Electron and Hole Transport Layers : Their Use in Inverted Bulk Heterojunction Polymer Solar Cells

Bulk heterojunction polymer solar cells (BHJ PSCs) are very promising organic-based devices for low-cost solar energy conversion, compatible with roll-to-roll or general printing methods for mass production. Nevertheless, to date, many issues should still be addressed, one of these being the poor stability in ambient conditions. One elegant way to overcome such an issue is the so-called “inverted” BHJ PSC, a device geometry in which the charge collection is reverted in comparison with the standard geometry device, i.e., the electrons are collected by the bottom electrode and the holes by the top electrode (in contact with air). This reverted geometry allows one to use a high work function top metal electrode, like silver or gold (thus avoiding its fast oxidation and degradation), and eliminates the need of a polymeric hole transport layer, typically of an acidic nature, on top of the transparent metal oxide bottom electrode. Moreover, this geometry is fully compatible with standard roll-to-roll manufacturing in air and is less demanding for a good post-production encapsulation process. To date, the external power conversion efficiencies of the inverted devices are generally comparable to their standard analogues, once both the electron transport layer and the hole transport layer are fully optimized for the particular device. Here, the most recent results on this particular optimization process will be reviewed, and a general outlook regarding the inverted BHJ PSC will be depicted.


Introduction
In 1977, Shirakawa, Louis, MacDiarmid, Chiang and Heeger reported on their discovery of electrically conductive polymers [1] (for which, in 2000, the Nobel Prize in Chemistry was awarded jointly to Heeger, MacDiarmid and Shirakawa).From that milestone, a great deal of research activity on conjugated polymer-based optoelectronics has been developed all over the world.To date, organic-based light-emitting diodes (OLED) have become commercially available with good performance, and polymer-based photovoltaics has reached high efficiency: over 9% in lab-scale devices [2,3].The exploiting of the peculiar properties of conjugated polymers in the photovoltaic (PV) field is more recent than in light emitting devices, and still, a lot of unsolved questions must be addressed in order to really develop a commercial route for polymer PV [4].Following the ideas and the experimental results in the pioneering works on small organic molecule-based photovoltaics [5], in 1993, Sariciftci reported on the evidence of a conjugated polymer/fullerene heterojunction bilayer solar cell [6,7].Soon after that, the first polymeric bulkheterojunction (BHJ) PV device was described by Yu et al. [8].In 1995, the new soluble fullerene derivatives [9] allowed the boosting of device performances [10].Starting from those first works, a lot of efforts have been made in order to increase the device power conversion efficiency (PCE) and stability, mainly working on synthesizing new polymers to be used in the active layer and on optimizing the device structure and geometry.The most commonly used device structure for a BHJ organic solar cell comprises a conductive transparent substrate-typically metal oxides, like indium tin oxide (ITO) or fluorine-doped tin oxide (FTO) on glass or plastic substrates-covered by a thin hole conducting layer, such as the polymer, poly(3,4-ethylenedioxythiophene)-poly(styrenesulfonate) (PEDOT:PSS), over which, the active layer is deposited, typically from a solution by means of spin coating, doctor blading, ink-jet printing, spray coating, etc. [11][12][13][14].Finally, a thin metal layer is realized (Al, Ca/Al, LiF/Al, for instance), usually by thermal evaporation.This structure has been used in the realization of devices with very good performance, reaching a PCE of around 9% [3].Despite these good efficiency reports, which, anyway, are far from being general and reproducible [4], such a structure suffers from several drawbacks: due to the required energy level alignment among all the cell components, a low work function metal electrode must be used on top of the device for the electron extraction, like Al.However, low work function metals undergo very fast oxidation when exposed to air, losing their conductivity, suddenly turning a working device into a faulty one.Moreover, the PEDOT:PSS is acidic in nature and is thus detrimental to the underlying metal oxide layer, which undergoes fast degradation [15,16].Finally, it has been widely reported that polymer/fullerene (as well as polymer/polymer blends [17]) are characterized by a stratified composition (vertical phase separation) during the film formation [18], with the fullerene phase-that is, the electron-conducting phase-mainly concentrated at the bottom of the film and the polymer phase-the hole-conducting one-concentrated mainly at the top of the film.Thus, the film vertical phase structure is opposite to the ideal one, where the electron-conducting phase must face the top low-work function electrode and the hole-conductive phase must face the bottom high work function electrode.All these problems can be avoided by reversing the collection process, i.e., collecting the holes by the top electrode and the electrons by the bottom electrode.In such a structure, the top metal electrode would be a high work function electrode, like silver or gold (Ag or Au), thus eliminating the oxidation problem, while the bottom electrode should be a transparent electron conducting layer (oxides, like zinc oxides (ZnO) or titanium oxide (TiO x ) are good examples), eliminating the problem of the acidic PEDOT:PSS on ITO or FTO.This inverted structure was first exploited in the fabrication of organic light emitting diodes, also initially referred to as an "upside-down" structure [19]; then, it started to be exploited also in the PV field [20,21].
After these pioneering works, it has been demonstrated that the inverted structure allows one to reach performances even better than the standard one [22,23].Despite this, the research on inverted polymer solar cells (PSCs) is a very small fraction of the total research in the PSC field.As of November 14, 2013, a simple search on the ISI Web of Knowledge using as key words "bulk heterojunction solar cell" (excluding the word "inverted") gives about 7400 papers, while when the key words "inverted" AND "bulk heterojunction solar cell" are used, only 207 results are shown, that is less than 3% of the published papers.The situation does not change much if the search string is not limited to bulk heterojunction, but to the more generic "polymer solar cell" (24,700 results) to be compared with the very small number of 514 results when searching for "inverted" AND "polymer solar cell" (less than 2% of the published papers on polymer solar cell), nor would a great variation of this ratio come out if the search were limited to the last two years.This was already pointed out by Krebs and coworkers very recently [4] (they estimated that less than 10% of published results were focused on the inverted structure).This is quite surprising, since the polymeric bulk heterojunction inverted solar cell (BHJ ISC) has by far better stability over time than the standard geometry devices [24].
Among these relative few papers, the review articles are very limited.The most recent and complete, by Zhang et al., published in 2011 [25], however, reviews general inverted organic devices, not focusing particularly on the polymer bulk heterojunction concept; while in the one by Hau et al. [24], published in 2010, there is no general focus on the electrode interfaces.
The aim of this review is to present a comprehensive description of the most recent results on the improvement in the performances of hole transport layers (HTLs) and electron transport layers (ETLs) used in BHJ ISC in order to tentatively give a strong baseline for further research directions and improvements.On the contrary, this paper will not consider: (1) small molecule-based devices; (2) bi-layered active materials or multi-layered active material devices; and (3) tandem devices; for all of which, there is an excellent scientific literature.
The paper will be structured as follows: first, a very brief and general overview of the working principles of the bulk heterojunction solar cell is given; then, a comparison between the standard geometry device and the inverted one is sketched, highlighting the pros and cons of both, following a detailed review of the various materials and strategies for optimizing the hole transport layer (HTL) and the electron transport layer (ETL); finally, a summary of the major findings is given.

The Bulk Heterojunction Solar Cell: Working Principles
The working principle of a polymer BHJ solar cell (see Figure 1) could barely be summarized as the creation of an exciton in the active layer, due to light absorption, and the separation of this exciton into two separate charge carriers at the interfaces between the species that constitute the active layer (typically, a binary blend of a polymer and a fullerene or two polymers, which act as the donor phase and the acceptor phase), with subsequent collection by the electrodes [26].Despite their apparent simplicity, all of these steps must obey very strict limitations in order to be as efficient as one would need.First of all, the generated excitons must hop between the molecules reaching an interface between the two phases before recombining (radiatively or non-radiatively).This means that the two phases should be mixed in an optimal structure, with phase domains usually in the order of 10-30 nm (the average exciton diffusion length in polymers [26]).Then, the position of the energy levels at the interface must be favorable for a fast exciton dissociation followed by charge separation (i.e., the electron in the acceptor phase and the hole in the donor phase without successive recombination [26]).After that, the charges must travel inside the respective phases, reaching the collecting electrodes again without a charge recombination: at this point, the energetic level structure at the electrode interfaces plays a fundamental role, ideally the interface being an ohmic contact [26].The standard geometry device is sketched in the upper part of Figure 2; as already mentioned in the Introduction, it typically consists of a transparent bottom electrode for the hole collection, a thin layer of active material and a top metal electrode for the electron collection.In the inverted structure (bottom of Figure 2), the role of the electrodes is swapped; thus, the electrons are collected by the bottom transparent electrode and the holes by the top metal electrode.This reversed collection implies that the work function of the top electrode must be high enough in order to match the donor highest occupied molecular orbital (HOMO) energy level, and the work function of the bottom electrode must be low enough in order to match the acceptor lowest unoccupied molecular orbital (LUMO) energy level.If the requirement of the top metal electrode can be simply fulfilled by selecting high work function metals, like gold or silver, the right selection of the bottom electrode is more tricky.In fact, the most used transparent metal oxides, like ITO or FTO, possess high work functions that do not match well with the LUMO level of the acceptors.The matching of the energy levels is obtained by modifying the bottom electrode with the deposition of a thin layer of suitable electron conducting (hole blocking) materials, like, for instance, ZnO or TiO x , which are, moreover, transparent to visible light.The effects of this electrode capsizing are multiple, and the working properties of the devices are strongly influenced, as will be described in the following sections.

Light Absorption and Electromagnetic Field Distribution: Comparison between Standard and Inverted Structure
The processes of exciton diffusion, exciton dissociation and charge separation, as well as charge recombination in the bulk active layer depend on the active layer properties and are thus expected to be the same for both of the structures.On the contrary, standard and inverted structures present strong differences in the electromagnetic field distribution inside the device.
The main criterion that should be fulfilled in every PV device is that the solar light should be well absorbed by the active layer (that means that, ideally, the main part of the solar spectrum reaching the Earth's surface should be harvested).This is not a well-solved issue yet in organic PV, since each polymer absorbs a narrow range of visible light, and typically, the ones used in BHJ solar cells have no or low absorption in the low energy side of the solar spectrum.Since the open circuit voltage (V oc ) of a BHJ PSC is believed to be determined mainly by the donor HOMO-acceptor LUMO energy gap [26], the exploitation of low band-gap donors typically decreases the V oc , and a trade-off must be reached between good light harvesting and good electrical parameters.Although this is a problem that affects both the standard geometry and the inverted one, it has been demonstrated that the inverted configuration better harvests the incoming light, due to a more favorable electromagnetic field distribution inside the active layer.In fact, Ameri et al. [22] considered a poly-3-hexyl thiophene (P3HT):phenyl-C61-butyric acid methyl ester (PCBM) standard geometry solar cell and an inverted analogue; they found that the inverted device performed about 15% better than the standard one.In order to understand the reasons underlying the better performance, they optically modeled the devices.They found that in the inverted structure, the number of absorbed photons by the active layer was increased, thus resulting in an increased number of charge carriers, thanks to the use of a TiO x bottom layer, which is transparent to visible light, instead of a PEDOT:PSS layer, which absorbs about 20% of the incoming light.With these observations, they were able to explain roughly a 10% increase in performance.The remaining 5% boost was instead ascribed to an unbalanced electron and hole mobilities that lead to a more efficient charge collection in the inverted structure than in the standard one.Indeed, assuming an exponential absorption profile, the charge carriers are generated mainly closer to the bottom electrode rather than to the top metal electrode; thus in the presence of unbalanced mobilities with the hole mobility lower than the electron one, a better charge collection is expected in the inverted configuration.What was still missing in the discussion was the findings that the P3HT:PCBM blend undergoes a vertical phase separation during the process of film formation, resulting in a stratified structure in which a PCBM-rich layer is formed on the bottom of the film and a P3HT-rich layer is formed on the top [18] (see Figure 3).As the electrons mainly travel inside the PCBM phase and the holes mainly travel inside the P3HT phase towards the respective electrodes, the inverted configuration is the best one, being characterized by a self-assembled ETL (hole blocking) at the electron extraction electrode (bottom) and a HTL (electron-blocking) at the hole extraction electrode (top), thus both increasing the charge extraction efficiency and reducing the bimolecular recombination at the electrodes.
These findings were partially confirmed by Chen et al. [27], who modeled a P3HT:PCBM-based inverted device both optically and electrically.They concluded that if one considers only the optical aspects of the structure (i.e., the field distribution inside the active layer), the inverted structure should be better performing than the standard one for the major part of the considered active layer thicknesses.The standard device should, on the contrary, perform better when the active layer thickness is just the ideal one for constructive interference inside the film between the incoming radiation and the reflected one from the top metal electrode.Concerning the electrical aspects, the electrical behavior of the device was modeled as a function of the charge drift length, L, a parameter that accounts for the charge lifetime and mobility, being defined as the maximum length that a charge carrier can drift within its lifetime before any possible recombination.The value of L influences the electromagnetic field distribution and the optical modulation inside the active layer for each thickness.They showed that the inverted structure is better than the standard one, only if the active layer thickness is comparable with L, while the standard geometry performs better if the active layer thickness is greater than L.This is due to the fact that L induces an "effective area" inside the active layer where the absorbed photons are efficiently transformed into charge carriers, while photons absorbed outside this area are mainly lost, not contributing to the photocurrent.Since the position of this effective area depends on the different charge transport properties peculiar to the active material, the optical field modulation inside the device together with the position of the effective area determine the performance differences between standard and inverted geometry.
This prediction, however, was slightly in contrast with the results of Aziz, Li and coworkers [28], who showed a remarkable difference between standard and inverted devices, especially when analyzing the short circuit current density, J sc .In both devices, the J sc increased with increasing the thickness of the active layer, but in their case, the J sc in the standard device started to decrease for an active layer thicker than 300 nm, while the J sc in the inverted one remained constant.They explained their results with a reduced recombination mechanism in the inverted structure (coherently with the conclusions of Ameri et al. [22]) and a reduction of the series resistance in the inverted device.This apparent contrast could be, however, explained with the differences in the device structure considering the nature, properties and thicknesses of all the constituent layers, which strongly influence the optical and electrical properties [27].
Even if those reports were based on a P3HT:PCBM active layer, many other research groups reported improved performances of the inverted structure compared to the standard analogue when other active materials were used.For instance, Ma et al. reported a boost in performances for an inverted device based on a blend of PBDT-12/PyT2 [23].
Overall, from both the optical and electrical analyses given by those reports, it appears clear that the inverted structure has many advantages with respect to the standard one, at least when thin layers are used, which is actually the standard in realizing organic-based optoelectronic devices.Given the general behavior of the devices, in the next sections, a description of the most recent results for the electrode interfaces is given.

ETLs and HTLs: Recent Developments
In both the standard and inverted devices, the active material plays the same role, that is, in short, to absorb photons and to convert them into free charge carriers (the influence of optical and electrical parameters on this process has been briefly sketched in the previous section).In order to achieve good performances, however, it is mandatory that a BHJ comprises both an HTL and an ETL.
A list of requirements that a good ETL, as well as a good HTL should fulfill (among which are transparency, good electrical properties and chemical stability) is given in a recent review on the synthesis methods of metal oxides by Litzov and Brabec [29]; although that review focuses on metal oxides only, the general list applies to every material to be used as the electron or hole-transport layer in a BHJ ISC.Here, a summary on the most recent results about different class of materials for the realization of the various ETLs and HTLs in BHJ ISCs is given.

Bottom Electrode: Electron Collection
The main requirements for the bottom electrode are a good electron transport property and, when the light is collected through it, transparency to the solar radiation.This limits the choice to a few materials, among which are the well-known and widely-used ZnO and TiO x [29] (which have also a good electronic level matching with the LUMOs of most of the polymers used in PV), other less used materials, like Cs 2 CO 3 [30], and new polymeric materials, like poly(2-(dimethylamino)ethyl methacrylate) (PDMAEMA) [31] and PFEN-Hg [32].The various realization methods of transparent metal oxides have been described in great detail in a very recent review [29].In the present work, the attention will be focused on the effects of the ETL and HTL layer properties on the device performances.

ZnO
Very thin layers of ZnO are easily realized by means of several deposition techniques, like sol-gel [33], spray-coating [34] and nanoparticle (NP) deposition [35].Most of the time, the as-deposited film needs a process of annealing usually at high temperatures in order to optimize the crystalline structure [36], but also, low annealing temperature processes have been recently reported [35,37,38].
Considering the P3HT:PCBM-based devices just as a prototypical PV system, the overall reported power conversion efficiency using ZnO as the ETL spans from about 2% to about 4% [29,35,39].This quite wide range is due to the differences in several factors, including the chosen HTL, the parameters involved in the active blend preparation and optimization and the properties of the ZnO ETL layer.In this section, the attention will be focused on the possible influences of the ETL layer on the device performances with various active layers.
One of the most important parameters that determines the overall performances of the device is the morphology (and consequently, the roughness) of the ZnO layer.Despite the several methods developed to deposit a thin ZnO layer on substrates, there is no "elective" method to obtain the optimal morphology; each method can afford the best results, once all the deposition parameters have been optimized.Yu et al. [40] showed that the power conversion efficiency of a P3HT:PCBM-based ISC was increased from 2.08% to 2.88% by increasing the surface roughness of the ZnO layer.Since the ZnO film realized by the sol-gel technique is affected by three processes during the annealing [41] (solvent evaporation, zinc acetate decomposition and crystallization), they changed the surface roughness of the layer by simply varying the annealing rate of the as-deposited (by sol-gel) ZnO layer, namely a slow annealing rate of 9 °C/min and a fast one of 56 °C/min up to 300 °C.They found both an increase in J sc and in FF for the slowly annealed sample.The boost in the J sc was explained by the enhanced light absorption, due to the rougher ZnO surface, which should provide an efficient light trapping mechanism: once being scattered by the ZnO layer, the light is reflected back by the upper Ag electrode in multiple directions, thus increasing the path length inside the active layer.The higher FF was explained considering the better crystallization of the ZnO, due to the slow annealing rate, which resulted in a more effective electron extraction and hole blocking (it is known that FF is strongly related to the charge extraction efficiency at the interfaces [42]).Dhibi et al. [43] reported the same conclusions in a PCDTBT:PC70BM-based ISC: the absorption of the blend cast onto the ITO/ZnO (sol-gel) substrate was higher than that of the one cast onto the bare ITO substrate.This effect was ascribed both to the light scattering mechanism, as well as to the larger surface area between the ZnO and the active blend.However, they noted that a too pronounced roughness of the sol-gel-derived ZnO films could be, on the other hand, detrimental to the overall device performance when a very thin active layer is needed (that is the case of the PCDTBT:PC70BM blend), due to the rising-up of a high contact resistance and a large leakage current [44].Moreover, they demonstrated that also the annealing temperature is a key factor in order to determine the morphology of the layer: they annealed the ZnO films at three different temperatures, namely 130 °C, 150 °C and 200 °C, and characterized them by photoluminescence and Raman measurements, showing that the 150 °C annealed film was affected by better crystallization, which could enhance the electron extraction and mobility, as well as suppress the leakage current into the device.Vijila and coworkers [45] showed that the annealing temperature greatly affects the overall crystalline morphology of the ZnO layer.They demonstrated that a ZnO film annealed at 240 °C was characterized by a higher crystallinity compared to a similar layer, but annealed at 160 °C.The device incorporating the 240 °C annealed ZnO layer performed better (with a 40% higher PCE) than the device realized with the 160 °C annealed ZnO layer.They explained these results, observing that the ZnO layer annealed at 240 °C was characterized by a lower trap depth, thus reflecting in higher charge mobility and better ohmic contact and, consequently, higher J sc and V oc .
These results indicate that both the annealing temperature and the annealing rate should be taken into account in order to optimize both the crystalline structure and the surface roughness of sol-gel-derived ZnO films, looking for a trade-off between increased light scattering by the surface roughness (thus enhancing the active layer absorption, and, hence, the J sc , and increasing the FF, due to the higher interface contact and charge extraction efficiency) and the crystal structure (better charge transport).
A further interesting contribution on this topic is the work by Hu et al. [46], showing that the ZnO morphology can also depend on the layer thickness.They realized ZnO layers by the metal organic chemical vapor deposition (MOCVD) technique, finding that the ZnO layer thickness did not influence the energy level alignment at the interfaces, but it had a strong impact on the crystal structure and, hence, on the electrical properties.In particular, with the increase of the thickness, the morphology of the ZnO layer changed from a pure smooth sphere-like structure for a thickness of <100 nm to a rough pyramid-like structure starting from 100 nm and above.They indirectly confirmed the above picture of the need for a trade-off between roughness and crystal structure; in fact, they showed that initially, an increase of the ZnO layer thickness, and hence, of the surface roughness, was beneficial for ISC performances (with PCE increasing from 1.60% for 40 nm of ZnO to 2.93% for 80 nm ZnO), while the PCE started to decrease for the thicker ZnO layer (ISCs were based on the P3HT:PCBM active layer).
It is noteworthy that the above results are commonly characterized by the use of no annealing or a relative low post deposition annealing temperature in order to obtain a good overall film morphology.This is quite different from the most commonly used high temperature post deposition annealing process; for example, Park et al. [47] used a temperature of 450 °C for annealing a sol-gel-derived ZnO layer to be used as the ETL in a P3HT:PCBM-and a P3HT:PCBM:polyoxyethylene tridecyl ether (PTE)-based device; they demonstrated by XRD that the so-annealed film showed two strong crystalline peaks at 30.3 °(100) and 34.3 °(002). .Nanostructured ZnO layers also appear to be promising as efficient ETL in ISC.Simple ZnO nanoparticles (NPs) are easily deposited onto substrates by spin coating [35,48] or slot-die coating [49]: Alstrup et al. [49] realized layers of ZnO starting from an aluminum-doped ZnO NP ink using a roll-to-roll compatible method, the slot-die coating.Their results showed that a very efficient ISC can be realized using low-cost techniques that are feasible for large area device realization, once each layer deposition's parameter has been optimized.Regarding the ZnO layer, they also showed that the overall PCE of the final device is correlated with the thickness of the ZnO layer itself.Ibrahem et al. [35] prepared thin layers of ZnO NPs using a simple method that avoids the use of any surfactants in order to stabilize the ZnO particles in solutions, simply dispersing ZnO powder in pure ethylene glycol followed by high energy ball milling at room temperature in a batch-type grinder.They showed that the ZnO particle size decreased with the increasing of the grinding time down to an average particle diameter of 25 nm.The NPs, after grinding, were stable in solution without any precipitation, obtaining very uniform and thin films upon spin coating and final device PCEs over 3.5%.Our group realized all polymer ISC using a thin layer (about 80 nm) of ZnO on fluorine-doped tin oxide (FTO) by simply spin coating a commercial (Sigma Aldrich, St. Louis, MO, USA) aqueous dispersion of ZnO NPs with an average diameter less than 100 nm [50], without any post deposition annealing process.The layer morphology resulted in a very uniform sphere-like texture (see Figure 4) without any cracking or pinholes, with an average surface roughness of 4.53 nm.Thus, it is demonstrated that ZnO NP dispersion in water or common organic solvents can be a convenient and roll-to-roll compatible method for obtaining good ZnO ETLs for efficient inverted solar cells.More complex structures have also been reported, like vertically aligned, one-dimensional nanostructures [51] and bi-dimensional nanowalls [52].The latter, in particular, should theoretically perform better than one-dimensional structures, since it provides a higher contact surface between the ZnO and the active layer (Figure 5).Liang et al. [52], however, showed that when compared with ZnO nanorods, the NWs structure does not perform clearly better.They tried to explain their results, observing that the overall growth condition and, hence, the surface chemistry of the two structures are different, thus implying a possible different mechanism of charge recombination and charge transfer through the surface.Nevertheless, their report is quite recent, and there is no systematic optimization of the ZnO NW layer parameters, such as the thickness, etching process and interface modification, in order to improve and promote the diffusion of the active blend inside the nanostructures, thus leaving a large amount of room for future improvement.Electrospun ZnO nanowires (NWs) were incorporated by Vijila et al. [53] in a sol-gel-derived ZnO layer to be used as the ETL in ISCs.With the NW plantation, the overall device performance was increased compared to a control device without NW.This was due to a higher FF (both lower series resistance and higher shunt resistance), as well as a beneficial effect on the active layer phase separation morphology (which was reflected in a higher carrier lifetime and a reduced charge recombination).
A proper doping of the ZnO layer [54,55] is another possible route to improve the charge collection efficiency and to reduce the charge recombination, due to charge trapping by defects or oxygen ions [29].Hu et al., in the already cited work [46], reported an improved overall performance of an ISC when the intrinsic ZnO (i-ZnO) was doped with boron during the MOCVD process, in particular showing that the J sc increased up to 11.31 mA cm −2 when the diborane flow rate was kept at 3 sccm.They explained the results considering a better charge collection at the electrodes (the average resistivity of the ZnO layer decreased from 4.04 × 10 5 Ω/sq of the i-ZnO to 2.46 × 10 3 Ω/sq of the best boron-doped ZnO).Alstrup et al. [49] doped ZnO with Al in order to improve the conductivity and quality of the ZnO film deposited by the slot-die coating method.Lithium fluoride (LiF) has been used for a long time as an ultra-thin layer underlying the top Al electrode in standard configuration bulk heterojunction solar cells, since it improves device performances [56] and device stability [57].Chang et al. [58], on the contrary, used it for doping a sol-gel-derived ZnO ETL to be used in ISCs based on a P3HT:PCBM active blend, demonstrating an improvement of the overall PCE of the device from 2.9% to 3.3%.Optimized LiF-doped ZnO showed a very high electron mobility up to 11.2 cm 2 V −1 s −1 .They concluded that Li atoms occupy interstitial sites, producing electrons, which results in higher electron mobility.At the same time, F atoms occupied in substitutional oxygen sites reduce the oxygen vacancies in the structure and lower the absorption in the visible region.Their conclusions are indeed supported by the fact that the better performance of the device is correlated with an increase in the J sc , all the other cell parameters being almost unchanged.Finally, Du et al. [59] showed that UV irradiation of a sol-gel-derived ZnO layer used as the ETL in PBDTTT-C:PC 71 BM-based ISC improved the device performance, due to the increase of the layer conductivity following the reduction of hole traps on ZnO.This is clearly reflected in the higher FF related to a decrease of the device series resistance.The drawback of UV irradiation consisted, however, in a gradual degradation of the polymeric active blend; so, they concluded that, practically, a trade-off between the early stage advantage of UV irradiation and long-term disadvantage due to polymer degradation should be considered in ISCs comprising a ZnO ETL.
Despite the several techniques and the different processes used to date for optimizing ZnO properties, it appears that the final device performance does not depend as much on the technique used for realizing the ZnO layer, but rather on the inner properties of the layer itself, such as the roughness and crystalline structure, both almost independent of the particular deposition method, once optimized.

TiO x
Although porous crystalline TiO 2 is the standard charge transport layer used in the so-called dye-sensitized solar cell (Grätzel cell) [60], it is, on the contrary, not well suited as the ETL in BHJ ISCs, due to the difficult and tricky deposition of a uniform and homogeneous layer on standard ITO substrates [61,62].Nevertheless, it has been used by Peng et al. [63] in ISCs, showing, in particular, that the overall power conversion efficiency of the device strongly depends on the TiO 2 crystalline structure.When the film is characterized by a well-defined morphology (the homogeneity of nanoparticle shapes and distribution) and a maximum crystal size of about 50 nm, the interface between the TiO 2 and the active material is characterized by an optimal morphology and interaction, thus favoring both exciton dissociation and charge separation, with improved FF (which means less charge recombination and improved charge transport through the inorganic layer).
However, amorphous TiO x (x ≤ 2) is as good as crystalline TiO 2 as the ETL, once optimized [64].Moreover, it is as good as ZnO as the ETL in ISCs, since it is characterized by high electron mobility and high transparency in the visible range [65].Hadipour et al. [66] showed that it is possible to realize a TiO x ETL layer at room temperature from solution without any post deposition annealing process.They demonstrated that such a layer was as good as calcium in extracting electrons from a bulk heterojunction device, confirming that TiO x possesses the right energy level to act efficiently as the ETL in combination with several active blends.Finally, they demonstrated that the properties of the deposited TiO x remain the same when the deposition process occurs both in air and in a glove box, which is important for the possible implementation of their deposition process into roll-to-roll manufacturing.
Inverted devices that utilize amorphous or crystalline TiO x (usually deposited by the sol-gel method [29]) are characterized by the same range of efficiencies seen in the case of the ZnO [29,66,67].
The performances of devices incorporating an amorphous TiO x layer can be boosted with proper tuning of the layer's physical and chemical properties.Choulis and co-workers [68] demonstrated that the PCE of inverted devices can be improved once an ultra-thin layer of polyoxyethylene tridecyl ether (PTE) is inserted as the interlayer between the ITO substrate and the TiO x layer, a mechanism that apparently resembles the improved electron collection in standard devices when an ultra-thin LiF layer is inserted between the active layer and the top aluminum electrode.They showed no variation in the V oc between samples with and without the PTE interlayer, thus suggesting that no recombination or morphological behavior differences should characterize the devices.An increase in the J sc was also indicative of reduced leakage current and series resistance.In order to explain such a result, they take into consideration the possible variation of the hydrophilicity of the TiO x surface, due to the PTE interlayer underneath.Indeed, they found that the TiO x layer becomes actually more hydrophilic when coated on top of the PTE layer, instead of directly on top of the ITO substrate, thus enhancing the vertical phase separation of the active blend with more of the fullerene phase (relatively more hydrophilic than P3HT [69]) on the bottom and, hence, the electron transport and collection [26,70].Their results demonstrate how it is possible to clearly improve the PCE of ISCs simply by adding a suitable interlayer on top of the ITO electrode.
They also demonstrated that, even when processed in air, the TiO x layer can be unaffected by the so-called "light-soaking effect" [71,72].This is an unusual "s-shaped" current-voltage curve, translating into an extremely low FF, due to the properties of the ITO/TiO x interface [73], which sometimes arises for air-processed layers.This effect, however, can be suppressed once the TiO x layer is exposed to UV light, as shown by Kim et al. [73].They proposed a possible mechanism that is responsible for this process: the work function of the TiO x slightly changes upon UV irradiation, due to the increase of its charge carrier (electrons) density following the suppression of O − 2 traps (adsorbed by TiO x during realization processes in air), filled with the holes generated by the absorption of the UV radiation.This change in the work function (increased electron density) induces a rearrangement of the Fermi level at the interface (the decreased energy barrier between ITO and TiO x , see Figure 6) in a direction favorable to a boost in charge transport and collection, resulting in a strong decrease of the series resistance of the device and improved J sc .This mechanism has been exploited to explain also the same findings upon UV irradiation of ZnO layers [59].Another elegant way to improve the charge transport and collection and to reduce the charge recombination is by enhancing the vertical phase separation by realizing self-assembled fullerene monolayers (SAM) on top of the TiO x film [67,74].Although this seems to be a very good strategy, it is necessary to be careful in the choice of the SAMs: in fact, the deposited layer must resist the subsequent active layer deposition process from solution, which means that the layer should not present at the end of the process localized defects due to the desorption of molecules [75].Hsu and coworkers [67] joined together the concepts of self-assembling structures and cross-linking, which were separately exploited previously [76][77][78][79], in order to overcome the cited drawback of SAMs, synthesizing a new fullerene-derivative, namely the bis(2-(trichlorosilyl)propyl)malonate C60 (TSMC).This new compound was able to both cross-link and self-assemble on top of the TiO x layer, without the need for annealing or UV irradiation.The modified TiO x layer was used in ISCs based on both PC 61 BM and PC 71 BM, showing a clear improvement of the device parameters, which was explained also by an enhanced vertical phase separation of the active blend by the SAM.Their approach seems to be quite interesting in the framework of facile and practical processes for mass production.Quite in parallel, Kim and coworkers [74] synthesized another fullerene derivative able to generate fullerene-based self-assembled monolayers (FSAMs) onto TiO x to be used as ETL in ISCs, namely bis-4-(2-ethylhexyloxy)- [6,6]-phenyl C61 butyric acid (bis-p-EHO-PCBA).They showed that the bis-FSAM was responsible for the sensible reduction of the series resistance compared to that of the TiO x layer alone, with an increase both in J sc and FF in polymer:fullerene-based ISC.This confirms the better wetting properties with the organic active layer (responsible, at the end, for the better device performances) when the TiO x layer is covered with fullerene-based SAMs.

Cs 2 CO 3
The use of Cs 2 CO 3 as the ETL is more recent: in 2004, a report from the Canon group [80] described its use in organic light emitting diodes.In 2006, Yang et al. [30] showed that such a material could be conveniently used as the ETL in BHJ ISCs: a P3HT:PCBM-based ISC exhibited a good PCE of 2.25% using Cs 2 CO 3 in combination with V 2 O 5 as the HTL.The interesting result they showed lies in the fact that all the device parameters remained almost unchanged when the Cs 2 CO 3 layer was realized by thermal evaporation or by spin coating from solution.They proved that the ETL provided an ohmic contact favorable to the device performance.Liao et al. [81] showed that the work function of Cs 2 CO 3 can be tuned by annealing the layer at relatively low temperatures up to 200 °C.This shift in the work function is mainly dependent on the intrinsic properties of the layer more than on some chemical reaction with the underlying ITO layer.They also showed a direct correlation between the annealing temperature and both the device PCE and the Cs 2 CO 3 surface contact angle, indicating a transition of the ETL from a more hydrophilic-type to a more hydrophobic-type layer, which is beneficial for the subsequent good deposition of the active polymer blend.They explained this work function shift by analyzing the results of XPS experiment, concluding that upon annealing, the Cs 2 CO 3 decomposes into Cs 2 O oxide doped with Cs 2 O 2 , thus lowering the WF, with a better matching with the acceptor LUMO and, consequently, the interface resistance.
Xin et al. [82] reported on the effect of UV-ozone treatment of Cs 2 CO 3 layers on the general performance of BHJ ISCs.They showed that the V oc monotonously decreased with increasing UV-ozone treatment time, while the J sc showed first an increase and then a decrease.The best device performance they obtained was thus attributed to the enhancement of carrier transport and collection by the Cs 2 CO 3 layer [83,84].However, while the effects of UV treatment on ZnO or TiOx layers have been widely studied, the precise mechanism underlying the UV-ozone treatment on the Cs 2 CO 3 layer is still unclear.
Other really interesting and recent results were reported by Barbot and coworkers [83]: they exploited the good ETL properties of both Cs 2 CO 3 and fullerenes, realizing a Cs 2 CO 3 -doped C60 ETL interlayer.In a previous work, they found that the electrical conductivity of undoped C60 was improved by eight orders of magnitude upon doping with Cs 2 CO 3 for a molar ratio of 1:0.35 [85].Indeed, they showed that the best BHJ ISC based on a P3HT:PCBM active blend was exactly the one that incorporated an ETL with such a molar concentration ratio, with an increase of V oc and a decrease of the series resistance.
Finally, Liu et al. [86] addressed one drawback of the Cs 2 CO 3 , that is its poor hole blocking ability.By incorporating graphene quantum dots (GQDs) into the Cs 2 CO 3 , they were able to improve the PCE of a BHJ ISC based on P3HT:PCBM from 2.57% for a reference device with a Cs 2 CO 3 ETL only to 3.17% for a device with a GQDs-Cs 2 CO 3 ETL.Furthermore, in this case, the better performance was correlated with a better match between the involved energy levels of the ETL and the PCBM, due to a shift of the ETL work function upon incorporation of the GQDs.This minimizes the energy losses at the interface almost without affecting the overall light absorption in the active blend.Moreover, they proposed that an additional exciton dissociation channel could occur at the P3HT/GQD-Cs 2 CO 3 interface, facilitating the charge separation.However this could also mean that using such an ETL would affect the vertical phase separation in the active blend, with a non-negligible P3HT phase content on the bottom side of the device that could enhance the charge recombination; this is actually a point that would need to be addressed.

Nb 2 O 5
Nb 2 O 5 has been used as the ETL in ISCs [87].A drawback of this metal oxide is that the energy level of its conduction band (about 3.7 eV [88]) is higher than the LUMO level of PCBM (about 3.8 eV [89]); thus, it should not perform as well as ETL in fullerene-based devices.Nevertheless, Cao et al. [87] showed that a PCE comparable to that obtained on a similar device, but incorporating a T iO 2 layer as the ETL, is achievable (2.7% compared to 2.8%).They explained this counterintuitive effect with an efficient tunneling process from the PCBM LUMO to the N b 2 O 5 conduction band.Especially when using this metal oxide, great care should be taken in obtaining the optimal layer thickness: a non-optimized layer thickness can even reduce the device PCE close to 0% [87].

Novel Alternatives
Although some of the previous materials dominate the field of electron transport layers in inverted bulk heterojunction solar cells, the surfaces of metal oxides possess hydroxyl groups that are the cause of possible charge trapping [24].Thus, new materials and concepts have been recently developed as promising alternatives.In particular, towards the development of an all-organic device, the use of polymeric material as ETLs seems to be quite interesting.Huang and coworkers [32] synthesized an amino-functionalized conjugated metallopolymer, namely PFEN-Hg, to be used as an efficient electron transport layer in inverted polymer solar cells.Their polymer was characterized by most of the requirements of an ETL, which are orthogonal solvent processability, transparency to visible light, good electron transport and hole blocking properties, good morphology and effective ITO work function tuning.With the incorporation of such an ETL, they were able to reach about 9% PCE on an ISC based on PTB7:PC 71 BM as the active material.
Vijila and coworkers [31] used as the ETL in a P3HT:PCBM-based ISC a solution processable cationic biopolymer, poly(2-(dimethylamino)ethyl methacrylate) (PDMAEMA).They compared the device performance with that of a reference device using a standard ZnO ETL layer, finding comparable PCE for a fresh device and a better stability of the polymeric ETL after eight weeks, with a PCE greater than that of the reference device.
Fang et al. [90] developed an ETL based on the hybrid material, CdS/2,9-dimethyl-4,7-diphenyl-1,10-phenanthroline (CdS-BCP), which also fulfills most of the cited requirements for good ETLs.They exploited it in PTB7:PC 71 BMor P3HT:PC 60 BM-based ISCs, showing an improvement in PCE compared to single CdS or BCP-based devices, with improved stability over time.
Jeon and coworkers [91] took advantage of the good electron transport properties of carbon nanotubes (CNTs) by improving their charge selectivity, coating them with an ultra-thin TiO x layer deposited by atomic layer deposition.The TiO x coating, due to its good hole-blocking properties, prevents the recombination of holes and electrons at the metallic part of the CNTs during the charge transport-collection process.This results in a final PCE improvement of about 30% compared to the bare CNTs-based ETL.Although being interesting results, the effectiveness of the TiO x -coated CNTs strongly depends on the quality (homogeneity and thickness) of the coating, which, if not optimized, can lead to a large amount of leakage current, due to the coating defects.
Kohlstadt et al. [92] demonstrated good PCE in ITO-free PTB7:PC 71 BM-based ISC when a metal-metal oxide multilayer is used as the electrode/ETL.They realized by DC magnetron sputtering a sequence of 25 nm of Al-doped ZnO (AZO), 9 nm of silver and, again, 20 nm of AZO.With such a sequential structure, they demonstrated a PCE of 6.1%, almost comparable with the PCE of a standard geometry device with an ITO/PEDOT:PSS substrate (which was reported to be 6.9%).They pointed out that the possible drawback of such a multilayered structure is an unfavorable field distribution inside the active layer and a lower absorption, issues that still need to be addressed.
Oo et al. [93] described a very good ETL, namely a zinc tin oxide (ZTO) layer used on top of FTO in P3HT:PCBM-based ISCs.They explained the better performance of the device incorporating the ZTO layer considering the reduced recombination at the interface, the improved electron transport and the hole blocking properties of the ZTO compared to standard ETL.

ETL: Conclusions
As a brief and overall general conclusion on this section, it appears quite clear that optimized (and possibly doped) ZnO and TiOx represent, to date, the best bet for realizing well performing BHJ ISCs, based on the reported averaged PCE of devices and on their fulfillment of most of the points in the "decalogue" established by Litzov and Brabec [29].Cs 2 CO 3 , despite its excellent electron transport properties, suffers from poor hole blocking properties [86], while Nb 2 O 5 possesses a probably unfavorable energy level diagram to be used generically as the ETL with systems that are not based on PCBM.The other novel alternatives seem really interesting and promising, especially the polymeric ETLs to be fully exploited in the framework of an all-polymer device realized by the roll-to-roll technique, but the results are too recent in order to be reasonably able to infer some general conclusion over the near future.

Top Electrode: Hole Collection
In BHJ ISCs, the top metal electrode is usually either gold or silver, due to their intrinsic stability in air.However, Ag is the most suitable top electrode, since its slow oxidation shifts its work function from about 4.3 eV to about 5 eV [94], favoring hole extraction.It has no absorption in the visible range contrary to the evaporated Au layer (slight absorption in the visible) and a higher reflectivity [95].In fact, Chen et al. [96] showed that, ceteris paribus, a device with an Ag top metal electrode performed better compared to Au or Al.This is also due to the better matching of the WF of Au (5.1 eV [97]) and the oxidized Ag (around 5.0 eV [94]) with the HOMO levels of the commonly used polymer donors, compared to Al (4.28 eV [97]).
Anyway, in order to obtain good performances from an ISC, the introduction of an HTL layer between the organic active layer and the top metal electrode is mandatory; otherwise, the device would show very poor performance or even no operation at all [98].To date, the most commonly used HTLs in BHJ ISCs are polymers, like PEDOT:PSS, or metal oxides, like MoO 3 , V 2 O 5 , WO 3 , NiO.

PEDOT:PSS
PEDOT:PSS is the most widely used HTL in standard geometry BHJ solar cells, due to its high work function (matching the HOMO level of commonly used donor polymers well), high transparency in the visible range (higher than 80%), good electrical conductivity and the ability to reduce the ITO surface roughness [99], while increasing its work function.However, PEDOT:PSS is highly hydrophilic; thus, bad film morphology and worse electrical properties have been observed when deposited as the HTL onto the hydrophobic organic layers in inverted devices [100,101].Nevertheless, it is widely used in BHJ ISCs despite this evident drawback.In order to address the wettability problem on hydrophobic surfaces, several strategies have been reported.First of all, the hydrophilic nature of PEDOT:PSS can be reduced by adding proper additives to the polymer dispersion, like isopropyl alcohol [102], or surfactants, like Zonyl FS-300 [103] and Triton X-100 [104], as well as surfactant mixtures [68].Moon et al. [105] added two amphiphilic surfactants, namely Sulfynol 104 EG or PA (dissolved in ethylene glycol or isopropyl alcohol respectively), showing a clear improvement of the PEDOT:PSS layer morphology.This was reflected in an improved P3HT:PCBM-based device performance of about 60% compared to a reference device incorporating the PEDOT:PSS layer without surfactants.The improvement was mainly due to a higher J sc and a higher FF.Indeed, they showed that upon adding the surfactant, the contact angle between the PEDOT:PSS and the underlying organic layer was decreased from 104°to about 53°.This means a clearly better film forming property, resulting in higher shunt resistance (less charge recombination at the electrode) and a lower series resistance (better charge extraction).They explained the better wettability of the additive-added PEDOT:PSS on the organic layer, observing that the used surfactants consist of a hydrophobic alkyl backbone and a hydrophilic hydroxyl group; thus, a van der Waals-type interaction between the hydrophobic alkyl backbone and the hydrophobic P3HT:PCBM surface occurs, making the organic layer hydrophilic with subsequent good PEDOT:PSS film forming properties on top.Ho and coworkers [106] proposed a novel fluorosurfactants to be used as a substitute to the Zonyl FS-300, since the latter usually needs to be used together with isopropyl alcohol in order to get the best results.The proposed Capstone Dupont FS-31 can be used alone as an additive to the PEDOT:PSS, giving very good results in terms of overall device performance when a relative concentration of about 5% is used.They also showed that the additive concentration in the PEDOT:PSS dispersion greatly affects the final film forming, thus reflecting in better or worse device performances.A concentration below 4% results in a non-uniform layer, due to the still low wettability onto the organic layer.On the contrary, if the concentration is too high (above 8%), the presence of excess surfactant reduces the charge conduction pathway through the PEDOT molecules, thus resulting in poor charge collection.Moreover, they showed that the surfactant also promotes a phase segregation between PEDOT and PSS, thus resulting in a well-formed PEDOT conduction pathway, while favoring a good physical contact with the underlying organic layer.
Another strategy to improve the PEDOT:PSS adhesion onto the organic layer is to exploit different deposition techniques rather than spin coating.Gupta and coworkers [107] obtained efficient P3HT:PCBM-based inverted OSCs by realizing the top PEDOT:PSS contact by stamp-transfer lamination.They showed that this process does not induce significant losses on the open circuit voltage, nor does it increase the overall resistance of the devices.
However, despite the progresses towards a better adhesion of PEDOT:PSS on top of organic layers, the acidic/hygroscopic nature of the compound is still detrimental for the top-side metal electrode.Indeed, Kim and coworkers clearly showed [108] that PEDOT:PSS is able to promote the corrosion and degradation of a thin layer of Ag as typically used in ISCs.They demonstrated that the hygroscopic nature of PEDOT:PSS promotes a very fast absorption of ambient water inside the layer.This absorbed water becomes corrosive, due to the acidic nature of the PEDOT:PSS, thus slowly corroding the silver layer, first promoting grain boundary grooves, then leading to grain segregation, forming separated particles.They, however, observed that such a process could be slow in a real device configuration and processing, although in the long term, it would cause strong degradation of the performances.
Even if, to date, PEDOT:PSS is still the most used HTL in inverted BHJ PSCs [29], in order to avoid all these drawbacks, in several works, PEDOT:PSS has been replaced by other efficient HTLs, like the following ones.

MoO 3
A thin HTL of molybdenum trioxide (MoO 3 ) greatly improves the performance of both inverted and conventional OLEDs [109][110][111].Early in 2008, Chen and coworkers [96] showed that MoO 3 increased the overall efficiency of a BHJ ISC based on both a P3HT:PCBM active layer and a nanocrystalline TiO 2 ETL.They reported that the V oc of the device changed only slightly when the top metal electrode was varied from Ag to Au to Al, while the optimal HTL thickness depends on the chosen metal.In particular, with Al as top electrode, an higher thickness of the evaporated MoO 3 layer is required in order to obtain the same performances of devices using Ag or Au.This was attributed to the chemical reaction between Al and MoO 3 during the evaporation process, forming a MoO 2 species, due to the reduction of MoO 3 .
The dependence of the overall efficiency of devices on the MoO 3 thickness has been also detailed by Zheng et al. [112] and by Wantz and coworkers [113].Zheng et al. showed that the PIFTBT8:PC 71 BM-based device parameters remained almost unchanged when the thickness of the MoO 3 layer was in the range of 10-20 nm, while a drop in both V oc and FF occurred when the thickness was decreased to 5 nm.They correlated such a behavior with the shunt and the series resistance, observing that a decrease of the thickness of the MoO 3 layer was followed by both an increase of the series resistance and a decrease of the shunt resistance.Moreover, increasing the thickness, the V oc changed from 1.00 V to 1.04 V, and the FF varied between 41.9% and 50%.Wantz et al. showed that the overall performance of a P3HT:PCBM-based device was only slightly affected by a variation of the MoO 3 thickness from 6 nm to 22 nm.Interestingly, they also demonstrated how the post-processing thermal treatment on the device, usually employed to improve the device performance [114], is, on the contrary, detrimental in BHJ ISCs when MoO 3 is used as the HTL.In fact, they observed that after 10 min at 170 °C, the cohesion between the MoO 3 layer and the silver electrode was lost, and the layer itself evolved into an alloy of silver and molybdenum oxide, with a diffusion of silver ions and oxygen inside the active layer.The thinner the MoO 3 layer, the faster and even worse was this process.It should then be concluded that an inverted BHJ incorporating an MoO 3 layer should not undergo a thermal treatment after the completion of the device.However, since in the case of P3HT:PCBM, the annealing process promotes the vertical phase separation in a way favorable to the inverted device working principle [115], such an annealing process should be performed prior to the deposition of both the HTL and the top metal electrode.Moreover, providing that the post-processing annealing treatment is avoided for the completed device, Gupta and coworkers [116] showed that MoO 3 acts as a barrier to the diffusion of metal ions and oxygen inside the active layer, thus preventing fast device degradation.The role of the environment on the effects of the thermal treatment of thin layers of MoO 3 was studied by Vijila and coworkers [117].When the complete device incorporating the MoO 3 layer as the HTL is annealed in vacuum, the overall stability and performance are higher compared to a similar device annealed in a nitrogen atmosphere.The difference in the PCE was increased upon aging, with the vacuum annealed device retaining 80% of its initial PCE value after 30 days compared to the 50% of the device annealed in nitrogen atmosphere.They explained this behavior observing that the metal oxide layer annealed in vacuum is characterized by an oxygen deficient structure that is the origin of an enhanced hole-conducting properties.
Jiang et al. [118] demonstrated that an interesting way to improve and tune the electrical and the optical properties of a MoO 3 layer can be by realizing a composite film with Au.They showed that a MoO 3 -Au (30%) layer is characterized by very good electrical and optical parameters.They exploited this layer as the HTL in a conventional geometry solar cell, in substitution of PEDOT:PSS.Although there is no report on the use of such a modified MoO 3 layer in inverted devices, one can speculate that such a layer could be conveniently exploited also in this type of solar cell.

WO 3
As shown earlier by Dong et al. [97], the high work function of WO 3 (−4.8eV) greatly enhances the hole transport and collection at the WO 3 /Ag interface.They showed that the optimum thickness for WO 3 in a P3HT:PCBM-based device is in the range between five and 10 nm.Lower values would results in poor coverage of the film, while higher values would increase the series resistance, thus lowering the overall device performance.They also showed that the V oc of the devices were substantially unaffected by the choice of the top metal electrode (Au, Ag or Al).Kwon et al. [119] compared the performances of MoO 3 and WO 3 inverted devices based on several active layers.They showed that in all cases, the device incorporating a WO 3 HTL layer (5 nm optimum thickness) performed slightly better than those with a MoO 3 HTL layer (10 nm optimum thickness).They explained this better behavior of the WO 3 layer with: (1) a better film surface morphology (a reduced average roughness measured by atomic force microscopy); (2) a better measured conductivity; and (3) better electron blocking and hole transport properties, due to the higher energy barrier between the donor LUMO and the HTL conduction band and the better matching between the donor HOMO level and valence band of the HTL.
Brabec et al. [120] demonstrated that a low temperature (80 °C) solution-processed nanoparticle WO 3 HTL was able to perform as well as PEDOT:PSS in a standard geometry device and to give very good performances on an inverted architecture.The solution was alcohol-based and free of any surfactant, and the realized HTL does not need any oxygen-plasma treatment, which is usually required to remove the dispersing agent in the original NP dispersion [121] and to improve its HTL properties.They concluded that the low temperature solution-processed WO 3 layer can be easily and conveniently exploited in flexible solar cell production processes.

V 2 O 5 and NiO
Even if V 2 O 5 has been used as the HTL in inverted devices,-its high toxicity [119] makes it not very suitable for easy and low cost device processing in ambient conditions.Moreover Riedl et al. [121] showed that care should be taken when realizing the V 2 O 5 HTL by the sol-gel process.They compared the performance of P3HT:PCBM-based devices with evaporated V 2 O 5 HTL or sol-gel-processed V 2 O 5 HTL, observing that the 10 nm-thick layer of the sol-gel-processed V 2 O 5 sensibly affects the working device, depending on the initial solution concentration and spin coating speed.In particular, a low concentration of the precursor solution (1:150) spin coated at slow speed formed a film that affected the device positively, showing parameters comparable to those obtained with an evaporated V 2 O 5 layer.On the contrary, a film with the same thickness, but obtained from a higher concentrated precursor solution (1:70), spin coated at a higher speed, greatly lowered the overall performance of the device.They excluded that such a strange behavior could be attributed to some effect of the solvent onto the active layer, while showing through XPS measurements that the higher concentrated precursor solution leads to a deep penetration of vanadium inside the active layer, up to more than 100 nm.This penetration of the precursor into the active layer (that was by far less evident in the case of the evaporated V 2 O 5 ) could damage the P3HT chains and/or lead to the quenching of excitons in the donor phase.This penetration was also indirectly confirmed by Dauskardt and coworkers [115] when comparing the roll-to-roll-processed inverted BHJ PSC incorporating both PEDOT:PSS or V 2 O 5 HTLs.
NiO has been also used as a good HTL in BHJ ISC.NiO possesses a valence band and a conduction band at −5.0 eV and −1.7 eV, respectively, thus acting as a good HTL and a good electron blocking layer in combination with a P3HT:PCBM, as shown by Chen and coworkers [122].The optimal HTL thickness in their case was under 2 nm, above which the device performance starts to decrease.They attributed this behavior to the intrinsic electronic properties of the NiO layer that was deposited in a high vacuum by thermal evaporation, thus affecting the oxygen concentration in the realized layer.A thin layer was also deposited from solution by Lee et al. [123], by simply dissolving NiO powder into various solvents (the best was found to be isopropyl alcohol) and spin coating it onto the P3HT:PCBM active layer.Upon optimization of the active layer thickness, they were able to obtain about 3% PCE, comparable with a reference device incorporating as the HTL a thin layer of PEDOT:PSS.The solution processable NiO layer could be compatible with roll-to-roll manufacturing in air.

Novel Alternatives
A solution processable mixture of V 2 O 5 and MoO 3 was exploited by Huang and coworkers [124] as an efficient HTL in a P3HT:PCBM-based inverted solar cell.They showed that the mixed layer causes a significant improvement of the device parameters compared to control devices with V 2 O 5 or WO 3 HTL alone.They explained these results with the efficient hole transport, due to the well matched donor HOMO-valence band of V 2 O 5 and, concomitantly, to the electron extraction from the top silver electrode through the conduction band of the WO 3 .This complementary mechanism positively affects the overall PCE.Solution-processable triindoles, namely triazatruxene (TAT) and N-trimethyltriindole (TMTI), with tailored HOMO and LUMO levels, in order to be used as the HTL in inverted devices, were presented by Ma et al. [125].They showed an improvement in the J sc and in the overall PCE of a P3HT:PCBM-based inverted device when a thin layer of TAT was spin coated onto the active layer.
Furthermore, graphene oxide has been exploited as the HTL in BHJ PSCs by Dai and coworker [126].Indeed, graphene is a bidimensional carbon structure that possesses excellent electronic and mechanical properties [127].Through the neutralization of the -COOH groups of the graphene oxide with Cs 2 CO 3 , they were able to tune the electronic properties of the material, giving both excellent ETL or HTL properties.They exploited these interesting characteristics, showing very good device performances in P3HT:PCBM-based devices with a PCE exceeding 3%.

HTLs: Conclusions
Averaging the general results, it should be concluded that the best HTL for BHJ ISCs should be a thin layer of either MoO 3 or WO 3 , due to their intrinsic good electrical and optical properties, their ability to prevent oxygen and metal ion diffusion into the active layer and the almost unchanged device parameters upon slight variation of their optimum thickness.On the contrary, PEDOT:PSS has to be avoided, due to its evident drawbacks, as mentioned.Dauskardt and coworkers [115] demonstrated how the poor adhesion of PEDOT:PSS onto the active layer results in a general loss of device performance in a roll-toroll processed device, due to the thermomechanical stress that every "real-life" device would undergo.
However, it appears clear (see, for example, the comparison between MoO 3 and WO 3 above) that the key factor in choosing the best HTL is the trade-off between the electrical/morphological properties and the energetics involved at the interfaces in the particular device.This means that for each particular donor-acceptor system, the matching between the HOMO and LUMO levels of the organic phases and the valence band and the conduction band of the HTL should be optimum.Not only the HOMO level of the donor should match with the valence band of the HTL, but also the latter should have a conduction band far from the active layers' LUMO level, otherwise resulting in an enhanced charge recombination and poor electron blocking behavior [112].Finally, when dealing with metal-oxides, it should be also considered (in the proper choosing of the best HTL for the particular system) that the work function of those materials strongly depends on the concentration of defects [129], with a decrease of the WF if oxygen is removed from the material, and an increase in the WF is observed with oxygen addition.Interestingly, Xie and Choy [130] developed a synthesis route for developing low-temperature solution-processed MoO 3 or V 2 O 5 HTL with the ability to control the oxygen vacancies, thus, in the end, controlling the final WF.
Finally, one key motivation in selecting the proper HTL should be the compatibility of its deposition process with standard roll-to-roll or printing techniques for the realization of complete devices in ambient conditions, without the need for vacuum processes, like thermal evaporation.Thus, the choice should come down to solution processable HTLs, which is no longer a limitation, since solution-processed WO 3 and MoO 3 HTLs have been exploited both in conventional and inverted devices, showing good overall performances [120,[131][132][133][134]. Furthermore, the very interesting polymeric novel alternatives cited above could be promising.

Device Stability: A Brief Comment
Considering the environmental stability in the working condition of a BHJ ISC as the key parameter, which is the best ETL and HTL choice?Actually it is not possible, to date, to give a precise answer.A detailed discussion on this issue is far beyond the scope of this paper: in one sentence, the stability of BHJ ISCs is still a matter of debate [29].For example, it is not clear how and to what extent the particular process used for preparing metal oxide layers can affect the final device stability [135], nor which is the particular "best" device structure [136].One comprehensive study on the stability of several ETL/HTL combinations has recently been performed by Voroshazi and coworkers [137].They demonstrated that ZnO-and TiO x -based devices were comparably stable in air for over 1000 h, while the analyzed organic ETLs were clearly affected by a lower stability over time.On the HTL side, they surprisingly showed that a PEDOT:PSS-based device was more stable than a MoO 3 -based device.However, Vijila and coworkers [117] demonstrated that the annealing condition of a MoO3-based device can greatly affect the final device stability.They demonstrated that a vacuum annealed device was characterized by a 10% increased stability compared to a device annealed in a nitrogen atmosphere, as discussed in the previous section.Furthermore, Sun et al. [38] showed that a device incorporating a thin layer of MoO 3 as the HTL retained 70% of its initial PCE after 30 days.
Finally, some recent reports on lifecycle analysis (LCA) and energy payback time (EPT) have been published by several groups [138][139][140].From the reported results, it can be concluded that it is quite difficult to assess a precise standard, to date, in order to definitively compare different kinds of devices, materials and procedures.Even the same devices are characterized by some variation in their working parameters when measured by different laboratories [136].A definition of a precise standard in this field should be particularly important to evaluate how much the (sometimes low) increase in the PCE obtained with a particular process/optimization of the ETLs and/or the HTLs is convenient from an economical point of view (considering the energy and time consumption needed for the layer realization/optimization).

General Conclusions
As a general conclusion, the BHJ ISC appears to be a very promising device structure to be exploited in the mass production of organic solar cells.Aside from the development and optimization of novel active materials, like novel acceptors or low band-gap polymer donors, the correct optimization and choice of the HTL together with the ETL play a fundamental role in order to get the best performance from a device.Many parameters, however, need to be traded off, as has been described.In particular, a necessary trade-off between better HTL and ETL electrical properties and overall device optical properties must be accomplished.Moreover, the right HTL, as well as the right ETL must be chosen, taking into account the energy level structure of the particular active blend used.The ideal optimum is the matching between the donor HOMO and the valence band of the HTL and between the acceptor LUMO and the conduction band of the ETL, together with a proper position of the HTL conduction band and the ETL valence band, in order to provide, at the same time, a good electron blocking or hole blocking behavior, respectively.The same features are also relevant in the case of polymeric or organic HTLs and ETLs (in this case, the matching should be between the HOMO and LUMO levels of the active layer and the HOMO and LUMO levels of the other layers).
As has been pointed out in the previous sections, the ETL and HTL performances can be affected by the realization parameters.Good control and knowledge of all these parameters should be necessary in order to get the maximum performance from a given system.
As a final remark, it is important to stress that novel polymeric alternatives could come in handy for bypassing the depicted criticality typical of the most used metal oxides, being also intrinsically compatible with low-cost device manufacturing processes.

Figure 1 .
Figure 1.Schematic basic working principle of a polymeric bulk heterojunction solar cell.

Figure 2 .
Figure 2. Schematic geometry for (top) a standard bulk heterojunction (BHJ) device and (bottom) an inverted BHJ device.The main component layers are sketched.ETL, electron transport layer; HTL, hole transport layer.

Figure 5 .
Figure 5. SEM images of (A) the ZnO seed layer, (B) the as-grown dense ZnO rods, (C) the etched ZnO nanowalls and the (D) cross-section of the nanowalls.Reprinted with permission from [52].

Figure 6 .
Figure 6.The energy diagram of the indium tin oxide (ITO)/TiO x interface before (a) and after (b) UV irradiation.Reprinted with permission from [73].