps are determined by exciton behavior, namely, exciton generation, exciton diffusion, and exciton dissociation. Therefore, it is of great importance to regulate exciton behavior in OSCs and QDSCs for achieving high power conversion efficiency. Due to the rapid development in materials and device fabrication, great progress has been made to manage the exciton behavior to achieve prolonged exciton diffusion length and improved exciton dissociation in recent years. In this review, we first introduce the parameters that affect exciton behavior, followed by the methods to measure exciton diffusion length. Then, we provide an overview of the recent advances with regard to exciton behavior investigation in OSCs and QDSCs, including exciton lifetime, exciton diffusion coefficient, and exciton dissociation. Finally, we propose future directions in deepening the understanding of exciton behavior and boosting the performance of OSCs and QDSCs.


| INTRODUCTION
Most energies we used, including fossil fuels, wind power, hydraulic energy, biomass, and so forth, are essentially derived from the sun.2][3][4][5][6] To realize photovoltaic conversion, it must go through the process that photosensitive materials in solar cells absorb sunlight via photon-electron coupling and are excited to form electron-hole pairs, namely, excitons.Excitons can be divided into the Wannier exciton and Frenkel exciton, according to the radius of excitons and the Coulombic binding energy between the excited electron and hole. 7he Wannier exciton has a binding energy, comparable to the thermal energy, and generally forms in inorganic semiconductors, such as 3.27 meV for GaAs, 20.4 meV for GaN, and 2-55 meV for perovskite (MAPbI 3 ). 8,9Those materials would produce free charge carriers upon absorbing photons since the thermal vibration can break bound excitons. 10It has been investigated thoroughly and will not be further discussed in the text. 8,9In contrast, the Frenkel exciton has a binding energy of hundred meV and generally exists in organic semiconductors and quantum dots (QDs). 7,11,12Extra driving force is needed to overcome the binding energy and compel the charge dissociation of Frenkel excitons. 13oreover, excitons must diffuse to special sites, for example, donor/acceptor interfaces, to realize charge dissociation before geminate recombination (the lifetime of excitons is generally 10 −9 -10 −7 s). 14Thus, the photoelectric conversion in Frenkel-exciton solar cells, such as organic solar cells (OSCs) and quantum-dot solar cells (QDSCs), can be divided into four key steps, each with its own efficiency: (1) photon absorption and exciton generation, (2) exciton diffusion, (3) exciton dissociation, and (4) charge carrier transport and collection. 15The product of the four efficiencies is the external quantum efficiency of solar cells, which determines the shortcircuit current density (J SC ), one of the three critical parameters defining the power conversion efficiencies (PCEs).Therefore, regulating exciton behavior is of great importance to achieve high PCEs for OSCs and QDSCs.
Due to the high extinction coefficient and excellent tunability of optical bandgap for organic semiconductors and QDs, OSCs and QDSCs can harvest sunlight and generate excitons efficiently.Long exciton diffusion length (L D ) of semiconductors, which exceeds 100 nm [16][17][18] and even reaches the magnitude of micrometer, 19,20 has been reported.However, most of the highperformance semiconductors for OSCs and QDSCs still suffer from relatively low L D (generally 10-20 nm), which is much shorter than the optical absorption length (~50 nm). 213][24] In addition, great progress has also been made in facilitating the exciton dissociation in OSCs and QDSCs.For example, the precondition, that the driving force for efficient exciton dissociation should be larger than 0.3 eV, has been broken up in OSCs as the emergence of high-performance nonfullerene acceptors. 25,26Therefore, it is necessary to retrospect and summarize the achievements in the understanding of exciton behavior for organic and QD semiconductor materials, which would pave the way for the further development of OSCs and QDSCs.
In this review, we first introduce the parameters those impact exciton behavior, followed by a summary of the methods, which are employed to measure exciton diffusion length.Then, we provide a critical overview of the recent advances in understanding the exciton behavior in OSCs and QDSCs, including exciton diffusion coefficient, exciton lifetime, and exciton dissociation.Finally, future directions are briefly outlined, focusing on exciton behavior manipulation and aiming to further boost the PCEs of OSCs and QDSCs.

| THEORY OF EXCITON DIFFUSION IN OSCs AND QDSCs
The formation of photon-generated excitons is a very fast process at the magnitude of femtosecond via photon-electron coupling.It is deemed that most excitons initially possess high energy and delocalize over several molecules, namely, hot excitons.However, the hot exciton undergoes a rapid relaxation on a sub-50 fs timescale and cools down to form either a localized exciton or polaron pair via intermolecular motion. 27The relaxation process for excitons loses a considerable amount of energy and leads to a fast increment in reorganization energy, which makes the diffusion rate of the localized exciton considerably lower. 28Despite the tight bound between the electron and hole, excitons are generated near-unity quantum efficiency in OSCs and QDSCs.Therefore, exciton generation will not be discussed further here.
There are mainly two types of energy transfer for exciton diffusion in OSCs and QDSCs, namely, Dexter transfer 29 and Förster transfer. 30Since it involves a direct exchange of electrons for Dexter transfer, the electronic overlap is required between the excited donor molecule and the nearby acceptor molecule, which occurs at a typical length scale of 1-10 Å.The rate of Dexter energy transfer (k D ) can be described as 31 where K is related to the specific orbital interaction; J is the spectral overlap integral normalized for the extinction coefficient of the ground state molecule; d is the spacing between donors and acceptors; and L is the van der Waals radius.It should be noted that the acceptor molecule here refers to a neighboring molecule, rather than the electron acceptor in OSCs.In addition, the rate of Dexter transfer is independent of the acceptor molecules' extinction coefficient.It can be found that the Dexter energy transfer decreases exponentially along with the increment of the spacing between donors and acceptors, which implies that the molecular packing has a significant impact on the exciton diffusion.Thus, as the spacing reaches over one or two molecular diameters, the rate of Dexter transfer can be negligible and Förster transfer dominates the process via weak dipole-dipole interaction.The rate of Förster transfer (k F ) can be written as follows 30 : where τ 0 is the intrinsic exciton lifetime of chromophores; R 0 is the Förster radius for energy transfer; and d is the spacing between donors and acceptors.It should be noted that x = 6 is for energy transfer between two point dipoles, 29 x = 4 is for energy transfer to a monolayer acceptor, 32 and x = 3 is for energy transfer to thicker quenching layers. 33Moreover, R 0 is defined as 34 where η PL is the photoluminescence (PL) efficiency of the excited state; κ is the dipole orientation factor; m is the refractive index; λ is the wavelength; F D is the normalized donor fluorescence spectrum; and σ A is the absorption cross-section of the acceptor.The integral part in Equation (3) represents the spectral overlap between the emission of donor molecules and the absorption of acceptor molecules.Thus, reducing Stokes shift to improve the spectral overlap and enhancing the PL quantum yield of materials can efficiently increase the rate of Förster transfer. 35Due to polymorph in the organic and QD films, it cannot exactly describe the exciton diffusion in one model.To simplify the exciton diffusion process, it is depicted as an ensemble of nearest-neighbor hopping events identical to a random walk. 36,37Therefore, the exciton diffusion length (L D ) can be defined as below: where Z is the dimensionality of exciton diffusion (Z = 1, 2, or 3 corresponds to one, two, and three dimensions, respectively); 35 D is the exciton diffusion coefficient; and τ f is the PL life in the solid film.However, the factor Z is often omitted.D can be described as 38 where A is the factor of disorder in the film; and k ET is the energy−transfer rate for a specified lattice point of set N. It has been reported that reducing the distribution of excitonic density of states can increase A to lengthen L D .In QDSCs, the diffusion of excitons with large binding energy mainly follows Förster energy transfer and L D can also be calculated as the square root of diffusion coefficient D and effective time constant τ eff 39,40 : where the slope S represents the PL ratio with acceptor concentration; l is the interdot distance; σ is the capture cross-section; τ bandedge is the lifetime of bandedge PL; and N t is the trap density.As discussed above, we can manipulate the relevant physical parameters to prolong the L D in OSCs and QDSCs through material design or device engineering.

| MEASUREMENTS OF EXCITON DIFFUSION IN OSCs AND QDSCs
The L D of photovoltaic materials is a key parameter that impacts the device performance.Many methods have been developed to determine L D .They can be divided into two categories: spectroscopic techniques (Figure 1A,B) and charge carrier techniques (Figure 1C,D).

| PL spectroscopic techniques
][49] The PL of the thin film is determined in the presence of a quenching site, either donor/acceptor interfaces or quenchers in the bulk, within these methods.The quencher should have enough energy level offset with respect to the sample to ensure that it is an exothermic process for charge transfer and obstruct long-range Förster energy transfer to simplify the L D fitting procedure.Generally, the corresponding donor/ acceptor material in solar cells is adopted.1][52] The PL ratio is constructed by comparing the film having quenched PL with the one that is not being quenched.With the data, D can be obtained via iteratively fitting to the following equation 35 : where n is the exciton density; ∇ 2 is the Laplace operator; r is the position at point n; G is the exciton generation rate; τ is the exciton lifetime; and γ is the annihilation rate constant.It should be noted that the term, γn(r n ) 2 , is taken into consideration when the exciton-exciton annihilation happens with relatively high exciton density, for example, exciton-exciton annihilation method.
In addition, γ is equal to 4πR a D, where R a is the annihilation radius, namely, the average distance between two excitons that undergo annihilation (R a = 1 nm is widely used). 53Different spectroscopic techniques have their advantages as well as disadvantages.The materials having relatively strong PL are required for spectroscopic techniques.For the steady-state PL surface quenching technique, it can be carried out at a steady state without ultrafast-detection testing equipment.However, the sharp and highly quenching interface is required, a statistically large number of samples and film thicknesses are needed for the measurement, and the data are iteratively fitted to Equation ( 8) and obtain L D .
For time-resolved PL surface quenching and timeresolved PL bulk quenching techniques, the number of samples would be significantly reduced to gain a confident measurement of L D .However, the fast detection at the timescale of a picosecond is indispensable for testing equipment.For exciton-exciton annihilation, only one pristine film is required to determine the PL decay under various excitation intensities.However, the annihilation cross-section of excitons in the film is hard to determine. 14,35The difference and applicability of spectroscopic techniques to measure L D will be not discussed here since it has been well summarized in the previous review. 35

| Charge carrier techniques
The charge carrier technique is one other kind of method to measure L D via measuring photogenerated charge carrier originating from quenched excitons, including device modeling, 54 time-resolved microwave conductivity, [55][56][57] photocurrent-ratio method, 58 and surface photoconductivity. 59n the device modeling technique, it is assumed in devices that all excitons dissociate into free charge carriers and are collected at electrodes. 54This method has the advantages of simple sample preparation and easy experimental measurements.The external quantum efficiency of planar heterojunction devices is determined and fitted by using the optical interference model and Equation (8).The L D can be calculated by using boundary conditions.However, the assumptions of exciton dissociation and charge collection would lead to an underestimation of L D since the trapped and recombined photogenerated charge carriers are treated as undissociated excitons.Time-resolved microwave conductivity can also measure L D without charge carriers collected at counter electrodes. 55This electrodeless technique measures the time-dependent change in reflected power from a microwave cavity as a result of a change in the conductivity of the quenching film.By using the boundary conditions of Equation ( 8), the functional relationships between the fraction of photons entering the film from frontside and backside and the L D of the film can be constructed.Plotting the four functions, the L D of the film can be obtained.The advantage of this technique is not susceptible to charge carrier collection loss.However, the conductivity of the quenching layer should be considerably higher than that of the measured material so that the differential microwave power reflected from the film results from the change in the conductivity of the photogenerated charge carriers presented in the quenching layer.The determined data have to be fitted to complicated modeling with many fitting parameters.In addition, exciton quenching at the surfaces would impact the value of L D .Zhang et al. 58 applied the photocurrent-ratio method and determined the relationship between the internal quantum efficiencies of the donor/acceptor (η IQE ) and the thickness of the donor/acceptor under the specific wavelength in bilayer devices.The η IQE ratio between the donor and acceptor was applied to cancel the unknown effect of charge separation.Combined with the η IQE ratio, the L D of the donor and acceptor can be both extracted via treatment with a one-dimensional (2D) steady-state diffusion model (Figure 1D).The advantage of this technique is that it can measure the intrinsic L D of semiconductors with/without emission.In addition, this method eliminates the impact of exciton dissociation.The surface photoconductivity technique is generally used to measure the L D of single crystals. 59The surface photoconductivity is proportional to the fraction of photogenerated excitons, which can reach the surface and dissociate into charge carriers.However, a large surface photoconductivity for measured samples is required.In addition, the sample with shorter L D will show a larger differential in surface photoconductivity, while samples with larger L D can overcome the issue of penetration depths and exhibit a smaller differential in surface photoconductivity.
So far, a series of techniques have been developed to measure L D of semiconductors.Each method has its own advantages as well as drawbacks.Provided that the technique is self-consistent, the trend of L D obtained from each method can still afford instructive feedback.However, it is essential to unify the technique and acquire the exact value of L D for semiconductors, which would provide more reliable guidance for material design and device fabrication.

| EXCITON DIFFUSION IN OSCs AND QDSCs
As mentioned above, the L D of semiconductors in OSCs and QDSCs is much shorter than the optical absorption length, which limits excitons to migrate to dissociation sites and thus the efficiency of photovoltaic conversion.Numerous studies were conducted to lengthen the L D of semiconductors in devices and boosted excitons diffusing to dissociation interfaces.According to Equation ( 4), we will describe specific studies about L D from the exciton diffusion coefficient and its lifetime herein (Table 1).

| Exciton diffusion coefficient
According to Equation ( 5), the exciton diffusion coefficient is proportional to the factor of disorder in the film.Lowering energetic disorder and preventing any structural relaxation in the excited state will bring the width of the density of states closer in energy and provide more pathways to benefit exciton transport.Akselrod et al. 60 investigated exciton diffusion in the single crystal and disorder thin films of tetracene via a spatial, temporal, and spectral visualization method.The spatial intensity distribution at each point in time has been normalized to a constant maximum value to emphasize the evolution in the width of the emission distribution (Figure 2A).The initial near-Gaussian intensity distribution has a standard deviation of σ Ι = 229 nm, which broadens rapidly within the first 2 μs, followed by a subsequent slowing down of the expansion and reaching σ I = 701 nm at 7 μs (Figure 2B).It indicated that the exciton transport would transform into a subdiffusion and the exciton diffusion coefficient significantly decreased, as the exciton was trapped (Figure 2C).They further finely controlled the crystallinity of tetracene films and pointed out that this transition to subdiffusive transport occurred at earlier times as the disorder increased.Thus, tetracene films with lower crystallinity suffer inferior exciton diffusion coefficients.Likewise, Lunt et al. 61 increased the crystalline domain sizes of 3,4,9,10-perylenetetracarboxylic dianhydride from 100 to 400 nm and found that the L D was enhanced from ~6 to ~22 nm, due to the reduced disorder and higher PL quantum yield.3][64][65][66] Samuel et al. 62 treated the PffBT4T-2OD:PC 71 BM blend film with thermal annealing at 100 °C for 5 min.The size of the crystalline domains of PffBT4T-2OD (Figure 3) increases from 13.2 to 17.8 nm.Thus, the exciton diffusion coefficient increases by twice,  3). 65herefore, the exciton diffusion coefficient of DR3TBDTT increases by threefold and the PCE of OSCs achieves a 20% enhancement.The ternary strategy was also used to improve the crystallinity in the blend film for enhancing the L D and device performance. 67,68Recently, Cai et al. 67 blended two high-performance nonfullerene acceptors, BTP-eC9 and L8-BO-F (Figure 3), and the blend film afforded improved crystalline coherence lengths (CCLs) of π-π and lamellar stacking compared with the pristine ones.Thus, the L8-BO-F:BTP-eC9 thin film achieves a larger exciton diffusion coefficient of 7.28 × 10 −2 cm 2 /s and thus prolonged L D of 47 nm.The enhanced L D for superior exciton diffusion and dissociation boosts PM6:L8-BO-F:BTP-eC9 ternary solar cells having a layer-by-layer structure to achieve higher PCEs of 18.53% with a film thickness of 120 nm and 15.21% with a film thickness of 500 nm.Since the crystallinity of materials is strongly related to side-chain length, the L D can be optimized via side-chain engineering. 69,70Bi et al. 69 developed three nonfullerene acceptors with different side chains.GS-ISO (Figure 3) has the shortest side chains and exhibits the strongest crystallinity, contributing to the smallest energetic disorder and the largest exciton diffusion coefficient of 7.60 × 10 −3 cm 2 /s.Thus, GS-ISO-based OSCs acquired the best PCE with the highest J SC .In addition, the homocoupling defects and low molecular mass for conjugated polymers are also detrimental to exciton diffusion, leading to the inferior exciton diffusion coefficient. 71I G U R E 3 Chemical structures of materials for OSCs are discussed in the text.OSC, organic solar cell.
Förster energy transfer is the main type of exciton diffusion in organic semiconductor films.4][75] Jo et al. 73 fabricated the sandwich devices with pentacene standing-up or lying-down orientation relative to the substrate.They found that the L D of pentacene with lying-down orientation achieved 83 nm and was almost as twice that with standing-up orientation (43 nm).It implies that the orientation of π−π stacking is preponderant for exciton diffusion, which is in line with the charge carrier behavior.It can be also found in Equation ( 3) that improving the PL quantum yield would increase the Förster radius for higher exciton diffusion coefficients.Mullenbach et al. 76 developed several rubrene derivatives, including f-rubrene, mm-rubrene, and fmrubrene.According to the crystallographic parameters, the intermolecular distance increases from rubrene, f-rubrene, mm-rubrene to fm-rubrene, successively.The material with larger intermolecular separation affords higher PL quantum yield by inhibiting the nonradiative decay process.Thus fm-rubrene thin film achieves the largest L D .Zhu et al. 77 removed the thiadiazole unit in Y6 and developed F1.The F1 film exhibits a higher PL quantum yield of 9.3% than the Y6 film (5.6%), contributing to a larger Förster radius of 3.3 nm (Figure 4).Thus, the F1 film presents a longer L D of 20 nm, which is nearly as twice that of the Y6 film (12 nm).In addition, Menke et al. 78 diluted the electron donor, subphthalocyanine chloride (SubPc) into a wide-energy-gap host material (UGH2) to optimize the intermolecular separation.As the ratio of SubPc decreases in the blend, the PL quantum yield is enhanced, due to the reduction in vibronic coupling between molecules for inhibited nonradiative decay, while the relative spectral overlap between absorption and emission of SubPc increases, owing to the reduced SubPc dimers and solid-state solvation effect, which improves the Förster radius from 1.0 ± 0.1 nm in the neat film to 3.8 ± 0.4 nm for the film containing 1 wt% SubPc in UGH2 and thus the L D from 10.7 to 15.4 nm.The strategy that enlarges the spectral overlap between the absorption of the acceptor and the emission of the donor is also used to increase the Förster radius for prolonged L D and thus improved energy transfer. 79,80Cnops et al. 79 inserted the SubNc (Figure 3) layer into the bilayer device based on SubPc and α-6T.Due to the large spectral overlap between the absorption of SubNc and the emission of SubPc, the Förster radius of SubPc increases from 1.5 to 7.5 nm.As a result, excitons can transfer from SubPc to SubNc efficiently and the three-layer device based on α-6T/SubNc/SubPc achieves a much higher PCE of 8.4% with enhanced external quantum efficiency, compared with the bilayer one.Recently, it has been reported that fused ring electron acceptors, such as IDIC, ITIC, IT-4F, and Y6 (Figure 3), possess the superiority of long L D in the range of 20-50 nm with high exciton diffusion coefficients, due to the low reorganization energy, high chromophore density, and low disorder (Figure 2D). 53,72,81This superiority for fused ring electron acceptors rationalizes their outstanding photovoltaic performance, especially, the excellent PCEs in layer-by-layer OSCs. 82So far, efficient strategies have been put forward to increase exciton diffusion coefficients based on the exciton transport model, and more efforts should be made to boost the advancement for prolonged L D .
F I G U R E 4 Compared with Y6, F1 without thiadiazole exhibits a higher photoluminescence quantum yield (PLQY) and longer L D .Longer L D ensures that more excitons within semiconductors can reach the interface and dissociate to charge carriers before recombination.The blue solid spheres represent electrons and the orange solid spheres represent holes.Solid blue and orange spheres surrounded by dashed lines represent excitons.The yellow lightning shapes represent exciton recombination.The thin green arrows represent diffusion, and the thick green arrows represent Förster's resonance energy transfer (FRET).Reproduced with permission: Copyright 2022, American Chemical Society. 77n QDSCs, the L D , which relies on the coupling transport, is a crucial parameter, and devices having excellent L D usually possess superior PCEs.The characteristics of exciton transport can be tuned by precisely controlling the interior crystal and surface structures in QD particles, among which the strong electrical coupling between the particles endows them with the capacity to transport efficiently. 83,84Akselrod et al. 85 adjusted the thickness of the inorganic shell and organic ligand length in CdSe/ZnCdS core/shell QD assemblies to explore the exciton transport process.They found that exciton diffusion in QD assemblies did not occur by a random process, but rather energetic disorder within the inhomogeneously broadened films, which led to the reduction of diffusivity over time and that diffusion was faster for excitons in the higher-energy portion of the inhomogeneous distribution than those near the bottom of the distribution.Yoon et al. 24 showed that localized exciton transport was supported in CdSe QD films and closer packing of QDs enabled enhanced the exciton diffusion coefficient of 2.5 × 10 −2 cm 2 /s in highly ordered superlattices, which was in good agreement with the Förster resonance energy transfer theory, while disordered and structural heterogeneity would cause the exciton diffusion coefficient to decrease (Figure 5).Furthermore, Penzo et al. 86 demonstrated that the 2D close-packed assembly of CsPbBr 3 nanocrystals possessed a superior exciton diffusion coefficient of 0.5 cm 2 /s and higher L D of 200 nm.In contrast, the exciton diffusion coefficient for chalcogen-based QDs was about 0.2 × 10 −3 -12 × 10 −3 cm 2 /s.In a word, perovskite QDs with higher exciton diffusion coefficients have several advantages, such as smaller interparticle distance, higher spectral overlap, well-aligned transition dipoles, higher photon absorption cross-section, near-unity quantum yield, and flatter energy landscape.
Recently, Akselrod et al. 60 and Zhang et al. 87 revealed the transport mechanism of excitons at the very beginning of their generation in a QD film, providing us a clearer view of how to achieve fast exciton transport: (1) there is an extremely fast transport process in the initial few hundred femtoseconds of exciton generation with a diffusion coefficient up to 10 2 cm 2 /s, which is nearly three orders of magnitude faster than the traditional hopping transport, and then, it will transform to the slower transport regimes; (2) the diffusivity of such ultrafast excitons slows down as the distance between QDs becomes shorter, which is inverse to the traditional hopping mechanism that shorter interdot distance brings about faster exciton transport; and (3) the process only occurs in QDs whose Bohr radius is much larger than the particle radius, such as PbS and PbSe, while the phenomenon has not been observed in materials whose Bohr radius is similar to their particle size, such as CdS, CdSe, and CsPbI 3 .Therefore, the regulation of the fast/ slow exciton transport regimes and their transition process can be realized by optimizing the packing density and heterogeneity of QDs.On the other hand, trap states on the surface of QDs hinder carrier transport.The Copyright 2016, American Chemical Society. 24ommon method to alleviate the problem is surface modification. 88,89Choi et al. 90 obtained PbS QDs with p-type and n-type transport properties by cascade surface modification and then mixed two-type QDs in a certain proportion to obtain the active layer with intrinsic homojunction.As a result, the L D increases by 1.5 times, compared with the reported planar QD films, to achieve the improved J SC of 30.2 mA/cm 2 and thus a higher PCE of 13.3% in QDSC.Moreover, the length of ligands and shell thickness also affect exciton transport and thus L D .
Currently, the exciton diffusion coefficient in OSCs is in the range of 10 −4 -10 −2 cm 2 /s, 72 while that in QDSCs is located at 10 −3 -1 cm 2 /s, 87,91 since excitons diffuse very fast in the crystal interior of QDs and exciton diffusion is also relatively fast as the reduction of QD distance (Table 1).

| Exciton lifetime
Exciton lifetime is the other parameter to determine L D .The exciton lifetime depends on the radiative and nonradiative photophysical processes and can be expressed as below 92 : where k R is the rate of radiative decay and k NR is the rate of nonradiative recombination decay.For organic semiconductors, the rate of nonradiative recombination decay dominates the exciton lifetime since k NR is generally significantly larger than k R .Nonradiative transitions in molecules include intersystem crossing, internal conversion, electron transfer, and electron-hole recombination.
The rate constants of these processes highly depend on the energy gap between the initial and final states involved, according to the Marcus theory and energy-gap law. 93However, the strategies to control the exciton lifetime are limited and vague yet currently from either molecular design or device processing. 92Triplet excitons have a much longer lifetime than singlet ones, because of their quantum-mechanically forbidden relaxation to the ground state.5][96][97][98] Du et al. 95 added a delayed fluorescence emitter, APDC-TPDA into PM6:Y6 and PTB7-Th:IEICO-4F pairs (Figure 3).APDC-TPDA efficiently prolongs the exciton lifetime of PM6 and PTB7-Th.They deemed that it would increase L D and give more time for exciton dissociation, thus contributing to higher J SC .
Kushto et al. 98 prepared several phthalocyanine derivatives with different metal cores.PdPc with the heavier atom shows a longer exciton lifetime due to the spin-orbit coupling and thus a longer L D of 10.1 nm, compared with CuPc, ZnPc, and so forth.As a result, it is contributing to the higher PCEs of OSCs.However, triplet excitons suffer from the following issues: (1) high binding energy requires a large driving force for charge dissociation; (2) a lower energy level of triplet than singlet would bring about larger energy loss; and (3) triplet excitons diffuse mainly via Dexter energy transfer, which decreases exponentially with the distance. 27ecently, Guo et al. 99 introduced a solid additive, FCA, into PBDB-T:IT-M blend (Figure 3) and increased the exciton lifetime of IT-M from 491 to 928 ps.They measured the steady-state (Figure 6A) and transient (Figure 6B Exciton lifetime is essentially dominated by the electronic structure and aggregation of organic semiconductors and strongly depends on their chemical structures.Umeyama et al. 100 developed three nonfullerene acceptors, TACIC-EH, TACIC-BO, and TACIC-HD, with the heteronanographene central core furnished with different-length side chains (Figure 3).Exciton lifetimes of the thin films increase from TACIC-EH (1.33 ns), TACIC-BO (1.59 ns) to TACIC-HD (2.33 ns), due to the reduction of the rate of nonradiative decay.Likewise, Kaushal et al. 101 synthesized a series of porphyrin derivatives with different side chains.Exciton lifetimes of the porphyrin derivatives also increase along with side-chain length.The probable reason for the improvement in exciton lifetime is the dilution effect of side chains on chromophore fragments of semiconductors.In contrast, Wen et al. 102 compared exciton lifetimes of Y6 and its derivatives with different side chains and found that the influence of alkyl chains, attached to the pyrrole ring, was weak on the exciton lifetimes of Y6 and its derivatives.They further pointed out that the chemical structure of end groups had a significant impact on the exciton lifetime.The thin films of Y5 and Y10 (Figure 3) with nonhalogenated electron-deficient terminals afford the longest exciton lifetime up to ~1100 ps.In addition, Dimitrov et al. 92 declare that increasing the crystallinity of conjugated polymers in films leads to the reduction in exciton lifetime.Some experiences have been summarized to manipulate the exciton lifetime of organic semiconductors, but the intrinsic factors to prolong exciton lifetime via material design are unclear yet so far.
4][105][106] When the size of QDs is smaller than its Bohr radius, the conduction band shifts down to the bottom, while the valence band moves to the top along with increasing the QD size, thus leading to the reduction in the optical bandgap.According to the bandgap law, the narrower bandgap would bring about the greater transition probability and thus shorter exciton lifetime.Ushakova et al. 103 finely manipulated the size of PbS QDs via controlling the synthetic time.They found that, when the size of PbS QDs increased from 2.5 to 8.8 nm, the exciton lifetime significantly decreased from ~2.5 to 0.25 μs (Figure 7A).Thus, it leads to a delicate balance between absorption and exciton diffusion for QDs.Ma et al. 106 finely controlled the size of PbSe QDs (1-3 nm) and balanced the light absorption, exciton transport, and carrier collection in ITO/PEDOT/PbSe/Al solar cells.The QDSCs based on PbSe QDs that had a diameter of ~2.3 nm and a bandgap of ~1.6 eV achieved the highest PCE of 4.57%.In addition, highly reactive dangling bonds in QDs, due to the huge specific surface area, result in a large number of defects, which has a great impact on exciton lifetime. 1079][110] Due to the unique staggered band alignment between the core and shell, the so-called type II band alignment, the carrier excitation holes and electrons are spatially separated at the core and shell, and thus, the exciton lifetime is strongly dependent on the thickness of the shell.Xu et al. 108 found that the exciton lifetime of core-shell ZnO-CdS increased along with the increment of CdS layer thickness.Wang et al. 110 synthesized the colloidal CsPbI 3 / PbSe QDs, and they found that CsPbI 3 /PbSe QDs exhibited improved absorption and lower trap density and prolonged the exciton lifetime from 33.4 to 42.6 ns, as compared with the bare CsPbI 3 QDs.As a result, the device based on CsPbI 3 /PbSe QDs acquired a high PCE of 13.9%.Organic ligands, such as oleic acid, glutathione, quaternary ammonium salt, sulfhydryl compound, alkylphosphonic acid, and alkyl sulfonic acid, were also used to passivate dangling bonds on QDs and thus to prolong the exciton lifetime. 88In addition, Chen et al. 111 investigated the impact of alkyl-chain length of organic ligands on the exciton lifetime of CsPbI 3 QDs.They found that replacing oleic acid partially with  99 octanoic acid or octylamine on CsPbI 3 QDs led to reduced concentration of surface defects and thus improved the exciton lifetime from 10.06 to 20.66 ns.Moreover, the shorter alkyl chains would reduce the distance of QDs and benefit charge carrier transport.As a result, C8/C18-CsPbI 3 QDbased solar cells obtained a higher PCE of 11.87%, compared with the C18-CsPbI 3 QD-based one (7.76%).Due to the bulkiness, vulnerability to oxidation, and thermal degradation of organic ligands, Tang et al. 112 developed the atomic ligand strategy that used monovalent halide anions to successfully passivate surface defects of PbS QDs and led to shallower trap state distribution than organic ligands (Figure 7B).The device based on halogenated PbS QDs exhibited an improved PCE of 6%.Likewise, Elibol 113 treated CdSeTe alloy QDs with CdCl 2 solution and improved its exciton lifetime from 28 to 33 ns.To date, many strategies have been reported to prolong the exciton lifetime of QDs, but a clear guideline for the rational design or selection of ligands to achieve high-efficiency QDSCs is still lacking.So far, the exciton lifetime for OSCs is in the range of 10 −9 -10 −7 s, 15 while that for QDSCs is 10 −8 -10 −6 s (Table 1). 114

| EXCITON DISSOCIATION IN OSCs AND QDSCs
Exciton dissociation would take place and convert into charge-transfer state, thus free charge carrier, after excitons diffuse to special sites, such as donor/acceptor interfaces.Since the low dielectric constant (2-4) and small electron-hole distance cause strong Coulombic force at the magnitude of hundred meV for Frenkel excitons, extra energy, called exciton binding energy (E b ), is demanded to drive excitons to dissociate into free charge carriers, which would lead to large energy loss and severe geminate recombination. 115,116Specifically, E b can be approximatively described as 117 where e is the elementary charge; ε 0 is the vacuum dielectric constant; ε r is the permittivity; and R is the average electron-hole distance.Increasing ε r of materials would reduce E b , while the distance R is strongly related to the electronic structure of materials.It should be noted that this method for E b calculation is generally used in inorganic semiconductors.Due to the disorder of organic materials in thin films, it is not suitable for most organic semiconductors.
In OSCs, it has been widely accepted and investigated previously that, in fullerene-based devices, the driving force should be larger than 0.3 eV to achieve fast and efficient exciton dissociation, which will not be further discussed here. 1180][121] Han et al. 122 conducted a series of theoretical studies and explained why efficient exciton dissociation could be achieved in such low-driving-force nonfullerene-acceptor-based OSCs from the molecular structure points of view: (1) the state-of-the-art nonfullerene acceptors having A-D-A structure were found to dock with donors mainly via local π-π interaction, and such interfacial geometries could provide strong electronic couplings to enable fast exciton dissociation.(2) The strong polarization of charge carriers for A-D-A acceptors remarkably reduced E b and the enhanced polarization energies of holes and electrons during charge separation led to reduced energy barrier, even barrierless charge separation (Figure 8A-H). 123(3) The intermolecular π-π stacking of end-groups reduced the singlet-triplet energy difference of A-D-A acceptors, which resulted in significantly decreased driving force and suppressed charge recombination via the triplet channel. 124In addition, Wu et al. 125 investigated the charge separation in the PM6:Y6 pair and pointed out that, due to the low energetic disorder, electrostatic interfacial fields, and low electronic energy offset in PM6:Y6 OSCs, entropy increase was the major energetic element for steady charge separation.Zhang et al. 126 evaluated the magnetic field-dependent photocurrent density and PL and proposed that the spin-dependent polaron pair dissociation at charge-transfer states proved critical for the photocurrent production, rather than the driving force from the energy offset.
Recently, considerable studies about exciton dissociation were focused on reducing E b and exploring the limitation for efficient charge separation in nonfullerene acceptor-based OSCs.By investigating E b of 14 typical nonfullerene acceptors, Zhu et al. 128 found that the driving forces for the dissociation of acceptor excitons into charge-transfer states were linearly correlated to the E b , and the smaller the E b , the lower the driving force was required.Gao et al. 129 combined the molecular asymmetry strategy with more polarizable selenium substitution to increase the dielectric constant of nonfullerene acceptors.The asymmetry BS3TSe-4F (Figure 3) exhibits a lower E b than the symmetric counterpart, which benefits exciton dissociation and suppresses charge recombination to contribute to the J SC improvement.Thus, D18:BS3TSe-4F-based OSCs achieved a higher PCE of 18.48%.Introducing highly polarizable substituent groups onto semiconductor molecules to improve the dielectric constant is another strategy to reduce E b . 127,130,131Li et al. 127 developed Y6-4O (Figure 3) via attaching a highly polarizable oligo-(ethylene glycol) side chain onto the pyrrole unit of Y6 and increased the dielectric constants from 3.36 to 5.13, thus reducing the E b from 0.30 eV (Y6) to 0.24 eV (Y6-4O) (Figure 8I).As a result, PM6:Y6-4O-based OSCs achieved a much higher PCE of 15.2% with higher charge separation yield, slower bimolecular recombination kinetics, and less voltage loss, as compared to PM6:Y6based ones.Furthermore, scientists also tried to find the limitation of the driving force for efficient exciton dissociation in nonfullerene acceptor-based OSCs.Zhong et al. 119 and Li et al. 120 systematically studied different donor/nonfullerene acceptor pairs and pointed out that efficient charge transfer sustained at timescales of a few hundred femtoseconds even at near-zero driving force.Zhang et al. 121 further used two nonfullerene acceptors with similar chemical structures to form molecular alloys and thus finely controlled the energy level of the acceptor in the blend film.They found that the minimum HOMO offset of ~40 meV was required to maintain the mostefficient exciton dissociation and photovoltaic performance.
In QDSCs, the increased E b due to the reduced dielectric constant, compared with their bulk counterparts, has a negative impact on the exciton dissociation, leading to the lower screening of electron-hole interactions in nanostructures.Efficient exciton dissociation usually occurs at the interface by means of electron transfer or within the QD film through the interdot coupling electric field.3][134] McDonald et al. 135 investigated the mixed system of QDs and organic semiconductors and found that the built-in electric field was formed in the mixed film and excitons could be separated at the boundary between QDs and the polymer under the driving force of conduction band energy level difference.To overcome the relatively low separation efficiency of photogenerated excitons in the Schottky structure, Leschkies et al. 136 introduced the p-n structure in heterojunction, and the exciton dissociation process occurred at the heterojunction interfaces.The depletion width in the active layer was expanded with the appearance of wide-bandgap metal oxides (such as TiO 2 and ZnO), and the high-density exciton was close to the heterojunction interface as photons entering from the transparent n-oxide film, which enabled efficient electron transport from QDs to metal oxides.Therefore, increasing the depletion width by QD structure optimization can effectively improve the efficiency of exciton dissociation. 137Furthermore, Zhao et al. 138 demonstrated that a structure of the ladder arrangement in the QD absorber layer, by changing the ratio of formamidine (FA) to Cs components in perovskite Cs 1−x FA x PbI 3 (Figure 9A), could promote the exciton dissociation.The step level distribution was obtained by depositing QD absorption layers with a layer-by-layer approach, which facilitated the charge separation at the internal interface of the device and improved the PCE up to 15.74%.
Inspired by the interpenetrating network structure in OSCs, Barkhouse et al. 140 prepared bulk heterojunction QDSCs based on n-type wide-bandgap oxide (TiO 2 ) and p-type PbS colloidal QDs, which efficiently increased the depletion region.Thus, the bulk heterojunction QDSCs delivered a higher PCE of 5.5%.Likewise, Rath et al. 141 blended n-type Bi 2 S 3 nanocrystals with p-type PbS QDs to form bulk nano-heterojunction devices, which also allowed the rapid exciton dissociation at interfaces.Thus, the bulk nanoheterojunction device afforded an over threefold improvement in J SC (24.2 mA/cm 2 ) and PCE (4.87%), compared with the bilayer p-n junction one.In addition to the oxide nanoparticles, ultrafast dissociation of excitons in QDs through electron transfer to molecular acceptors was reported.Baek et al. 139 prepared the QDorganic hybrid solar cell via introducing a small molecule, IEICO, into the QD/organic (PbS/PBDTT-E-T) stacked structure.IEICO complemented the QD absorption and created an exciton cascade with the host polymer PBDTT-E-T, thus enabling efficient energy transfer and promoting exciton dissociation at heterointerfaces (Figure 9B).The hybrid solar cell with smallmolecule bridges achieved a superior PCE of 13.1% with remarkably enhanced J SC .Similarly, Hu et al. 142 prepared a hybrid interfacial architecture with the CsPbI 3 QD/ phenyl C 61 butyric acid methyl ester (PCBM) heterojunction, which also promoted charge transfer from the CsPbI 3 QD layer to the PCBM layer.The devices delivered an excellent PCE of 15.1%.In a word, F I G U R E 9 (A) Schematic overview of layerby-layer assembly showing a perovskite quantum-dot (QD) film composed of different layers of QDs.Reproduced under the terms of the CC-BY license: Copyright 2019, the authors. 138(B) Illustration of PbS QD/polymer (bilayer and bulk heterojunction) and PbS QD/ polymer-SM-bridge structures.CQD represents colloidal quantum dot, BHJ represents bulk heterojunction, and SM represents.smallmolecule acceptor Reproduced with permission: Copyright 2019, Springer Nature Limited. 139onstructing heterojunction architecture with proper materials is a facile and effective approach to promote charge transfer in QDSCs for improved PCEs.

| CONCLUSION
Great progress has been achieved to manage the exciton behavior in OSCs and QDSCs in recent years.In this review, we first introduce the parameters that impact the exciton behavior and the techniques for exciton diffusion measurements.Then, we focus on the recent advancements in exciton diffusion and exciton dissociation in OSCs and QDSCs.In OSCs, improving the crystallinity of materials, reducing disorder, optimizing molecular orientation, enhancing PLQY of materials, and enlarging the spectral overlap can efficiently increase the exciton diffusion coefficient.In contrast, the methods to manipulate exciton lifetime are limited and empiric, including introducing triplet sensitizers, diluting effect, and end group engineering.As for exciton dissociation, different mechanisms were developed to rationalize the near-zero driving force in high-efficiency OSCs with nonfullerene acceptors.Material design and device engineering were also applied to reduce E b , approaching the limitation of the driving force for exciton dissociation.In QDSCs, the exciton lifetime can be easily managed by particle size control and surface modification.Although the exciton diffusion coefficient can be enhanced by reducing the packing distance of QDs and optimizing the heterogeneity of QDs and/or surface modification, the strategies are still empiric.In addition, exciton dissociation in QDSCs can be effectively promoted by increasing the width of a depletion region, regulating the band arrangement, and introducing heterojunction architectures.
Since exciton behavior plays a crucial role in photovoltaic conversion of OSCs and QDSCs, more efforts should be made to regulate it and then boost PCEs: (1) developing unified and reliable methods to measure the intrinsic L D and that in devices; (2) seeking efficient strategies to prolong the exciton lifetime of organic semiconductors and QDs; (3) according to the theoretical model, increasing the exciton diffusion coefficient of organic semiconductors via material design and device engineering; (4) establishing reasonable models for exciton transport in QDSCs and thus efficiently improving exciton transport; (5) reducing the driving force for exciton dissociation as low as possible without sacrificing J SC to decrease the energy loss of devices and exploring the inner mechanism; and (6) building ideal interpenetrating network structures via device engineering to further improve exciton dissociation in QDSCs.

F
I G U R E 1 Schematic of techniques to measure L D : (A) photoluminescence (PL) quenching, (B) transient absorption spectroscopy, (C) time-resolved microwave conductivity, and (D) device-based photocurrent-ratio method.EBL, electron-beam lithography.

F
I G U R E 5 Transient absorption microscopy images of exciton diffusion.Images formed by spatially scanning the probe beam relative to a pump beam.Two-dimensional normalized carrier density profile distribution on the fcc(111) plane of a CdSe quantum-dot (QD) superlattice at two different time delays: (A) 1 ps and (B) 2.5 ns.The transmission value (ΔT/T) was normalized to the maxima at 1 ps.(C) Normalized one-dimensional (1D) exciton density profiles of the superlattice fitted with Gaussian functions at two different delay times: 1 ps and 2.5 ns.(D) Normalized 1D exciton density profiles of a disordered film at two different delay times: 1 ps and 3 ns.Reproduced with permission: ) infrared absorption spectroscopy of IT-M, FCA, and IT-M:FCA thin films.Compared with the neat films, the position and intensity of the vibration band in the IT-M:FCA blend film exhibit obvious variation, which demonstrates the strong intermolecular vibrational coupling between IT-M and FCA molecules.Moreover, the ring vibrations and C=O vibrations in the excited state of IT-M decay more rapidly in the IT-M:FCA film (Figure6C,D), which indicates that the vibrational modes are easier to relax to lower-frequency modes or to surrounding environments in the blend film, suppressing vibrations and torsions of IT-M.Thus, they proposed that the reinforced rigidity of IT-M molecules efficiently limited intramolecular motion and suppressed nonradiative decay, improving the exciton lifetime of IT-M and benefitting excitons to transport to donor/acceptor interfaces (Figure6E,F).As a result, the PCE of OSCs based on PBDB-T:IT-M increases from 11.45% to 12.31% with enhanced J SC by adding FCA.

F
I G U R E 6 Characterization of molecular vibrations in electronic ground/excited states and the proposed mechanism.(A) Steady-state infrared (IR) spectra of the IT-M, FCA neat, and blend films.(B) Transient IR spectra of the IT-M neat and blend films (recorded at 0 ps time delay).(C, D) Time-dependent changes in the intensities of peaks at different wavelengths in IT-M neat and blend films.(E, F) Mechanism underlying the prolongation of the lifetime of photogenerated excitons and exciton behavior in IT-M neat and blend films.Reproduced with permission: Copyright 2020, Wiley-VCH GmbH.