Inorganic–organic interfaces in hybrid solar cells

In this review, we present important concepts to describe inorganic–organic interfaces in hybrid solar cells. We discuss the formation of hybrid interfaces, provide an introduction to the ground-state electronic structure of the individual components, and detail the overall electronic landscape after combining into a hybrid material for different relevant cases. We then explore the impact of hybrid interfaces on photophysical processes that are crucial for the photovoltaic performance of hybrid solar cells. Within this framework, we discuss methods for hybrid interface modification toward the optimization of hybrid solar cells, such as doping, the application of interlayers, and morphological control.


Motivation: why go hybrid?
A hybrid solar cell utilizes organic and inorganic semiconductors (OSCs/ISCs) to convert sunlight into electricity using the photovoltaic effect. The interfaces that are formed between the inorganic and organic material domains of these devices, which control many of their key attributes, are naturally called hybrid interfaces.
Hybrid interfaces are often implemented to improve certain interface properties or facilitate functions that would be impossible in a non-hybrid configuration. Two important examples are shown in figure 1. But hybrid interfaces may also come into being through pure necessity, for example, the need to collect charges from an organic bulk heterojunction at an (inorganic) metal oxide electrode. This review covers both these varieties of hybrid interface, though it is the promise of the former that is our main motivation.
What could be the appeal of hybrid solar cells in today's photovoltaics landscape? For decades now, multiple junction cells featuring III-V semiconductors have been the champions of ultimate solar cell power conversion efficiency, while the more economical cells based on silicon, CdTe, CIGS(e), and more recently lead halide perovskites continue to gain efficiency [1]. With these technologies standing dominant, and indium tin oxide (ITO) and other effective inorganic transparent conductors in place, what are the properties through which OSCs can complement or even out-perform ISCs?

Chemical composition and compositional tuning
The virtually infinite flexibility in adjusting the chemical structures of organic materials allows the fine-tuning of their chemical, electronic, and mechanical properties. This allows combining controlled surface bonding, efficient passivation, and selective charge transport in a single molecule. It also permits alteration of the solubility characteristics, which opens up a large variety of processing routes. Solar cells are slated to eventually supply a large fraction of the world's electricity demand [2]. To provide the required terawatts of electrical power, the sheer scale of solar cell production necessitates replacement of solar cell components employing scarce resources, like indium and gallium, with more earth-abundant materials. Since organic materials are comprised largely of carbon and hydrogen, replacing their inorganic counterparts in solar cells can address some issues of material scarcity. Furthermore, introducing innovative materials that are cheaper or require lower processing temperatures harbors the potential to reduce the costs and energy requirements for fabrication and recycling. Such advances can further improve the ecological and economic benefit of solar cells and accelerate the transition to a sustainable way of life.

Photophysics
π-conjugated organic molecules and polymers can exhibit very strong electronic absorption bands. While the bandwidth of these absorption features in a single chromophore is typically not comparable to the continuum absorption of a bulk crystalline semiconductor, the many sources of broadening that act upon these electronic bands, particularly in the solid state, as well as the increasing density of allowed electronic transitions above the optical gap can lead to what is effectively a quasi-continuous absorption band. Therefore, the thickness of a π-conjugated organic film can be a fraction of that of inorganic absorbers while still achieving a comparable absorbance. Since thinner solar cell absorber layers favor a higher device voltage due to a reduction in bulk defect-mediated recombination, thinning the inorganic absorber layer by the incorporation of a light-harvesting organic chromophore layer may be a useful device strategy [3].
OSCs are excitonic, and thus can exhibit the spectral conversion processes singlet fission and triplet fusion. When used in conjunction with an already-mature high-efficiency ISC, these processes offer a promising path toward surpassing the Shockley-Queisser limit [4], which describes the efficiency of an ideal single-junction solar cell, wherein the only loss mechanism is that of radiative recombination, under one-sun excitation [5,6]. Pushing the efficiency of mainstream solar cells beyond this limit is important, because continuously falling solar cell prices imply that the balance of systems costs (the costs of the utilities and services required for electricity production from solar cells) account for an increasing fraction of the total system costs. In the limit of the cost of the solar cell itself being negligible and fixed balance-of-systems costs, improving the solar cell efficiency-making more electricity per module-is the only route to a reduction in the cost per unit of electricity.
Given the complexity of hybrid inorganic-organic systems, it is no surprise that their comprehensive treatment is still in its infancy. In this review, we aim to identify concepts to describe interfaces of hybrid solar cells both in the ground and excited states. We first define hybrid interfaces and discuss their formation. This is followed by an introduction to the electronic structure of the individual components and their impact on the overall electronic landscape after combining into a hybrid material. We later explore the impact of hybrid interfaces on various photophysical processes that occur in photovoltaics. Finally, we discuss methods for hybrid interface modification that enable tuning of the excited state dynamics, such as doping, the application of interlayers, and morphological control. To keep this review focused, the details of solar cell structure outside of the hybrid interface are considered only insofar as they provide the necessary background for discussing the interface. Other aspects are covered by other hybrid solar cell reviews [12][13][14][15][16][17][18][19][20][21], as well as recent reviews on organic [22][23][24] and inorganic [25][26][27][28] solar cells. For a technology analysis we refer the reader to reference 29.

The hybrid interface
Inorganic-organic composites rely on the synergistic effects of combining two materials classes that have typically been treated separately according to their physical properties. This is done in order to improve different aspects of performance, often in ways that deviate from commonly employed theories, such as mean field theory or effective medium approximations. The rule of mixtures is a foundational concept in materials science that defines upper and lower boundaries for the properties that composites can express. It states that composite properties can be estimated as the volume weighted average of the properties of the included phases. This rule most frequently applies when considering macroscale composites. The IUPAC defines a hybrid material as a 'material composed of an intimate mixture of inorganic components, organic components, or both types of components' for which 'the components usually interpenetrate on scales of less than 1 μm' [30]. As the dimensions decrease, strong deviations from the properties of the individual materials can often be observed. For instance, an electrical conductivity has been reported for nanoparticle/polymer hybrid materials that cannot be described as a combination of the individual components, and instead is dependent on an unusually high conductivity at the hybrid interface [31,32]. Important properties can arise depending on whether the inorganic and organic materials share orbital overlap or interact via van der Waals forces. While many properties of hybrid materials are accessible by just relying on intimate contact, certain exotic properties necessitate orbital overlap. This intermixing can result in profound changes in the optoelectronic properties as a result of the inherent differences in the properties of the individual constituent materials, such as their exciton degeneracy, increased absorption coefficients for hybrid excitons, dipole strength, or the lifetime of charge carriers [33]. One example is the formation of hybrid excitons at hybrid interfaces that are predicted to give rise to a greatly enhanced optical non-linearity [34]. As a result, inorganic-organic hybrid materials have been identified as key technologies for the further development of optoelectronic devices.
Hybrid solar cells consist of ISCs [35] and OSCs [36]. ICSs and OSCs express significantly different properties that are ultimately derived from their different bonding characteristics. Where inorganic materials are typically comprised of atoms bonding predominately via covalent or ionic mechanisms, organic semiconducting materials are more likely to be comprised of a series of alternating π bonds with molecules interacting via weaker van der Waals bonding. These differences in bond energies have a profound effect on the optoelectronic properties (as discussed in sections 3 and 4) and morphologies. ISCs will readily form hard crystalline structures (although their amorphous counterparts are also technologically important) with open shell structures that have highly reactive surfaces that require passivation in order to control the desired properties. The morphology of interfaces between ISCs and organic materials (not necessarily semiconductors) is discussed in the remaining part of this section. While organic materials encompass a wide variety of form and function, we discuss only those of particular importance for optoelectronic applications, namely ligand shells, π-conjugated polymers, and small molecules.

Interface chemistry: ligands and SAMs
To control the termination of an inorganic material, organic molecules can be attached to their surface via coordination or covalent bonds. Understanding ligation effects is of central importance for detailing the consequent evolution of the optoelectronic properties in hybrid materials and interfaces. The functional headgroup can have varying iono-covalent or weak binding characteristics with the inorganic surface. M L H Green presented a formalism for understanding this interaction at the molecular level via the covalent bond classification (CBC) theory, which was later expanded to inorganic nanoparticle systems [37][38][39]. According to CBC theory, ligands can be classified into three different categories according to the number of electrons they contribute to the interfacial bond. L-type ligands are neutral donors that rely on lone pairs for coordination with inorganic surfaces, resulting in relatively weak, dative bonds, and contributing two electrons to the inorganic constituent. organic molecular beam deposition and (c) solvent-based film casting. The orange color indicates an interlayer (e.g. a dielectric) that modifies the interface transfer processes. Nanoparticles with (d) self-assembled monolayer (SAM) and (e) ligand shell. The orange interface unit is a linker that facilitates the attachment of the organic unit and can additionally control transfer processes. planarity of the molecule, the structure of the side chains, the presence or absence of a coordinating head group, and the orientation of the crystal facet that the organic material is coming into contact with. The morphology of the molecules at the inorganic-organic interface does not necessarily represent that of the bulk of the OSC. In many π-conjugated systems, and in particular for π-conjugated polymers, the ordering is described in terms of π-π stacking or lamellar stacking, which corresponds to packing in the direction of the π-orbitals or the side chains, respectively. Charge transport in π-conjugated materials can occur along the π-conjugated backbone (where it is fastest) or in the π-π stacking direction (where the reliance on interchain hopping results in slower transport), and is frequently impeded in the lamellar stacking direction due to the insulating aliphatic chains making interchain hopping in this direction unfavorable [72]. There has been a strong drive to develop methods for controlling the orientation and stacking motifs of these π-conjugated systems in order to improve aggregation behavior and achieve desired molecular orientations for improved performance.
When fabricating monolithic solar cells, layers of organic small molecules are often grown by organic molecular beam deposition or physical vapor deposition onto the inorganic counterpart that acts as substrate. In this case, the geometric structure of the first organic layer(s) and the growth mode strongly depend on the interfacial interaction. Organic film growth both on atomically clean as well as passivated Si have been actively studied for more than thirty years [73][74][75][76][77][78][79] and will be used here as pertinent example. In the chemisorption limit, the epitaxial registry and orientation of the first molecular layer is dominated by specific organic-inorganic interactions [80][81][82][83]. In contrast, in the physisorption limit the intermolecular interaction dominates the molecular assembly. However, the interface can induce the formation of so-called thin film phases [84,85]. The resulting film morphologies sensitively depend on the deposition rate, as well as the reactivity, crystallinity, and temperature of the ISC surface [86,87]. For weak organic-inorganic interactions, Volmer-Weber growth, dewetting, and desorption is often observed at temperatures around room temperature [88][89][90][91][92], especially under vacuum conditions. This can be a severe problem when attempting to probe interface properties, since a large fraction of the OSC is not, in fact, in contact with the ISC even for low film thicknesses. Sometimes a small decrease of the substrate temperature is enough to prevent the dewetting and desorption [85,90]. As another instructive example we mention p-sexiphenyl (6P) on ZnO, for which it was shown that the 6P molecular orientation can be switched by using different ZnO crystal faces of a given ISC [93] and that the growth mode can be controlled by partial fluorination of 6P [94]. Quite generally, partial fluorination is a potent strategy for subtly tuning organic film morphologies [95,96].
The reverse deposition sequence (inorganic on organic) has also been successfully demonstrated [97,98], but is usually avoided due to the high risk of damaging the organic material. Low-temperature deposition schemes exist for a number of ISCs that can help reduce detrimental effects on the OSC.
Due to their high molecular weight, the thermal evaporation of polymers is difficult. Deposition of polymers and large and/or sensitive molecules under vacuum conditions is possible via electrospray deposition [99][100][101][102]. More typically though, polymers are deposited from solution processing methods, such as spincoating, blade coating, or printing, among others, which results in edge-on or face-on stacking. When an end-on morphology is desired, polymers can also be chemically attached via either 'grafting to' or 'grafting from' approaches [103,104]. The 'grafting to' approach utilizes pre-formed polymers that associate to the inorganic surface via chemical attachment, but can result in poor coverage due to steric effects. The 'grafting from' technique takes advantage of a surface-bound initiator from which the polymer is grown via chain growth polymerization. End-on orientation can also be achieved by use of a fluorinated surface-segregated layer and/or suitable SAMs [105,106].

Electronic structure
We now turn to a discussion of the optoelectronic properties of hybrid interfaces, starting with the ground state electronic structure, which controls the fate of photo-generated charge carriers.

Intrinsic energy levels of ISCs and OSCs
The relevant energy levels for optoelectronic devices are those that derive from the valence shell of the atomic constituents. Intermixing of atomic orbitals gives rise to occupied and unoccupied molecular orbitals. In ISCs the atoms of the whole lattice overlap, which leads to formation of a continuum of energy levels known as energy bands. The highest occupied band and the lowest unoccupied band are the valence band (VB) and conduction band (CB), respectively. When the dimensions of inorganic materials are reduced from the bulk to lower than the Bohr radius of the exciton, they can eventually experience quantum confinement in one [108][109][110] or more [111][112][113][114] dimensions. As particle size is reduced, there are fewer available electronic states in the CB and VB, and eventually this results in the widening of the bandgap and development of discreet states near the band edges. Depending on the number of quantum confined dimensions, there can be profound effects on the density of states (DOS), as shown schematically in figure 4, and this can therefore significantly influence the optoelectronic properties, especially at heterointerfaces. Within the strong confinement regime, the dielectric constant will also be impacted by particle size, and in core-shell type structures, the dielectric constant of the matrix material will significantly impact the properties of the quantum confined material. This has the result of modulating the absorption coefficient, donor energy levels, and electric field concentration within the dots, among other effects [115][116][117], which in turn will determine the electronic properties at heterointerfaces.
In contrast, π-conjugated organic molecules in the condensed state feature large orbital intermixing only within the molecular building blocks, whereas intermolecular hybridization is typically weak. This gives rise to comparably smaller intermolecular band dispersion [118,119]. Accordingly, the electronic energy levels are primarily determined by the occupancy of the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO). π-conjugated polymers feature bands that derive from the hybridization along the extended polymer chain. State energies additionally depend on the local electrostatic potential, which is influenced, e.g. by molecular sliding [120] and disorder [121].
The surface disrupts the periodic structure of a solid. The resulting asymmetric potential gives rise to additional electronic states (Tamm/Shockley states) [122]. In addition, the different number of bonding partners at the surface compared to the bulk can result in dangling bonds, surface reconstructions, and the formation  [135,136]. (b) Similar to (a) but with a doped OSC, which gives rise to band/energy-level bending in the OSC and ISC. Example: α-NPD:F6TCNNQ/ZnO [137]. (c) The LUMO of organic acceptor molecules overlaps energetically with the ISC occupied DOS. This gives rise to an ISC-to-OSC electron transfer and, in turn, a potential difference across the interface. This can be divided into two contributions: an abrupt potential step at the interface (interface dipole, Δϕ ID ) and a gradual potential gradient that extends into the ISC bulk (band bending, Δϕ BB ). The shares of these two contributions depends on the density of surface states and bulk dopant concentration. Example: F6TCNNQ on an ISC [138,139]. Additional details for each example are given in the main text below. of distinct surface crystal phases, as well as a structural relaxation of the surface layers. A different molecular orientation can be present at the surface than expected from the bulk structure [123]. Native defects in the near-surface region can sensitively influence the surface potential of ISCs, as observed, e.g. for ZnO [124][125][126]. Due to these effects, the surface electronic structure differs from that of the bulk, which can manifest itself as increased energetic broadening of the DOS or the formation of additional distinct states, leading to shallow and deep traps for charge carriers. The surface structures of many atomically clean and passivated ISC surfaces have been established in the last decades. However, this is not true in all cases. One example is the ZnO(0001) surface (i.e. the Zn-terminated facet of ZnO), for which it is still not fully clear whether and how the observed periodic structures [127] and oxygen species [128,129] at the surface derive from surface reconstructions [130], adsorbates like hydroxyl groups [131], sub-surface hydrogen [132], or a combination thereof [133,134]. Figure 5 presents one-particle energy-level diagrams for three characteristic hybrid interface configurations. Such diagrams are frequently used to depict the energy-level alignment (ELA) at interfaces. To that end, the energies of occupied (unoccupied) orbitals/bands are determined by measuring their ionization energies (electron affinities), which correspond to the energy difference between the neutral ground state and the final state with one extra hole (electron) in the orbital/band, respectively. These are also the energies which are most relevant for charge transport.
Upon the removal of an electron, the polarization of the environment screens the bare charge. This dielectric stabilization of the hole reduces the ionization energy (IE). Conversely, the addition of an electron generates a distortion in the local electronic and atomic structure that stabilizes the electron affinity (EA) [140,141]. The charge and the response of the environment is jointly referred to as a polaron. Polaronic effects have a large influence on the charge transport in OSCs [72,[142][143][144][145][146][147][148] and quantum dots [149,150].
ISCs typically have relatively large dielectric constants, which results in effective screening of the charge carriers [151]. This leads to the working principle that bulk 3D ISCs are populated exclusively by free charge carriers at room temperature. Polaronic effects can in most cases duly be ignored for this category of materials, at least at room temperature and above.
Energy levels are typically given either with respect to the Fermi level or the vacuum level. Note that in the context of interfaces, the vacuum level refers to a location just on top of the surface of the materials and not at infinity [152][153][154]. For this reason, energy levels with respect to the vacuum level are sensitive to the surface potential that the electron traverses when leaving the material. This also applies to the work function, which is defined as the energy difference between the Fermi level and vacuum level. Changes in the surface potential directly translate to a corresponding change in the work function and the energy levels referenced to the vacuum level. These energies thus depend, e.g. on the facet of the ISC crystal or the molecular orientation in the organic layer [155][156][157]. This effect is particularly pronounced in organics, since they feature a very anisotropic charge distribution, with the π-orbitals giving rise to a negative partial charge above and below the π-conjugated plane [158]. This effect can be exploited to tune the energy levels via the implementation of an extra surface potential. This can be achieved, e.g. via a polar surface termination in the case of ISCs [136,159,160] and intra-molecular polar bonds in the case of OSCs [156,158,161,162].

Interface formation
It is often instructive to arrive at the final ELA of a particular interface by first aligning the energy-levels of the individual components assuming vacuum-level alignment, and second incorporating effects induced by the formation of the interface (see figure 5). In the following we initially assume that the intrinsic electronic structures of either constituent also hold for their surfaces and that no new states form upon interface formation.
One effect that has to be considered is a possible difference in screening efficiency at the interface compared to the ISC and OSC bulk. It was found that the electronic screening for small molecules is more efficient when they are on a metal surface than when surrounded by other molecules, which gives rise to a decreased electronic gap at the metal-molecule interface compared to the bulk [163][164][165][166][167][168][169][170][171][172][173]. Since ISCs, polymers, and small molecules can have very different dielectric constants, a similar effect can be expected at hybrid interfaces [174,175].
Next, we consider the ELA in the absence of any significant electronic interaction at the intimate interface. This means we first ignore the possibility of the formation of an interface dipole. In this case, vacuum-level alignment persists also after interface formation and controls the ELA directly at the interface. In contrast, the ELA far away from the interfaces is controlled by the Fermi-levels of the two materials. A possible misalignment of the Fermi levels for assumed vacuum-level alignment gives rise to a charge-carrier redistribution across the interface and within the materials to guarantee electronic equilibrium. This leads to a gradual potential drop that facilitates the alignment of the two Fermi-levels. This potential drop is referred to as band bending. In cases where no bands form, the term energy-level bending is appropriate. Importantly, sufficient free charge carriers are required for full alignment of the Fermi-levels, which is not guaranteed for OSCs, as discussed in the following.
Undoped OSC on passivated ISC (figure 5(a)): small molecule OSCs often have gaps >2 eV [36,176] and, thus, very small intrinsic charge carrier densities. In these cases and the absence of (intentional or unintentional) doping, vacuum-level alignment persists well into the bulk and no significant band bending occurs for organic layer thicknesses relevant for most opto-electronic devices [152,177]. In this case, the Schottky-Mott rule applies according to which the hole-injection barrier (HIB, the energy difference between the HOMO of the OSC and the Fermi-level of the ISC) is equal to the difference of the IE of the OSC and the work function of the ISC (ϕ), HIB = IE − ϕ.
As a prototypical example, we discuss tetracene (Tc) on crystalline silicon (c-Si) with hydrogen termination (H-Si). The ELA of this interface was experimentally derived via ultraviolet photoelectron spectroscopy (UPS), as shown in figure 6(a) [135,136]. Indeed, no significant work function change was observed upon interface formation, evidencing vacuum-level alignment at the Tc/H-Si interface. Notably, the experimentally derived ELA deviates from theoretical calculations shown in panel (b) of the same figure [178]. While the latter predicts the Tc HOMO to be lower in energy than the Si VB, the inverse configuration is observed experimentally. Possible reasons for the discrepancy are differences in Si doping (n-type in the experiments vs intrinsic in the calculations) and defects. In addition, the calculations were performed for the intimate interface and include the full screening of the Si surface. In contrast, due to the pronounced Vollmer-Weber growth mode of Tc on H-Si [90], the UPS experiments mostly probe Tc in multilayers, for which the screening is only by other Tc molecules. Lastly, the Tc orientation that was considered in the calculation is more upright than the one that was recently derived for ultrathin Tc on H-Si [90]. Due to the electrostatic effects discussed above, IEs for Tc films with the molecules in an inclined configuration should be higher than for those with more upright molecules, as has been observed for related molecules [179,180].
Greiner et al systematically investigated a number of OSC/metal oxide interfaces [183] where ϕ covered a large energy range. They found characteristic regimes for the relationship between the HIB on the one hand and IE and ϕ on the other hand. The behavior is essentially the same as what had been reported previously for N,N -di(1-naphthyl)-N,N -diphenyl-(1,1 -biphenyl)-4,4 -diamine (NPB, another name is α-NPD, which is used below) on a variety of organic and inorganic substrates [184] and is shown in figure 7. Below a Reproduced from [178]. © IOP Publishing Ltd. All rights reserved.
critical ϕ value, the Schottky-Mott rule applies. Note, however, that in contrast to the case of Tc/H-Si discussed above, additional interface dipoles can occur that do not depend on the ELA. Two examples that are discussed in more detail in the corresponding subsection below are the electrostatic polarization due to the interaction of the molecular quadrupole moment and the ISC surface as well as the 'push-back' effect that is caused by Pauli repulsion. We define an effective work function ϕ eff that accounts for these effects. Above the critical ϕ eff , Fermi-level pinning sets in. This behavior can be explained with electrostatic considerations alone [182,185,186], by assuming that the charge equilibrium between the electronic systems of the organic molecules and the underlying substrate is expressed by the Fermi-Dirac distribution function. As illustrated in figure 8, once the OSC HOMO DOS and the ISC Fermi-level start overlapping substantially (ϕ ≈ IE), electronic equilibrium requires charge redistribution across the interface to guarantee overall charge neutrality. The two regimes are characterized by slope parameters S = d(ΔE H )/dϕ eff of S = 0 and S = 1. As illustrated in figure 9, an abrupt transition between the two extreme S values is only observed for sufficiently thick films and narrow HOMO DOS. The less abrupt transition observed for realistic interfaces is sometimes referred to as 'soft' Fermi-level pinning [187,188]. Compared to the bulk, interface layers can have significantly higher degrees of disorder [18,189], giving rise to a broadened DOS. This affects the interfacial ELA [190] and makes the Fermi-level pinning softer [187]. Molecules at the interface can feature a different molecular-orientation distribution than those in the bulk [191,192], and are subject to a different electronic screening [167,193,194]. As explained above, both effects influence IE and EA and, thereby, the pinning behavior at the interface. For example, it is the combination of more pronounced screening at the hybrid interface (estimated to be 0.3 eV) and the adoption of a flat-lying molecular orientation that is thought to drive the 6P/ZnO(0001) interface into the pinning regime, which is different than what is expected from the ELA between multilayer 6P films and ZnO(0001) [195].
Zahn et al investigated the ELA of a number of interfaces between OSCs and H-Si, Se-GaAs, and S-GaAs that, assuming vacuum-level alignment, have a large energetic distance between the Fermi level of the ISC and the molecular orbitals of the OSC, but were found to not obey the Schottky-Mott rule [196,197]. Defects in the ISC [198] and a molecular rearrangement as a function of coverage [196] were discussed as possible reasons for the observed behavior. Later, one of these molecules, perylene-3,4,9,10-tetracarboxylic dianhydride (PTCDA), was investigated for additional substrates that span a large ϕ range. After formation of the PTCDA film, all samples exhibited essentially the same work function and ΔE H , independent of the initial ϕ and reactivity of the substrate (see figure 7). This was attributed to a high density of interface states for this OSC [199]. The additional DOS can be accounted for via a Lorentzian broadening and gives rise to intermediate values of S, as shown in figure 9 and further detailed in [182,[199][200][201]. Alternatively, a new DOS in the gap of the OSC can also describe the observed behavior. The effect is that the gap of the OSC essentially becomes much narrower, and so does the ϕ range without pinning. The identical substrate-independent pinning was also observed for 1,4,5,8,9,12-hexaazatriphenylene-2,3,6,7,10,11-hexacarbonitrile (HATCN) [181] and is included in figure 7. However, for HATCN the intra-gap DOS was also observed beyond the interface region. For this reason, a chemical defect rather than interface states was inferred in this case.  This occurrence of Fermi-level pinning is related to the observation that the Schottky barrier height at metal/ISC interfaces cannot be readily tuned via the metal work function, which has puzzled scientists and plagued device specialists for the better part of the second half of the twentieth century. Also in this case, the resulting interface chemistry is often (but usually not adequately [202]) accounted for via a Lorentzian broadening of the DOS.
In cases where the OSC is Fermi-level pinned, its intrinsically low free charge carrier density is significantly increased due to contact doping from the ISC. The additional carriers give rise to a drastically increased energylevel bending in the OSC [203,204]. A complex distribution of the charge carriers can result for OSC bilayers. In these cases, the Fermi level of the conductive substrate controls the energy levels of the whole organic layer stack [205][206][207][208][209] (NB: also under illumination [210,211]). This situation can give rise to energy-level shifts upon formation of the organic hetero-interface that can be mistaken for interface dipoles at that interface. However, these are not due to the interaction at the interface at all, but rather a charge carrier redistribution that guarantees electronic equilibrium throughout the whole layer stack [203,212].  Doped OSC (figure 5(b)): additional charge carriers in organic layers can be induced via doping of the organic films [213][214][215][216][217]. While it was initially believed that molecular doping proceeds via integer charge transfer, there is strong evidence that a charge-transfer complex formation occurs instead in cases where the dopant and matrix consist of small molecules in intimate contact [218,219]. It is still being debated under which conditions integer charge transfer or charge transfer complex formation prevails. As a note of caution, we would like to point out that diffusion of dopant molecules to the interface has been observed [220][221][222][223]. In these cases, the interaction between dopant molecules and the ISC can mimic doping-related effects.
For doped OSCs, significant band/energy-level bending occurs also in thin films. As can be seen in figure 10(a), doping significantly decreases the width of the space charge layer in the OSC and gives rise to a different ELA between multilayer films of the OSC and ITO [224]. Figure 10(b) shows band bending in the OSC for substrates with different work functions [208]. Here, band bending in the OSC achieves a common ELA between the HOMO and E F within a few nm.
In the above example, the inorganic ITO is degenerately doped while the OSC is not. The opposite configuration is observed for poly(ethylenedioxythiophene):poly(styrenesulfonate) (PEDOT:PSS) on n-type H-Si. PEDOT:PSS is a highly doped organic polymer that is used as a high-performance charge-selective transport material, and depending on the formulation, can be prepared from several solvents, including water [227]. Formation of the PEDOT:PSS/H-Si interface leads to strong band bending in Si and the formation of an inversion layer at its surface [225], as shown in figure 11(a). Notably, the height of the Schottky barrier is directly related to the PEDOT:PSS work function. This is very different from the case of metal/Si contacts, where Schottky barriers significantly deviate from what is expected from the metal work functions [202,228,229]. This strongly indicates that interface states are much less important for H-Si contacts to PEDOT:PSS than to metals. Figure 11(b) shows a series of PEDOT:PSS/H-Si interfaces with different PEDOT:PSS formulations [226]. While no indication for band bending in PEDOT:PSS was observed, it was found that potential variations in the PEDOT:PSS layer do occur. These were attributed to a dipole layer either at the PEDOT:PSS/H-Si interface or at the PEDOT:PSS surface.
A case where band-bending was observed both in the inorganic and OSCs is discussed in reference [137] and shown in figure 12. Here, n-type ZnO and p-doped α-NPD form a hybrid pn-junction.
We note that molecular doping can also be used to passivate charge carrier traps in OSCs, and thereby improve the carrier mobility by orders of magnitude [230][231][232]. Interface dipoles: interface dipoles arise in cases of a preferential vertical charge distribution at the interface [155,233]. If matter replaces the vacuum on top of surfaces of inorganic materials, Pauli repulsion gives rise to the 'push-back' or 'pillow' effect, which is well-known for organic/metal interfaces [173,234,235]. The name refers to the fact that the adsorbate displaces the electron-density that spills out of clean metal surfaces. The electron density redistribution gives rise to an interface dipole. The identical phenomenon was reported to occur at metal oxide/organic interfaces. Here, an interface dipole of approximately 0.3 eV was consistently observed for a number of different interfaces [137,183,224,236] and attributed to the push-back effect.
In some cases, interface formation was found to induce an intramolecular charge density rearrangement. At the interface between 6P and the non-polar (1010) facet of the ionic semiconductor ZnO, Della Sala et al reported an electrostatic coupling of the molecule with the periodic dipolar electric field generated by the Zn-O surface dimers that gives rise to an induced vertical dipole moment [237]. Additional evidence for the importance of the molecular quadrupole moment comes from organic heterojunctions. Calculations predict that at the C 60 /pentacene interface, the non-zero multipole moments of pentacene gives rise to electrostatic polarization of C 60 [238]. A particularly strong lateral quadrupolar interaction was measured for a blend formed by a single-monolayer of flat-lying perfluoropentacene and pentacene molecules [162]. DFT calculations for the OSC/SAM/gold interface indicate that intramolecular polarization also occurs at the OSC/SAM interface [239].
Different to such induced dipoles, polar molecules have a permanent dipole moment. If polar molecules exhibit a preferential alignment at the interface, they give rise to an interface dipole [240,241], as will be further discussed in the context of organic interlayers below.
An interface dipole can also be caused by substrate-to-adsorbate ground state charge transfer at the interface. On the one hand, states of OSCs and ISCs can hybridize, which leads to fractional ground state charge transfer at the intimate interface. On the other hand, interfacial charge transfer can be also due to integer charge transfer between localized states in the ISC (e.g. defects) and molecular orbitals localized in the OSC. In many cases, a detailed understanding of the quantum-mechanical nature of the ground state charge transfer is still not established. This is due to mainly two reasons: (a) A theoretical description is challenging. For example, the ELA and, thus, the amount of interfacial electron transfer (IET) depends on the employed theoretical method [242]. The sensitivity to the employed method also extends to describing fractional and integer charge transfer [243]. Recent calculations found that both transfer mechanisms can be present at hybrid interfaces simultaneously [244]. (b) The real interface structure is often unclear. However, the structure of the inorganic surface (defects, reconstructions, terminating atom type, presence of adsorbates) and the adsorption geometry of the organic molecules (adsorption site, molecular orientation, adsorption-induced changes in the molecular conformation) control the electronic structure at the interface and which part of the inorganic and organic materials interact most intimately.
Experiments that probed the electronic (and in some cases also the geometric) structures of a number of different moderate molecular acceptors on transition metal oxides identified a prominent role of ISC nearsurface defects for the observed interfacial ground state charge transfer [124,126,134,[245][246][247][248]. The charge transfer is due to the misalignment between OSC EA and ISC Fermi-level (cf figure 8) and larger EAs give  [137]. © IOP Publishing Ltd. All rights reserved. Figure 13. (a) Calculated band bending contribution to the overall work function change as a function of surface state density for F6TCNNQ deposited on undoped GaN. Circles indicate the measured band bending contributions for the indicated different surface preparation conditions. The surface state density obtained from this procedure agrees well with the value that can be determined from the bulk donor concentration and the initial surface band bending (before F6TCNNQ deposition) for the same samples, as further detailed in reference [138]. (b) Band bending contribution for ZnO(000-1) with different bulk dopant concentrations. Reprinted from [138], with the permission of AIP Publishing. rise to a more pronounced effect. Therefore, strong acceptor molecules offer a means of quantifying the ISC bulk and surface DOS [138,139,188,249,250]. To that end, the fraction of charges originating from the ISC surface is compared to the fraction stemming from its bulk. As illustrated in figure 5(c)), electron transfer from the bulk and the surface both give rise to a potential drop across the hybrid interface, but only the latter also registers as band bending. The potential drop (bend bending) directly corresponds to the work function change (ISC energy level shifts) upon interface formation, both of which can be probed with photoelectron spectroscopy. This procedure was used to systematically investigate how surface treatments and the bulk dopant concentration affect the ISC surface DOS [138]. As can be seen in figure 13, a strong effect of the surface state density is observed in all investigated cases. Notably, a pronounced influence of the bulk doping is observed only at very high dopant concentrations.
Note that while the described procedure allows quantifying the influence of the surface states on the charge redistribution across the interfaces, it does not allow a resolution of the nature of the interaction at the hybrid interface. The two very strong acceptors, 2,3,5,6-tetrafluoro-7,7,8,8-tetracyanoquinodimethane (F4TCNNQ) and 2,2-(perfluoronaphthalene-2,6-diylidene)dimalononitrile (F6TCNNQ) are an interesting case. While a pronounced chemical shift in the nitrogen atoms of molecules at the interface (indicative for charge transfer) are observed for H-Si [251], this is not the case for ZnO [249,252]. Therefore, despite further insightful investigations with other techniques [253,254], the exact nature of the interfacial interaction at these organic/ZnO interfaces is not yet completely resolved.
The introduction of insulating interlayers induces the transition from fractional to integer charge transfer [255]. The thickness of the interlayer controls the number of transferred charges [205,207,209,256], which shows that an interface dipole is formed between substrate and overlayer, whereas the insulating interlayer is neutral. The opposite trend in interface chemistry strength can be induced by covalently binding the organic and inorganic materials at the hybrid interface. The ELA at these interfaces has been studied as well and we list examples for Si [63,257] and TiO 2 [258,259].
We conclude this section by noting that organic and hybrid materials are often prone to chemical changes and structural rearrangements. The rate of these aging processes can be significantly increased by exposure to vacuum and irradiation. References [260,261] give a good overview about how these effects should be considered when determining the relevant energy levels for organo-halide perovskites, which is a particularly dynamic material class.

Photoinduced excited state dynamics
The purpose of a solar cell is to convert incident light into electricity via the photovoltaic effect. In broad terms, this is achieved by delivering electrons and holes that have been generated by photoexcitation of a semiconductor to separate electrical contacts, where an attached electrical load can be driven. A well-optimized solar cell operated at the maximum power point maintains a chemical potential difference between the electron and hole populations that is close to the band gap energy of the semiconductor. Charge carrier recombination within the solar cell beyond the amount necessary to satisfy thermodynamic constraints is a loss process, which reduces the chemical potential difference between electrons and holes, expressed as a reduction in the quasi-Fermi level splitting, and hence the achievable voltage of the solar cell.
The exact process by which a solar cell converts light into usable electricity depends on the details of the cell design, and this must be disambiguated in order to give a concise account of the relevant excited state processes. Our scope in this work is restricted to the different variations of hybrid inorganic-organic solar cells, and the hybrid interfaces of these devices in particular. We first make an important distinction regarding the nature of the semiconductor light absorber. For an OSC absorber layer, photoexcitation leads to exciton formation, and the mechanics of charge separation are chiefly concerned with driving efficient exciton separation at heterojunctions while providing a device structure that supports electron and hole diffusion away from the heterojunction. An OSC absorber also allows for spectral conversion processes, such as singlet fission, to be implemented. In the case of the semiconductor light absorber having a negligibly low exciton binding energy (this corresponds with all classical bulk ISCs, but also with the bulk hybrid lead halide perovskite family), free electron-hole pairs are generated directly upon photoexcitation.
In the following, we briefly discuss exciton formation and transfer mechanisms, and then collect the different possible exciton evolution pathways available in hybrid solar cells (see sketch in figure 14). We then separately discuss the role of the hybrid interface in the two different cases described above, namely that (i) excitons or (ii) free charge carriers arrive at a hybrid interface.

Exciton formation within OSC and ISC layers
If a photon reaches an absorber of the solar cell, it may get absorbed by inducing an electronic transition for which the energy difference between initial and final state corresponds to the photon energy. The immediate effect of the electronic transition can be described as the filling of one formerly unoccupied higher-lying electron level and the simultaneous creation of a hole in the lower-lying level. In response to the electronic transition, the electronic and atomic structure of the local environment adapts to the new charge configuration and stabilizes it. This is referred to as electronic and lattice relaxation, respectively, and the former is much faster than the latter. An exciton is a quasi-particle that consists of the electron-hole pair as well as the environmental response. The exciton binding energy is the energy difference between the bound electron-hole pair and the situation where the electron and hole are sufficiently separated for their Coulomb interaction to be negligible. The exciton immediately after the transition is 'hot', since it typically contains excess vibrational and/or electronic energy compared to the lowest excited state of the same multiplicity. Accordingly, the exciton is considered 'cold' once it has thermalized by shedding this excess energy.
As mentioned previously, the relatively large dielectric constants of bulk crystalline ISCs reduces the Coulomb attraction between geminate electron-hole pairs as a result of dielectric screening. These materials are often referred to as non-excitonic since the exciton binding energy is small, on the order of a few meV [262], which is less than the thermal energy available at room temperature. The electron-hole pairs produced by photoexcitation rapidly dissociate into free charge carriers, and excitonic effects can be largely ignored. Despite different initial expectations, hybrid lead halide perovskites also appear to be non-excitonic by this definition [263,264]. This is in contrast to what occurs in the case of a typical π-conjugated organic material [265,266] and lowdimensional ISCs [114,[267][268][269][270][271][272]. Here, two factors combine to produce distinctive excitonic effects. A low dielectric constant for organic and organic-dominated hybrid materials results in weak dielectric screening and hence a strong Coulomb interaction between electrons and holes. In addition, a high degree of spatial confinement in one or more dimensions occurs, resulting in localizing and hence increasing the Coulomb attraction of electrons and holes [273]. This leads to the formation of strongly bound and localized (Frenkel) excitons upon photoexcitation. These necessitate dissociation in order to generate photocurrent in a solar cell.
Frenkel excitons exhibit significant correlation effects, meaning the total electron spin of the exciton sensitively affects the state energy (equivalently, the exciton binding energy). Neglecting for simplicity the various spectroscopic selection rules that come into play, the allowed optical transitions for such a system will be between the electronic ground state and exciton states with the same total electron spin, while optical transitions between unlike spin states are forbidden. According to Hund's rule for maximum multiplicity, triplet excited states (corresponding to a total electron spin of 1) will always be lower in energy than singlet excited states (total electron spin of zero) for a given principle quantum number. The substantial energy difference between singlet and triplet excited states in some OSCs can be exploited for useful functions. Singlet fission is a process by which a singlet exciton splits into two triplet excitons of approximately half the energy, located on adjacent chromophores [274,275]. The inverse process is named triplet fusion or triplet-triplet annihilation. Both processes have the potential to support solar cell efficiency beyond the Shockley-Queisser limit in hybrid solar cells, with singlet fission serving as an exciton multiplication scheme that can reduce thermalization losses [4,135], while triplet fusion can be used in concert with a triplet sensitizing light absorber to generate additional photocurrent from light that is sub-resonant with the optical gap of a solar cell [276,277]. Several species of quantum dots, especially those from the metal chalcogenide family, have been employed to similar effect in exciton multiplication and triplet fusion schemes [278][279][280].
Experiments that probe excitons usually measure energy differences (e.g. optical transitions) that do not directly correspond to energy-level differences of the two non-interacting single particles. For example, the difference between the HOMO IE and the LUMO EA (the transport gap) deviates significantly from the energy of the first optical excitation (the optical gap). This is because even in their simplest representation, excitons have to be described as two bound particles (electron and hole) that interact with themselves in addition to their environment. Therefore, one has to be particularly careful when discussing excited states of organic molecules in a one-particle energy-level diagram, as detailed in reference [281].

Free charge carrier collection
Some hybrid solar cell working principles rely on free charge carriers arriving at hybrid interfaces. This is the case when only non-excitonic absorber materials are used (case IX). However, it also applies to cases where the organic layer itself contains a heterojunction which facilitates exciton dissociation (case OXb). Hybrid interfaces in these devices are formed between the absorber layer and an electrode or transport layer. Examples are an ISC absorber in contact with an organic charge selective contact, or an OSC heterojunction in contact with metal oxides. In a tandem cell architecture, both the organic and inorganic layers produce free charge carriers. In this case, the hybrid interface between these two absorber layers serves as a recombination layer [282][283][284][285][286][287].
Free charge carriers are generated at heterojunctions in excitonic systems, or directly upon photoexcitation of a non-excitonic semiconductor. In the former case, electrons and holes diffuse through separate material domains to their respective electrodes. In the latter, electrons and holes have to be separated by steering their respective currents toward separate electrodes. The current densities for electrons and holes are controlled by their conductivities and the gradients of their electrochemical potentials, which, in turn, is equal to their quasi Fermi levels [288]. Maximizing the carrier current to the appropriate contact thus requires a negative gradient of the respective quasi Fermi level and a sufficiently high conductivity, while minimizing the current in the reverse direction requires the opposite. To achieve this anisotropy, at least two distinct regions of the device are required.
A Si pn homojunction is a frequent realization of this requirement. Here the doping level across a monolithic silicon slab is deliberately varied in order to implement the required spatial variation in carrier conductivity. Another strategy for charge carrier separation is the implementation of heterojunctions, where the semiconductor absorber is interfaced with unlike materials, which impart the functions of carrier selectivity [288]. In perhaps the simplest example of this concept, high/low work function metals can be used to create charge-selective heterojunctions on silicon. However, the high density of gap states at the interfaces gives rise to non-ohmic contacts [202,228,229] and leads to substantial surface recombination losses [289]. A more successful approach is found in silicon heterojunction cells employing thin amorphous silicon heterojunctions [290]. Dopant-free hetero-contacts are another approach [291][292][293]. A number of hybrid heterojunctions consisting of organic materials on crystalline silicon have been reported [294][295][296][297]. In this role, the strength of organics are their weak interface chemistries, which may reduce the detrimental effects of Fermi-level pinning (compared to metals) and give rise to better passivation of the ISC (compared to inorganic dopant-free hetero contacts). PEDOT:PSS has shown great promise as a hole-selective contact on silicon [298][299][300][301]. In section 3 we discussed the ELA at the PEDOT:PSS/H-Si interface. Here, we discuss the charge transport properties. Two different models could apply: in a heterojunction between a p-type semiconductor (PEDOT:PSS) and n-type semiconductor (c-Si), the transport processes are dominated by the diffusion of minority carriers. In contrast, a contact between a metal-like semiconductor (PEDOT:PSS) and a high-mobility semiconductor (c-Si) should behave as a Schottky diode, where thermionic emission over the Schottky barrier controls the charge transport. Jäckle et al found that the PEDOT:PSS/H-Si interface behaves like an abrupt pn heterojunction rather than a Schottky diode [225].
The ELA at the hybrid interface controls a possible contact doping into the interface-near region of the absorber layer, which selectively influences the local conductivity for electrons and holes [288,[302][303][304][305], but also leads to non-radiative recombination for materials with low mobility, like OSCs [204,306]. For sufficiently low bimolecular recombination rates, the large increase in mobility already for ultralow doping [232], together with a positive effect on the built-in field, outbalances the increased recombination and warrants employing electrode materials with work functions that force the contacts beyond the onset of ohmic transition [187].
The current density-voltage (J -V) characteristics and open circuit voltage V OC at a hybrid heterojunction between an organic p-type semiconductor (i.e. the Fermi level of the contact material is closer to the OSC HOMO than to its LUMO) and an inorganic n-type semiconductor was semiclassically calculated by Renshaw and Forrest [151]: Here, n O is the ideality factor that depends on the trap density. J ph is the total photocurrent, which is the sum of the exciton current generated in the OSC times the exciton dissociation efficiency at the interface and the direct carrier generation current in the ISC. χ(V) is the bias-dependent reverse-bias factor with χ(0) → 1. The interface energy gap ΔE OI is the energy difference between IE of the OSC and the EA of the ISC. k r (k d ) are the hybrid CT state recombination (dissociation) rate. The field dependence k d can be described by the Onsager-Braun model [307,308]. a is the characteristic width of the hybrid CT state at the organic-inorganic interface. N HOMO (N c ) is the OSC HOMO (ISC VB) effective DOS. k rec is the bimolecular recombination rate. Further details and formulas for OSC/ISC interfaces with other type configurations can be found in reference [151].
Charge transport through SAMs on ISCs in a metal/organic/ISC configuration was studied for a number of inorganic-organic combinations and modeled as thermionic emission over a Schottky barrier attenuated by tunneling through a 'molecular insulator' [63,257,309], akin to the transport through a metal-oxidesemiconductor junction [310]. Recently, a SAM of methyl-substituted carbazole on ITO has been utilized as the hole-selective layer of a lead halide perovskite absorber used within a silicon/perovskite tandem solar cell [311]. The SAM made a key contribution to the record-setting device performance and high photostability through fast hole extraction and minimized nonradiative recombination at the hole-selective interface. The compositional flexibility and self-limiting monolayer growth of the carbazole derivative, the attributes of which belong to organic molecules in particular, appear instrumental in this function, and make this a good example of the potential boon of hybrid approaches to solar cell interfaces.

Exciton diffusion and harvest at hybrid interfaces
While single electrons and holes carry a net charge and therefore respond to a gradual electrostatic potential, like band bending, or to a gradient in the quasi-Fermi level, this is not the case for excitons, which are electrically neutral. The Coulombically bound pair of charges that constitute an exciton cannot spontaneously drift apart from one another, at least when the exciton binding energy exceeds kT. An abrupt change in some electronic property-for instance, the EA-is required to drive exciton dissociation, as is typically implemented at heterojunctions. Here we discuss the case that excitons must travel to a hybrid interface prior to their conversion into free charge carriers (cases OXa and OXc) through dissociation at the hybrid interface (cases HXa and HXa'), or exciton transfer across the hybrid interface followed by spontaneous dissociation in a low-exciton binding energy SC (case HXb).
Excitons are typically required to migrate throughout the absorbing medium before encountering a suitable interface. Exciton transfer, migration, or diffusion, as distinct from exciton dissociation, refers to the concerted, resonant relaxation and excitation of electronic transitions in nearby absorbers giving rise to the appearance of an exciton moving throughout a system. Many such transfers may take place before an exciton ultimately expires via recombination or dissociation. The mechanism by which excitons undergo transfer depends on the nature of the electronic transitions involved in the donor (the initially-excited excitonic center) and the acceptor (which may be excitonic or non-excitonic).
In the case of both donor and acceptor transitions being electric dipole allowed, the dominating exciton transfer process is usually Förster resonance energy tranfer (FRET), wherein the transition dipoles of the exciton donor and acceptor couple through the Coulomb interaction, analogous to classical dipole-dipole coupling, and electronic energy is exchanged through a virtual photon. The effective range of FRET is expressed as a critical radius, R 0 , defined as the donor-acceptor separation at which the FRET quantum yield reaches 0.5; R 0 values of around 5 nm are typical in molecular solids [312]. While FRET was first formulated in terms of distinct molecular or atomic entities as donor and acceptor, FRET at hybrid interfaces is an important topic in the research of hybrid optoelectronic devices [34,44,313]. It has been invoked to describe the transfer of electronic energy to bulk crystalline silicon from metal chalcogenide nanocrystals [314,315] as well as from fluorescent organic molecules [316].
In the case that the exciton donor and acceptor lack an electronic dipole allowed transition at the relevant frequency, Dexter energy transfer may become significant. The Dexter transfer mechanism is based on the exchange interaction between adjacent chromophores that bear resonant electronic transitions, and permits a non-negligible rate of resonance energy transfer between molecules in spite of the lack of an electric dipole allowed transition [317][318][319]. Dexter transfer is the mechanism by which triplet excitons may migrate throughout materials when the radiative relaxation of triplet excitons to a singlet ground state is strictly forbidden. Since this applies exactly to singlet fission materials, both in the exciton diffusion and transfer stage, Dexter transfer is particularly relevant for incorporating singlet fission into solar cells [320].
FRET typically occurs at a much higher rate and at longer range than Dexter energy transfer, hence for materials in which FRET is active, the rate due to Dexter is usually negligible. On the other hand, where radiative relaxation of triplets is strictly forbidden, the triplet lifetime is significantly greater than that of singlet excitons, hence the low energy transfer rate may be counteracted. A key metric is the exciton diffusion length, which depends on both the lifetime and the hopping rate. It has been shown that, given optimal energetics of singlet and triplet excitons, triplet exciton diffusion can occur via a cooperative singlet and triplet exciton transport mechanism [321].
Exciton dissociation relies on IET. For relevant cases it is often assumed that IET can be described in the limit of weak electronic coupling between donor and acceptor, and, thus, non-adiabatically [318,[322][323][324][325][326]. In the framework of Marcus theory, this yields for the IET rate at a hybrid interface: Here ΔG is the free energy difference between electron acceptor and donor and relates to the ELA. ρ(E) is the ISC DOS at energy E, H(E) is the electronic coupling matrix element between donor and acceptor at energy E, and λ is the reorganization energy [327][328][329] that is related to polaronic effects.
Equation (2) is instructive since it combines several interface parameters that can, at least conceptually, be individually modified. In the following we substantiate the relevance of these parameters.
There is clear evidence that the interfacial ELA sensitively affects exciton harvest. The ELA at the organic/ZnO interface was shown to control the exciton transfer efficiency [330]. For organic heterojunctions, the ELA and the V OC of corresponding solar cells were found to correlate [331] (although with a significant offset of ∼0.5 eV that is thought to be linked to intramolecular vibrations [332,333]). An optimal driving force has been reported for converting excitons into free carriers in excitonic solar cells [334]. A similar driving force optimum was also found for triplet exciton energy transfer at the interface between pentacene and PbSe nanocrystals [335]. For organic/metal oxide DSSCs, a correlation between the electron injection and the driving force was observed [323].
The electronic coupling, H, was identified as another crucial parameter [336]. The influence of H was systematically studied for DSSCs via the linker type and the incorporation of insulating interlayers [323,337]. Equation (2) also expresses the fact that the donating OSC state should overlap with a dense manifold of states in the accepting band of the ISC in order to achieve a high IET rate. We note that ρ(k) has to be considered as well, since it was shown that FRET from quantum dots into the indirect ISC Si relies on phonon-assistance [314].
While instructive, equation (2) might not be adequate to describe the most relevant interface processes for exciton harvest.
Firstly, the suitability of Marcus theory in some circumstances has been questioned [338], partially due to the observation of fast transfer times that requires the transfer to be described adiabatically. Such small transfer times were observed for organic/TiO 2 [339] and organic/Si [340] interfaces and also for organic heterojunctions [341][342][343].
Secondly, the relationship between the parameters of equation (2) and the overall exciton dissociation efficiency becomes less simple if the forward process must strongly compete with other parasitic processes, for instance, recombination from a charge transfer state. For example, exciton dissociation at the organic/TiO 2 interface can be injection or diffusion limited [344]. A good interface has to facilitate efficient injection, meaning a sufficient rate of electron transfer, but more crucially, a minimal rate of the reverse process. Achieving optimum performance may entail striking a compromise between competing rates.
Lastly, the driving force for exciton dissociation via electron transfer and the rate of subsequent charge separation away from the interface may be related. After the charge transfer over a hybrid interface, electron and hole reside on opposite sides of the interface. Such a state is referred to as CT state ('CT exciton' and 'bound charge pair' are also sometimes used). If injection and subsequent CT state separation occurs coherently, the relevant free energy difference is that for the complete process instead of that for the initial IET. For example, a higher driving force for the initial IET increases the probability for hot CT states, which in turn might increase the probability for successful exciton dissociation. CT states have been observed in the following subset of hybrid interface systems: Zn(Mg)O/organic [345,346], (In)GaN/organic [109], CdS/organic [347], and transition-metal dichalcogenide/organic [348,349]. We note in passing that it might be instructive to additionally analyze the CT excitons in van der Waals heterostructures [350], since they allow to study the response of hybrid excitons to a local potential [351], their intra-layer and interfacial interaction [352], the role of momentum conservation [353,354], and Coulomb engineering of the bandgap and excitons [355]. However, it is still debated whether hot or cold CT states drive exciton dissociation in these materials, and the answer may well be material-dependent. CT states as mediators of recombination was reported for some hybrid systems [356][357][358]. Non-coherent exciton dissociation via cold CT states has been reported for the poly(3-hexylthiophene-2,5-diyl) (P3HT)/TiO 2 interface [359].
While we note that qualitative differences between organic/organic and organic/ZnO interfaces have been reported [359,362], it is instructive to consider findings from all-organic heterointerfaces, which have been more widely studied. In particular, the body of work on the role of hot CT states in charge separation is significantly larger [326,363,364]. Driving force, delocalization, and electron-hole separation were reported to improve the performance for organic heterojunctions and hint at the importance of hot CT states [341,342,[365][366][367][368][369][370]. The pathway to separated charges via a hot CT state is illustrated in figure 15. On the other hand, current generation with high efficiency was also observed from relaxed CT states [371][372][373]. Possible processes that derive from the interface morphology and can assist charge separation from a cold CT state are illustrated in figure 16 [361]. In addition, screening of the charge carriers [364], the mesoscopic structure [374,375], entropy [376], and different mobilities of electron and holes [377] can assist exciton dissociation. These findings indicate in particular the importance of mesoscopic interface structures that give rise  The same as (a) when an electric field is present. The electrical gap and the optical gap are shown. Their difference is due to the Coulomb attraction between electron and hole. The Coulomb energy is indicated by the gray line. The fact that the energies are actually broadened is indicated by the Gaussian line shape that superimposes on the levels. The electrostatic potential gradient reduces the energy barrier by ΔE 1 . Energetic disorder further reduces the barrier by ΔE 2 . ΔE 3 is a compensation resulting from the zero-point oscillation energy of the electron-hole pair in the Coulomb potential. E A is the energy that still needs to be overcome to separate charges. [ to gradients in crystallinity and intermixed phases that facilitate charge delocalization and favorable energy cascades [378].

Modifying hybrid interfaces
The chemical and electronic structure of hybrid interfaces is highly tunable by virtue of the many interoperating components, and many approaches have been undertaken to improve their optoelectronic functions. Figure 17(a) summarizes schematically the three interface modification categories discussed in the following. Example modification materials are given in the figure caption. We discuss the effects of the modification schemes via the expected influence on the parameters in equation (2), namely the driving force ΔG, the electronic coupling H, and the density of acceptor states of the ISC ρ. Echoing a point made previously, enhancing the yield of the desired forward process can also be achieved by reducing the rate of competing loss processes, and interface modifications can also be made to this end.

Doping
Modifying the ISC by doping allows tuning its electronic interface structure. The electronic structure controls the DOS that are available for accepting charge transfer. This is expressed by ρ in equation (2). Notably, ρ for acceptor states that give rise to bound CT states are to be avoided. As shown in figure 18, Cs doping of ZnO reduces the density of these gap states and the associated effect of electron trapping at the interface [358]. Also N doping of ZnO was shown to facilitate de-trapping of electrons [381].  (Si:B). The inorganic interlayer (IL) can represent a dielectric layer (SiO 2 , TiO 2 , ZrO 2 , Al 2 O 3 , MgO, ZnO, Y 2 O 3 , Nb 2 O 5 , HfO x N y ) or a (highly electronegative) surface-passivating agent (TiCl 4 , NF 3 ). The π-conjugated core of the SAM can be identical to the molecules that make up the OSC. In this case, conceptually only the linker of the SAM serves as interface modifier. Alternatively, the whole SAM can act as modifier, e.g. by facilitating an interface dipole or energy cascade. The green-red-striped filling accounts for both cases. (b) Illustration of how a linker with suitable chemical structure and an inorganic interlayer with optimized thickness allows for efficient injection and effectively prevents geminate recombination. (c) Illustration of an inorganic interlayer additionally modified with a sub-monolayar of acceptor molecules (F4TCNQ, F6TCNNQ, HATCN) that induce an interface dipole (μ). Dipoles with opposite orientations can also be introduced via donor molecules (19-electron organometallic sandwich compounds [379,380]). Extrinsic doping with Sn [382] as well as intrinsic doping from surface defects was shown to control the ELA at oxide/organic interfaces [124,126,183]. Changes in the ELA have direct implications for ΔG for forward and backward IET processes as well as for the available ρ [382].
A suitable boron (B) doping profile at Si surfaces gives rise to electron transfer from unsaturated dangling bonds on the Si surface to sub-surface B atoms, thereby passivating the surface Si dangling bond state. This de-activated ISC surface was found to induce a face-on orientation for some organic adsorbates [383][384][385], indicative of a larger H than in the case of chemically passivated surfaces.
Doping in the alloy regime allows tuning the band gap of the ISC. For example, the combination of ZnMgO and ZnO can be used to form a quantum well structure at the ISC surface and, which enhances FRET to an organic overlayer [313,386].

Inorganic interlayers
As shown in figure 19, several insulating oxide interlayers (SiO 2 , TiO 2 , ZrO 2 , Al 2 O 3 , [387,388] MgO [388,389], ZnO, Y 2 O 3 [388]) were found to improve the efficiency of the exciton dissociation process of DSSCs. Suppressed recombination due to an increased electron-hole separation after injection was suggested as the primary reason for the improvements. Passivation of surface states has also been found to be important [390].
At the P3HT/TiO 2 interface, a Nb 2 O 5 interlayer was shown to suppress the formation of CT states [391]. Also in this case, the interlayer was shown to reduce non-geminate recombination both when P3HT and PbSe colloidal quantum dots were used as absorber.
Already in the 1980s, Ng and Card reported an asymmetry in the SiO 2 tunneling barriers to electrons (low barrier) and holes (high barrier) [392], showing that interlayers can be used to achieve charge-selective transport. Essentially the same effect was also reported for TiO 2 [393,394].   Figure 20 presents an example for an unwanted oxide interlayer. As resolved with hard x-ray photoelectron spectroscopy, the buried Si-PEDOT:PSS interface contains a non-intentional sub-stoichiometric silicon oxide layer immediately after fabrication of the polymer layer [395]. When stored under N 2 , the interface continues to oxidize and forms stoichiometric SiO 2 . The oxide formation likely derives from residual water in the PEDOT:PSS film and a more stable surface passivation should be chosen, since the uncontrolled silicon oxide growth is expected to deteriorate the interface properties. These results were recently substantiated by a more extensive study [396].
As shown in figure 21, the triplet harvest yield at the Si-organic interface could be enhanced by using a sub-nm hafnium oxynitride (HfO x N y ) interlayer [397]. By contrast, no improvement was reported when using LiF interlayers at the same interface [398]. However, the thinnest interlayer in the latter study was 5 nm, significantly beyond the narrow thickness range for which a positive effect of the interlayer was reported in reference [397]. The HfO x N y interlayer was also found to reduce the surface recombination in Si via fieldeffect passivation. The passivation effect was most pronounced in the spectral region where Tc absorbs light. This indicates that charges from Tc excitons accumulate at the HfO x N y -modified interface, which might also mediate triplet dissociation. Since the interlayer has a larger dielectric constant than the ISC in this case, a more efficient screening of the Coulomb potential of the electron-hole pair might also contribute.
Another kind of interlayer is the terminating surface layer. For example, TiCl 4 is a common passivation reagent for TiO 2 that can be used for trap-state engineering [399]. For elemental ISCs like Si, the simplest termination consists of hydrogen. In comparison, more electronegative groups are expected to give rise to an increase of the work function by up to more than 2 eV [159,160]. Such a large work function increase has indeed been realized experimentally by exposing H-Si to NF 3 [136]. In the same work, it was shown that this polar surface termination allows tuning the ELA at the hybrid interface in much the same way as discussed below for organic interlayers.
We note that quantum dots can also be used as acceptor interlayers in energy cascades that funnel charges [400] or excitons [315] from an OSC.

Organic interlayers
Organic interlayers can be used to tune the work function of ISCs within a range of several eV via the introduction of surface dipoles. This can be achieved via induced dipoles generated by donor or acceptor molecules that drive a preferential electron rearrangement between the ISC bulk and its surface [139,249,251,379]. The molecular modifiers are typically prepared via vacuum-deposition, but can also be grown from solution [401]. Alternatively, molecules that bear a permanent dipole moment and are preferentially aligned, e.g. by anchoring to the ISC surface, also give rise to a surface dipole [52,258,345,402,403].
The change in work function can be used to tune the ELA at the hybrid interface, as illustrated in figure 17(c), 22, and 23. This was observed, e.g. at organic/ZnO [249,330,345,406] organic/H-Si [136,380], and organic/TiO 2 [258,[407][408][409] interfaces. A change in ELA can be used to increase (decrease) ΔG as well as the ρ overlap between the OSC and ISC for electron injection (recombination), thereby enhancing exciton dissociation [410]. As can be seen in figure 24, this modification enhances the short circuit current (J SC ) but the same change in ELA also reduces the V OC [405,406]. This follows from the fact that equation (1b) depends on the interface energy gap ΔE OI . In a related case, the opposite approach was employed for a nonsolar cell structure. Here, ELA tuning toward smaller ΔG was used to enhance the exciton transfer yield at the ZnO-organic interface and attributed to the fact that the transferred excitons undergo dissociation at the modified hybrid interface with significantly reduced rates [330]. In another study, ELA tuning via molecular acceptors did not notably improve the triplet exciton dissociation at the organic/H-Si interface [136]. C 60 was one interlayer material examined in this study. In this case, the interface dipole at the ISC/interlayer interface was compensated by a second, unintentional interface dipole at the interlayer/organic interface. This example shows the importance of experimentally confirming an ELA tuning scheme. Co-adsorption of modifier and dye was used in DSSCs to obtain SAM blends [411]. In a recent example, this strategy facilitated an increase of the cell efficiency from 11.1% to 12.9% [412].
Molecular acceptors not only induce ground state charge transfer from the substrate, but can also accept photogenerated charges from the OSC. This resembles closely the standard operation mode of conventional all-organic solar cells. In hybrid solar cells, this can be exploited to assist exciton harvest via an energy cascade, as further discussed in relation to surface-bonded dyes below. Here we mention that for a zinc phthalocyanine  (b) One-dimensional slices of the data in (a) at the indicated times. The S 1 and T 1 positions were determined by visual inspection and comparison with the time series measured for another Tc film reported previously [404]. (c) (Left) One-particle energy-level diagrams for the unmodified Tc/H-Si(111) interface. The energy levels for the Tc HOMO and H-Si(111) VB were derived from the data shown in (b) and figure 6. Unoccupied states do not have an IE and the Tc LUMO is indicated only for completeness. (Center) and (right) Energy-level diagrams with shifted Tc/c-Si ELA due to the introduction of interface dipoles of two different magnitudes. Reprinted with permission from [136]. Copyright (2020) American Chemical Society.
(ZnPc) interlayer, theoretical and experimental results revealed that ZnPc molecules in direct contact with a ZnO surface insert new energy levels due to a strong ZnPc/ZnO coupling, and that electron injection from these new energy levels of ZnPc into ZnO is highly efficient [406]. This possibly contributes to the observed increase in photocurrent and an overall increase in power conversion efficiency.
Similar to inorganic interlayers, organic interlayers can reduce the charge carrier recombination at the hybrid interface via a decreased coupling between organic and ISC [406]. However, an organic interlayer does not necessarily prevent the formation of CT states across the hybrid, as evident from CT emission and absorption observed for SAM-modified organic/ZnO interfaces [345]. In this case it was found that the change in ELA induced by the dipole moment of the SAM induces a corresponding change in CT state energy, as shown in figure 22(b).
IET relies on electronically coupling the donor and acceptor. Covalently attaching the organic chromophore to the semiconductor surface allows to systematically control H (as introduced in equation (2)) via the linker. The unoccupied DOS fraction on the linker calculated for the isolated molecule was found to be highly correlated with the initial IET efficiency [344]. It has been shown that increasing the distance between the electron-donating group of the organic and the electron accepting TiO 2 by introducing saturated CH 2 units in the linker (thereby decreasing H) considerably decreases the rate of electron injection in DSSCs [413]. π-conjugated linkers were found to result in much higher IET rates than when using saturated CH 2 units [414,415]. Given a strong-enough dye-inorganic coupling, light absorption and charge transfer can occur as a coherent process [339].
Notably, different H's for initial IET and interfacial recombination can be achieved by rational design. This is because the initial IET between organic and inorganic constituents proceeds via formerly unoccupied states (the LUMOs of the neutral molecule, LUMO, LUMO+1, . . . , and the CB of the ISC) while the final state of the interfacial recombination is the HOMO of the molecular cation or the VB of the ISC. Therefore, it is possible to manipulate the spatial distribution of the electronic states so as to permit the forward IET while shutting off the recombination channel. Figure 25 shows a clear correlation between charge recombination half-time and spatial separation of the dye cation HOMO orbital from the TiO 2 surface. From this, an exponential dependence of the electronic transfer rate constant on the separation can be derived [337].
In a similar way, asymmetric substitution of otherwise symmetric molecules can induce directionality of the excited state energy and improve injection into the ISC [416], as shown for zinc tetraphenylporphyrin (ZnTPP) in figure 26. A common structural motif for asymmetric organic sensitizers is a donor-π-bridgeacceptor configuration.
The attachment of dye molecules is also important for quantum dots. If quantum dots act as absorbers, transfer from the quantum dots to the OSC can be used to extract their energy. Triplet exciton transfer to attached organic molecules was observed for quantum dots made, e.g. from PbS [417], Si [68], and perovskites [70,418,419]. Exciton dissociation was observed, e.g. for CdSe quantum dots via hole transfer to phenothiazine [420]. For the other chemical structures, the reader is referred to [337]. Reproduced from [337] with permission of The Royal Society of Chemistry. Surface-bound dyes can be used as mediators for exciton harvest from an organic overlayer by acting as electron acceptors [405]. The investigation of a series of acceptor interlayers substantiated that a suitable energy cascade can improve exciton dissociation at organic/TiO 2 interfaces [421]. However, the same scheme does not work for ZnO, which the authors ascribe to the more pronounced and unfavorable interface dipole in that case. A different study that employed diethylphosphonate-benzodiathiazole terminated thiophene oligomers of different length highlighted the role of the energetics across the molecularly modified hybrid interface and suggested that both a large energy offset and a suitable energy cascade facilitate exciton dissociation and prevent  recombination [422]. Energy transfer from P3HT to a surface-bonded dye, followed by electron injection from the dye into TiO 2 represents an alternative pathway for exciton harvest [423].
In the electronic structure section we discussed ground-state charge transfer between organic layers and inorganic surface states. This suggests that organic interlayers can passivate ISC surface states. Another passivation mechanism is the chemical attachment of stable moieties that saturate surface radicals and prevent the formation of detrimental chemical species like sub-oxides. While much work has been done on inorganic-inorganic passivation schemes as a result of the growth of the semiconductor industry, inorganic-organic passivation schemes are gaining importance in the optoelectronic fields. Examples for chemisorbed and physisorbed organic interlayers that were found to passivate ZnO defects are mentioned in [424]. Other cases where surface-bonded organic molecules were used for surface passivation comprise perovskite solar cells [311,425,426] and Si [427,428].

Role of interface morphology changes
The same organic layers in contact with the pristine or modified ISC surface may give rise to very different OSC morphologies (molecular orientation, disorder, growth mode, preferred crystal phases). As shown in figure 27, for Tc on H-Si(111), a 2 nm C 60 interlayer was found to dramatically increase the roughness and reduce the crystallinity of the Tc films for the same deposition conditions. Accordingly, devices with C 60 interlayers of more than one monolayer essentially do not work [136]. Organic interlayers can also improve the crystallinity of the OSC film, as shown in figure 28: C 60 grows as an amorphous film when deposited directly on sapphire, whereas a pentacene interlayer gives rise to a highly crystalline C 60 film [429]. Similarly, diindenoperylenecovered SiO 2 gives rise to enhanced C 60 crystallinity compared to the pristine substrate [430].
These effects are important, because film morphology and polymorphism sensitively influence, e.g. exciton dissociation [148,189,374,375,431,432] and singlet fission [191,[433][434][435]. The molecular orientation at the hybrid interface was shown to influence exciton quenching rates [436] and the lifetime of charge-separated states [437]. The interlayer-induced change of OSC morphology can be exploited in rational ways to better control the interface structure [18,438,439].

Conclusions & outlook
Progress in the field of hybrid solar cells is a mixed bag. On the one hand, the use of OSCs in combination with ISC absorbers to realize purely electronic functions like charge-selective transport and surface passivation has come a long way, and reaping benefits like reduced production cost and increased stability seems within reach. In particular, using OSCs as dopant free hetero-contacts holds enormous potential to ease fabrication of c-Si solar cells with the potential extra benefit of increased efficiency.
On the other hand, harvesting excitons at hybrid interfaces still remains very inefficient, aside from in DSSCs and quantum dot solar cells. Initially, the higher degree of charge delocalization in ISCs compared to OSCs was one appealing aspect of hybrid solar cells, since the associated more efficient screening seems to promise efficient exciton dissociation at hybrid interfaces based on experimental results [440] and semiclassical modeling [151]. However, experiments [440] and calculations [362,[441][442][443] also indicate that the ISC surface can act as a trap for charge carriers, and thus hinder exciton dissociation at hybrid interfaces. A similar trapping is not observed, e.g. for the corresponding organic/fullerene interface [362,440]. More importantly, experiments find that at organic heterojunctions, often no significant driving force is needed to efficiently split thermalized charge transfer states. This means that the interface with an ISC is not, in fact, required for efficient exciton harvest from an OSC.
However, exciton-harvesting hybrid interfaces have other potential advantages. For example, if singlet fission in an excitonic SC can be tapped via exciton harvesting at a hybrid interface, this opens up a non-tandem cell approach to exceeding the Shockley-Queisser single junction limit with silicon, the quintessential solar cell material. Notably in such a device concept, current matching between absorber layers is not required. However, the hybrid interface in such a device has steep requirements, and must support efficient exciton dissociation or injection, excellent passivation of both semiconductor materials, and in the case of a heterojunction devices, high selectivity charge extraction from silicon. These commingled functions pose a challenge to device designers.
In attempting to review this diverse field, we have discussed prevalent concepts used to describe relevant interfaces in hybrid solar cell technologies. We have presented a comprehensive discussion of mechanisms that govern the electronic structure and excited state dynamics at these interfaces, as well as strategies to control them. However, this is still a far cry from a complete description of the complex property-function relationships in hybrid solar cells. This is due partly to the sheer breadth of the field, which prohibits an all-encompassing review, but also to the inherent complexity of those interfaces that do indeed function successfully [444]. The following is a shortlist of these complexity-compounding phenomena: • 3D morphologies, such as bulk heterojunctions, the nanostructured interface of a DSSC, and the surfaces of thin film perovskites, imply complications for investigations, both theoretical and experimental. Theoretical modeling that can account for local quantum effects while capturing the relevant spatial and temporal dimensions is being attempted, but whether the employed structures are representative of the real 3D arrangements still needs to be established. Experiments are often performed on planar interfaces that enable controlled interface preparation and straightforward data interpretation. However, these might miss the role of the surface curvature [441] as well as low-dimensional features like defects that gain in importance for 3D structures, highlighting the necessity of techniques capable of probing buried interfaces. • Research on organic bulk-heterojunctions indicate that gradients in crystallinity at the interface, as well as phases that feature intermixing, facilitate energy cascades and charge delocalization. Such mesoscopic structures are challenging to rationally implement at hybrid interfaces. • Calculations for organic hetero-interfaces suggest the need to control the electrostatic potential landscape beyond interface dipoles [374,375,432]. • The triplet exciton harvest at the interface between Tc and HfO x N y -covered Si might be assisted by transient interface dipoles that arise from accumulation of photo-generated charges [397]. This suggests that ISC surface states do not just act as recombination centers and sensitively influence the ELA and surface band bending, but can also give rise to transient local fields. These considerations indicate that controlling both the morphological microstructure as well as the electrostatic potential in space and time is required to realize an efficient exciton harvest. To get a handle on this complexity requires experimental methods capable of probing all of these properties and processes.
A number of methods are available to study solar cells on different spatial and temporal scales [28]. However, different effects might contribute to what can be observed by the different methods [445]. This hampers directly linking derived parameters, like energies from optical and (inverse) photoelectron spectroscopy [194,345]. Probing both ground state and excited state energies with the same technique should alleviate this issue, as laid out in reference [281].
For resolving the excited state dynamics, our current understanding is largely based on all-optical pump-probe spectroscopy, which can routinely achieve sub-ps resolution. In comparison, time-resolved x-ray spectroscopy is still underdeveloped. However, spectroscopy in the x-ray regime has a number of important advantages over all-optical characterization techniques. It provides elemental, chemical, and surface sensitivity, allowing selective probing of the materials building the hybrid interfaces, rather than the less informative bulk. In addition, the high spatial resolving power of x-ray microcopy (a few 10 nm) allows to selectively probe single grains of the typically multi-crystalline organic films, and can thus help disentangle some of the structural complexity at the interfaces. Laser pump and x-ray probe spectroscopy at storage ring synchrotron sources can readily resolve ns to μs dynamics from interface state charging [446] and surface photovoltage build-up [447]. Recently, pioneering experiments have resolved triplet signatures in the sub-100 ps timescale [448]. By not relying on a pump-probe scheme but instead using the core-hole lifetime as an internal clock, IET can be resolved on a sub-10 fs timescale [449,450]. Given the advent of next-generation synchrotron sources, free electron lasers, and high harmonic generation, as well as advances in computational methods, the future of excited state dynamics investigations at hybrid interfaces looks promising.