3.1.

Ultrathin Metal Layers

There is a trade-off between the transparency and conductivity for ultrathin metal layers. Though a sufficiently thin metal layer allows most of the light to be transmitted (high transparency), the discontinuity in film morphology results in poor electrical conductivity for ultrathin metal films. ^{41 , 106 , 107} Wilken et al. measured the sheet resistance of Au thin films on glass substrates with different nominal film thicknesses (from 10 to 60 nm). As shown in Fig. 3(a), while the experimental data (open symbols) qualitatively follow the predicted trend from the Fuchs-Sondheimer and Mayadas-Shatzkes models based on the scattering of free charge carriers, ¹⁰⁴ the Au films with thicknesses $>20\mathrm{nm}$ have low sheet resistances of $<5\mathrm{\Omega }/\square$ ; however, the sheet resistance dramatically increases to $>50\mathrm{\Omega }/\square$ for films $<10\mathrm{nm}$ . Shown in the inset of Fig. 3(a) is a comparison of transmittance of the Au films with different thicknesses. A maximum transmittance of $\sim 70\%$ at $\lambda =500\mathrm{nm}$ was obtained for a 7-nm-thick Au film, which was reduced to $<60\%$ for a 12-nm-thick film. In addition, the transmittance exhibits strong wavelength dependence. For example, the transmittance is only 50% at $\lambda =700\mathrm{nm}$ compared to the peak of 70% at $\lambda =500\mathrm{nm}$ for the 7-nm-thick Au film.

Fig. 3

(a) Sheet resistance of ultrathin Au films with various thicknesses (open dots are experimental results and solid lines are the prediction based on Fuchs-Sondheimer and Mayadas-Shatzkes mode). Inset: Transmittance of Au films with differet thicknesses. Reprinted with permission from Ref. 104. Copyright 2012 Elsevier. (b) A comparison of $\mathrm{J}-\mathrm{V}$ characteristics of CuPc/C60 devices with ITO and Ag thin films as bottom transparent electrodes. Reprinted with permission from Ref. 103. Copyright 2008 AIP Publishing LLC.

With a 9-nm-thick Ag layer as the bottom TE, O’Connor et al. successfully demonstrated a $\mathrm{CuPc}/{\mathrm{C}}_{60}$ bilayer OPV device, which exhibited a similar PV performance to that with an ITO electrode [Fig. 3(b)]. ¹⁰³ Although there have been a few studies aimed toward increasing the electrical conductivity without sacrificing the optical transparency, ^{46 , 108 , 109} such as applying a seed layer to improve the continuity of thin metal layers, ¹¹⁰ the relatively low optical transparency of thin metals is still one of the limitations in achieving high-performance devices with them. ^{111 , 112}

3.2.

Dielectric/Thin-Metal/Dielectric Electrodes

Since the first report by Fan et al. in 1974, ¹¹³ dielectric/thin-metal/dielectric (DMD) multilayer structures have been extensively studied for achieving highly transparent and conductive electrodes. ^{114 – 117} In these DMD electrodes, the intermediate thin metal layer provides the electrical conductance for the entire structure, whereas the two dielectric layers improve the overall transparency due to optical interference within the multilayer structure and the surface plasmonic effects at the two metal/dielectric interfaces. ^{44 , 118 – 120} As shown in Figs. 4(b) and 4(c), an ${\mathrm{MoO}}_x/\mathrm{Au}/{\mathrm{MoO}}_x$ structure shows a maximum transmittance of $T=80\%$ at $\lambda =650\mathrm{nm}$ (the corresponding reflectance is $R=5\%$ ), which is about two times higher than that of the bare Au layer ( $T=40\%$ and $R=25\%$ ). These experimental results qualitatively agree with the calculated transmittance and reflectance [open symbols in Fig. 4(b)] based on the transfer-matrix model with the optical constants listed in Fig. 4(a), which further confirms the transparency enhancement of the DMD structure as a result of the strong optical interference effects within the multilayer structure. ⁴⁴ In addition to ${\mathrm{MoO}}_x$ , ^{115 , 121} other choices for the dielectric layers include metal oxides (ZnO, ¹²² ${\mathrm{WO}}_3$ , ^{123 , 124} etc.), metal sulfides [e.g., ZnS (Refs. 125 and 126], and organic materials (bathocuproine). ¹²⁷ For example, Cho et al. applied the $\mathrm{ZnS}/\mathrm{Ag}/{\mathrm{WO}}_3$ electrode in OLEDs with tris(8-hydroxyquinolinato) aluminum ( ${\mathrm{Alq}}_3$ ) active layers, which exhibited near-Lambertian emission with comparable current efficiency to ITO-based reference devices. ¹²⁵ The inner dielectric layer in these DMD electrodes determines the carrier injection/collection polarity of the whole structure. Other than the rare and noble metal materials (e.g., Au and Ag), some other metals, such as Al and Cu, have also been utilized in the electrodes without affecting their electrical properties, which can further reduce the material cost for the entire structure. ¹²⁸ All the layers in the DMD structures are deposited by low-temperature deposition techniques, such as thermal vacuum deposition and spin coating; hence, these DMD electrodes can be utilized as top TEs with negligible damage to the underlying organic layers. ^{115 , 116 , 129} Wrzesniewski et al. successfully demonstrated a top-emitting OLED (TE-OLED) with a thermal evaporated ${\mathrm{MoO}}_x/\mathrm{Au}/{\mathrm{MoO}}_x$ trilayer structure as the top TE, which enabled a very high light extraction efficiency. ¹¹⁶ With transparent and close-packed microlens arrays attached to the DMD electrode, the light extraction efficiency as well as the external quantum efficiency of such TE-OLED devices was increased by up to 2.6 times compared to the ones without the microlens arrays. ¹¹⁶

Fig. 4

(a) Optical constants of thin layers of ${\mathrm{MoO}}_x$ and Au for transfer-matix simulations. (b) Comparison of the transmittance of ${\mathrm{MoO}}_x/\mathrm{Au}/{\mathrm{MoO}}_x$ trilayer electrode and a 15-nm-thick Au thin layers. (c) Comparison of the reflectance of the trilayer and Au thin layers, where solid symbols are experimental results and open symbols are calculated results. Reprinted with permission from Ref. 44. Copyright 2012 Elsevier.

Furthermore, such DMD multilayer electrodes show excellent mechanical flexibility in comparison to ITO electrodes because of the good ductility of metal layers, enabling their application in flexible optoelectronic devices. ¹²² Cao et al. successfully demonstrated a flexible poly(3-hexylthiophene):phenyl- ${\mathrm{C}}_{61}$ -butyric acid methyl ester (P3HT:PCBM) device with ${\mathrm{MoO}}_x/\mathrm{Au}/{\mathrm{MoO}}_x$ electrodes on poly(ethylene terephthalate) (PET) substrates [Fig. 5(a)], which shows a nearly identical photovoltaic performance as the similar device fabricated on a glass substrate [Fig. 5(b)]. Cao et al. further studied the mechanical flexibility of the flexible device using a simple bending test method. ⁴⁴ Only an $\sim 6\%$ drop in efficiency was observed after bending the device 500 times with a bending radius of 1.3 cm [Fig. 5(c)], which is far better than ITO electrodes, which can only sustain a few bending cycles with large bending angles. ⁴⁴

Fig. 5

(a) A photograph of a flexible poly(3-hexylthiophene):phenyl-C61-butyric acid methyl ester (P3HT:PCBM) device with an ${\mathrm{MoO}}_x/\mathrm{Au}/{\mathrm{MoO}}_x$ transparent electrode. (b) A comparison of $\mathrm{J}-\mathrm{V}$ characteristics of P3HT:PCBM devices on glass and plastic substrates. (c)  $\mathrm{J}-\mathrm{V}$ characteristics of devices under different bending conditions. Reprinted with permission from Ref. 44. Copyright 2012 Elsevier.

In summary, the multilayer structure with a thin metal layer sandwiched between two dielectric layers shows a similar electrical conductivity but an enhanced optical transparency compared to the metal-only semitransparent electrodes. With some advantages over ITO electrodes, such as better flexibility in flexible devices and less damage to organic layers when used as top TEs, these multilayer electrodes have been investigated as potential ITO substitutes. So far, sputter and thermal evaporation are the two major deposition methods for metal and dielectric layers in these DMD electrodes. Achieving the desired thickness for each layer in DMD structures may present some cost issues for fabricating low-cost optoelectronic devices. ¹³⁰ Moreover, transparency and conductivity of the DMD layers are strongly dependent on the thickness of the multiple thin layers, especially the intermediate metal layer, which is only $\sim 10\mathrm{nm}$ thick. Therefore, the uniformity of these ultrathin layers is vitally important for optimal properties; better deposition methods are required to make them compatible with large-area OPV and OLED modules.

3.3.

Metal Grids

Macroscopic metal grids have been widely utilized as the front contact in many inorganic-based solar cells, such as silicon solar cells and copper indium gallium selenide photovoltaic cells, where the metal fingers conduct away the generated current and the openings between the figures become the light windows for light absorption. ^{131 , 132} Here, microscopic metal grids composed of ordered metal lines that are $<1\mu \mathrm{m}$ wide are discussed and considered as a potential replacement for semitransparent continuous metal films for use as TEs in organic optoelectronic devices. ^{133 – 135} One of the advantages of metal grids is the increased transparency compared to continuous metal layers, where the gaps between metal lines are blank (100% in transparency) and contribute to the high transparency. The transmittance of metal grids is determined by the percentage of the blank areas on the film, which is related to the metal line width, spacing between lines, and the number of total lines within a unit area. The low sheet resistance of the metal grids can be achieved by using thick metal lines, 100 nm or even thicker, although this comes at the expense of film planarity. ⁵² However, the openings in the metal grids result in poor electrical contact with the organic materials, which typically have very low conductivity, thus leading to very inefficient charge injection and/or collection at the electrode/organic interface. Generally, a continuous conductive buffer layer (e.g., PEDOT:PSS) is required to planarize the surface of the metal grids and improve electrode contact with the active materials when used in optoelectronic devices. ¹³⁶

Conventional photolithography has been one of the predominant patterning methods used to fabricate such ordered metal grids, although it is somewhat complicated, costly, and incompatible with flexible substrates. ^{137 , 138} In 2007, Kang and Guo demonstrated a nanoimprint lithography (NIL) technique to achieve these electrodes on both rigid and flexible substrates. ^{139 , 140} As schematically illustrated in Fig. 6(a), a soft mold (e.g., polydimethylsiloxane) is first created with the replication of desired features from a hard template patterned using conventional photolithography. The soft mold is a reusable transfer media or stamp which transfers a thin metal layer deposited on its surface to the substrates by applying proper pressure and heat. ^{139 , 141} Different metal grids, such as Au, Cu, and Ag, have been successfully fabricated with this NIL method. ²³ These printed grids with 70-nm line width and 700 nm periods [Fig. 6(b)] showed an average transmittance of 84, 83, and 78% in the visible spectral range and sheet resistances of 24, 28, and $23\mathrm{\Omega }/\square$ for Au, Cu, and Ag electrodes, respectively [Fig. 6(c)]. With these metal grids as TEs, both ITO-free OPVs and OLEDs were successfully demonstrated with comparable device performance as the ITO-reference cells. ^{136 , 137 , 140 , 142} For example, an ${\eta }_P$ of $\sim 2.0\%$ was achieved by using the above metal grids as the transparent anodes in P3HT:PCBM devices, which is comparable to the ITO-based devices fabricated under the same conditions. ²³

Fig. 6

(a) Schematic illustration of a nanoimprint lithography (NIL) technique to produce metal grids. (b) SEM images of Cu grids fabricated by the NIL method. (c) Optical transmittance of Au, Cu, and Ag grid electrodes and a conventional ITO electrode. Reprinted with permission from Ref. 23. Copyright 2008 John Wiley and Sons.

Although the development of the NIL technique presents a great opportunity for using metal grids as TEs in electronic devices, there are still several factors that influence the large-scale applications. The extremely rough surfaces (generally in the range of tens of nanometers) limit their application in devices with thin organic layers. It is also difficult to use such metal grids as top TEs in OPVs and OLEDs since the transfer processes (heat or pressure) may cause damage to the underlying organic layers. Moreover, more effort is needed to integrate the NIL techniques with roll-to-roll processing, which is another significant challenge for utilizing metal grids as TEs in large-scale devices.

5.1.

Carbon Nanotubes

Since the first report on CNTs in the early 1990s by Iijima, ¹⁷¹ CNTs have attracted extensive academic and industrial attention due to their unique and promising mechanical and electrical properties. Individual single-walled CNTs (SWCNTs) have charge mobilities of ${10}^5{\mathrm{cm}}^2/\mathrm{Vs}$ and electrical conductivities in the 1 to $3\times {10}^6\mathrm{S}/\mathrm{m}$ range. ^{10 , 172} Despite the excellent electrical properties of individual tubes, thin films with a network of CNTs show more than three orders of magnitude lower conductivity and mobility (the highest reported conductivity was $\sim 6000\mathrm{S}/\mathrm{m}$ and mobility was $\sim 10{\mathrm{cm}}^2/\mathrm{Vs}$ ) as a result of the large contact resistance between two CNTs. ¹⁰ Since the early 2000 s, there have been extensive discussions regarding the potential applications of CNTs as TEs in optoelectronic devices; ^{173 – 175} however, several impediments including CNT synthesis, purification, and film processes have hindered their commercial use.

Solution processibility is one of the great advantages for nanoscale materials, which can be integrated with fast, large-scale, and high-throughput manufacturing methods. Therefore, dispersing the CNTs in compatible solvents or obtaining printable inks of CNTs is required for most solution-based deposition techniques. ¹⁷⁶ Several studies have shown that unfunctionalized CNTs can be dispersed in some organic solvents, such as chloroform, dichlorobenzene, dimethylformamide, and cyclohexylpyrrolidone. However, the concentration of these CNT solutions has only reached 0.1 to $2\mathrm{mg}/\mathrm{mL}$ , which limits their application for large-scale manufacturing. ^{176 – 179} Some dispersion agents or surfactants have been utilized to better promote the suspension of CNTs in organic solvents or even some aqueous media. ^{180 – 184} However, most surfactants in CNT solutions become impurities after forming solid films, leading to a significant decrease in film conductivity. ¹⁸⁵ Some research groups have demonstrated that it is possible to achieve high concentration and well-dispersed CNT solutions by chemically modifying the CNT surface with covalently bonded functional groups. ^{176 , 186} Though such chemical modification assists the dispersion of CNTs and prevents the possible rebundling mechanisms, the anchored small molecules break up the ordered ${\mathrm{sp}}^2$ bonding structures and become defects in single CNTs, affecting the conductivity of films made from these solutions. Therefore, it is necessary to deeply understand the relationship between conductivity, surfactants, and molecular structures, which will provide guidance for achieving stable CNT solutions with controllable concentrations while maintaining the good electrical properties of single CNTs and CNT films.

Several solution process techniques, including spin coating, ^{187 , 188} spray coating, ^{30 , 189} dip coating, ¹⁹⁰ and solution-related soft lithography methods, ^{191 – 193} have been demonstrated to fabricate uniform transparent conductive CNT films. For example, Wu et al. reported a simple filtration technique to deposit uniform SWCNT films on various substrates. An SWCNT film was first vacuum-filtered on a filtration membrane from a dilute suspension of nanotubes, followed by purification processes to remove possible surfactants and impurities. The SWCNT film was then transferred onto any required substrates after dissolving the membrane in solvent. Benefiting from the vacuum filtration process, this method can provide homogeneous SWCNT films with precisely controllable thicknesses by varying the solution concentration and volume. ¹⁷⁴ However, these deposition techniques are only effective for producing films on small area substrates and there are still many challenges to develop cheap and fast deposition techniques for large-scale and uniform CNT films. ^{194 , 195}

Besides the processing challenges, the rough surfaces of CNT films combined with the high sheet resistance ( $>200\mathrm{\Omega }/\square$ ) as a result of the possible defects and the poor interconnection between CNTs limit their applications in optoelectronic devices. In early reports, OPV devices with CNT films as TEs had ${\eta }_P$ of only $<1\%$ , compared to 3 to 5% in standard ITO-based devices. ^{19 , 196} Many film treatment methods were developed for achieving CNT electrodes with higher electrical conductivity and smoother film surface. ^{20 , 26 , 30 , 197 , 198} Tenent et al. reported uniform SWCNT films by using an ultrasonic spray method [see the photograph in Fig. 8(a)] with the sheet resistance of $\sim 150\mathrm{\Omega }/\square$ , which was reduced to $<50\mathrm{\Omega }/\square$ after nitric acid treatment. As shown in Fig. 8(b), an ${\eta }_P$ of 3.1% was achieved for a P3HT:PCBM device with the CNT electrode, which was only slightly lower than the ITO-based cells ( ${\eta }_P=3.6\%$ ). ¹⁸⁹ Ou et al. demonstrated surface-modified nanotube films with a surface roughness $<6.0\mathrm{nm}$ and sheet resistance of $\sim 100\mathrm{\Omega }/\square$ using PEDOT:PSS modification, ${\mathrm{HNO}}_3$ acid soaking, and polymer coating. With the modified CNT anodes, a maximum current efficiency of $\sim 10\mathrm{cd}/\mathrm{A}$ was achieved for an Alq3-based OLED, similar to the ITO-based reference devices. ¹⁹⁹

Fig. 8

(a) A photograph of single-walled carbon nanotube (SWNT) films ( $6\times 6{\mathrm{in.}}^2$ ) by a spray deposition method. (b)  $\mathrm{J}-\mathrm{V}$ characteristics of P3HT:PCBM devices with ITO and SWNT transparent electrodes. Reprinted with permission from Ref. 189. Copyright 2009 John Wiley and Sons.

Although there are a few laboratory demonstrations using CNT networks as TEs in OPVs and OLEDs, it is still a long journey for their application in large-scale optoelectronic modules. First, achieving high-quality and low-cost CNTs over a large scale has been one of the challenges in the past 20 years, and will continue to be the research focus of this field in the future. ¹⁷³ Though some scale-up technologies have been demonstrated by different research groups that have the potential for CNT mass productions, complicated purification processes, such as routes to remove remained impurities after synthesis and methods to distinguish the semiconducting and metallic components, still impede the scalability of high-quality CNTs and, hence, a reduction of the cost. ^{200 , 201} In terms of CNT thin films, the relatively low conductivity and transparency are still the two major factors that limit the device performance. The current deposition techniques for CNT films still need to be improved in their compatibility with large-scale roll-to-roll processing. Moreover, there is very limited study of the mechanical flexibility and environmental stability of devices with CNT electrodes.

5.2.

Graphene

Graphene is a two-dimensional material of close-packed carbon atoms with a honeycomb crystal lattice, which can be considered as a single atomic layer of graphite. ²⁰² Since the great breakthrough from its theoretical study to experimental studies in 2004, which led to the Nobel Prize in physics in 2010, ²⁰³ graphene has attracted tremendous attention and presented many possible applications in different areas. ^{204 , 205} Similar to CNTs where all carbon atoms are connected with ${\mathrm{sp}}^2$ -bonds, a single graphene sheet exhibits high in-plane conductance since all valence electrons are delocalized over the entire sheet. Moreover, graphene shows excellent optical properties: a single graphene layer has a theoretical transmittance of 97.7% (with a reflectance of $<0.1\%$ and absorbance of $\sim 2.3\%$ ). ^{34 , 206} Besides the excellent optical transparency and electrical conductivity, graphene also exhibits other impressive properties, such as great thermal and chemical stability, excellent mechanical properties (high flexibility and stretchability), and good interfacial contact with organic materials, which enables its potential application as TE in optoelectronic devices. ^{14 , 207 – 210}

Despite the promising properties of graphene as a TE, there are still several challenges remaining for its utilization in actual devices. Achieving high-quality graphene thin films and depositing graphene on required substrates is one of the biggest challenges. Since the pioneering report by Novoselov et al. in 2004, ²¹¹ mechanical exfoliation from bulk graphite materials is the basic technique to obtain high-quality graphene sheets for fundamental studies and small device demonstrations. ^{202 , 212 , 213} However, it is not suitable for high-throughput and large-scale industrial production lines. Chemical vapor deposition (CVD) is another successful technique for achieving graphene films, ^{32 , 47 , 214 – 216} which, despite its relatively high-cost deposition method with high treatment temperature, is still the most widely used fabrication method to obtain large graphene sheets. As shown in Figs. 9(a) and 9(b), Bae et al demonstrated large-scale (30-in.) CVD-grown graphene films with good flexibility and transparency. A monolayer of graphene film was first synthesized on copper foil using a CVD system, followed by some thermal and gas treatments for achieving high-quality films. A thermal-release tape was then attached with the graphene/copper foil for film transfer. After etching away the copper foil with copper etchant and cleaning the graphene surface with deionized water, the high-quality graphene film standing on the tape can be transferred to any target substrate. ²¹⁵ Some solution-based deposition methods, such as liquid phase exfoliation and chemical reduction of graphene oxide, have also been developed to produce graphene films in the past several years. Direct exfoliation of graphite in specific solvents by sonication is the simple and low-cost method for graphene inks so far. ^{207 , 217} However, the concentration of graphene dispersed in these solvents, up to $1.5\mathrm{mg}/\mathrm{mL}$ , is not high enough for some solution processes. Moreover, the mean flake size, generally on the order of 1 to $10\mu \mathrm{m}$ , is relatively small due to the sonication; therefore, films prepared from these solutions exhibit higher resistance as a result of significant junction resistances. ²¹⁸ This approach can be further improved by adding volatile agents (e.g., alkali salt) and organic surfactants into the original graphite dispersions, where the ions or organic molecules can intercalate between graphite layers and expand the interlayer spacing, leading to improved exfoliation and a more stabilized and better-dispersed graphene solution with a higher concentration (up to $10\mathrm{mg}/\mathrm{mL}$ ). ^{219 – 221} Another alternative low-cost method is to reduce graphene oxide to graphene as demonstrated by several groups in the past few years, which provides a great opportunity to fabricate graphene inks or films on a large scale. ^{222 – 225} However, these solution methods, either exfoliation with metal ions and organic molecules or chemical reduction from oxides, induce a number of defects into the graphene sheets and interfaces, leading to a dramatic increase in the sheet resistance of the related graphene films.

Fig. 9

(a) and (b) Photographs of flexible graphene layers grown by a chemical vapor deposition technique. Reprinted with permission from Ref. 215. Copyright 2010 Nature Publishing Group. (c) Device structure of flexible phosphorescent OLED with a graphene electrode. (d) Current efficiency and luminance (insert) of phosphorescent OLEDs with ITO and graphene as transparent electrodes. Reprinted with permission from Ref. 47. Copyright 2012 Nature Publishing Group.

Although a single graphene sheet exhibits a low sheet resistance of $<100\mathrm{\Omega }/\square$ , most graphene films produced by the above methods show a high sheet resistance on the order of a $1\mathrm{k}\mathrm{\Omega }/\square$ range as a result of the possible structural defects and contact resistances between graphene flakes. ^{14 , 207 , 222} De Arco et al. reported high-quality graphene films synthesized by CVD with a sheet resistance of $230\mathrm{\Omega }/\square$ and transmittance of 72% at $\lambda =550\mathrm{nm}$ . With such graphene films as TEs, a small molecule $\mathrm{CuPc}/{\mathrm{C}}_{60}$ PV device with an ${\eta }_P$ of 1.18% was demonstrated, which was slightly lower than the ITO-reference cell ( ${\eta }_P=1.27\%$ ). ²¹⁴ Some chemicals, such as ${\mathrm{AuCl}}_4$ , ^{31 , 32} ${\mathrm{HNO}}_3$ , ²¹⁵ and thionyl chloride, ²²⁶ have shown doping effects to improve the conductivity of graphene layers while maintaining the high levels of transmittance. For example, a CVD-grown, four-layer graphene film doped with ${\mathrm{HNO}}_3$ was demonstrated with a sheet resistance as low as $30\mathrm{\Omega }/\square$ at $\sim 90\%$ transparency. ²¹⁵ Han et al. used ${\mathrm{HNO}}_3$ and ${\mathrm{AuCl}}_4$ as p-dopants in CVD-grown graphene films and applied similar highly conductive graphene films to replace ITO in flexible OLEDs, ⁴⁷ resulting in a decrease in sheet resistance from 87 to 54 and $34\mathrm{\Omega }/\square$ , respectively. These resistance values are now approaching that of ITO and thin metal layers, and are much higher than that of PEDOT:PSS and CNT layers. Using such highly conductive graphene to replace ITO in a green phosphorescent OLED [see Figs. 9(c) and 9(d)], a very high current efficiency phosphorescent OLED of up to $98.1\mathrm{cd}/\mathrm{A}$ was achieved, which was appreciably higher than the ITO-reference cells ( $81.8\mathrm{cd}/\mathrm{A}$ ). ⁴⁷

Apart from improving the conductivity of graphene films, interfacial modification between graphene electrodes and active layers is another challenge in efficient optoelectronic devices. Similar to other possible TE candidates, polymers, such as PEDOT:PSS, and metal oxides, such as ${\mathrm{MoO}}_3$ , ${\mathrm{TiO}}_2$ , and ZnO, are widely utilized to modify graphene interface and minimize the possible energy barriers for efficient injection/extraction of electrons and holes. In some cases, some interfacial layers are required to improve the surface wettability for solution-processed transport/active layers. For instance, Zhang et al. introduced a layer of Al nanoclusters on a single-layer graphene to overcome its hydrophobic properties for ${\mathrm{TiO}}_2$ deposition. As a result of such surface modification, the P3HT:PCBM devices with graphene as TE showed an ${\eta }_P$ of $\sim 2.6\%$ , which was better than those without the modification and was close to that of the ITO-based references ( ${\eta }_P=3.45\%$ ). ²²⁷ These results suggest the great potential of using graphene as the TE in flexible organic optoelectronics. Moreover, graphene electrodes also possess outstanding mechanical flexibility. ^{14 , 216} De Arco et al. compared the graphene and ITO conductance under different bending conditions: only minor changes in the film morphology and conductance were observed for graphene films after bending the films to 160 deg, whereas ITO films exhibited irreversible cracks and a decrease in conductance under 60 deg of bending.

Nevertheless, for large-area commercial applications, the resistance of the graphene layers needs to be further reduced, along with simplifying the graphene fabrication/transfer processes. Therefore, more effort is needed to obtain highly conductive graphene films with low-cost and large-scale deposition methods to enable its industrial application.

5.3.

Metal Nanowires

Besides the carbon-based nanomaterials, films with random networks of metal NWs are also attractive candidates for transparent conductive electrodes. ^{18 , 228 , 229} Although many metal nanostructures, such as copper, ^{230 , 231} gold, ^{232 , 233} and cupronickel, ²³⁴ have been demonstrated with promising properties and possibility as electrodes, Ag NWs have been the focus of research in this area due to the excellent electrical properties of bulk Ag materials and the synthesis scalability of Ag NWs. Ag NWs with a typical diameter of 20 to 40 nm and length of $>20\mu \mathrm{m}$ dispersed in water or organic solvents are now commercially available.

Since metal NWs with proper surface treatment can be well dispersed in many solvents, many solution-based processing techniques, such as spin coating, ³⁹ spray coating, ^{235 , 236} drop casting, ⁵⁰ doctor blade coating, ^{11 , 237} Mayer rod coating, ⁴⁰ brush painting, ²³⁸ etc., have been reported by different reach groups to achieve films with random NW networks. Similar to the discussion in CNTs, electrical and optical properties of the random NW networks are strongly dependent on the electrical properties of individual NWs, the interconnection between NWs, and the NW density in the resulting films. In general, the conductivity of NW networks increases as the length and diameter of an individual wire increases. ²³⁹ However, it is difficult to achieve NWs with an extremely long length and the typical reported length of Ag NWs is in the range of 1 to $50\mu \mathrm{m}$ . It is also challenging to maintain the integrity of very long NWs in the solid-state film, as they tend to break during the dispersion and deposition processes. Controlling the diameter of metal NWs is also critical for achieving high-quality films. NWs in thin films provide the conductive pathways for charge carrier transport, but they also function as scatter centers to incident light affecting the optical properties of the films. Therefore, the diameter of metal NWs should be minimized in order to achieve high optical transparency. Moreover, the diameter of individual NWs determines the surface roughness of films with random NW networks, which should also be reduced for most optoelectronic devices. Furthermore, an optimized wire density is required to balance between the electrical and optical properties for specific applications.

The junction resistance between the jointed wires is another dominant factor that affects the film resistance, which is similar to that for CNTs. ^{10 , 49} Several modification methods, such as thermal annealing, ^{35 , 228} optical sintering, ²⁴⁰ mechanical pressing, ²⁴¹ and fusing with other materials, ⁴² have been demonstrated to reduce the junction resistance and improve the film conductivity. For example, Tokuno et al. applied a mechanical pressure of 25 MPa on as-deposited Ag NW films for 5 s at room temperature, which leads to a dramatic decrease in sheet resistance from $>{10}^4$ to $8.6\mathrm{\Omega }/\square$ . ²⁴¹

Several research groups have reported metal NW films with transmittance $>85\%$ and sheet resistance $<20\mathrm{\Omega }/\square$ [Figs. 10(a) and 10(b)], which are comparable with standard ITO electrodes, and explored them as TEs in organic optoelectronic devices. ^{11 , 39 , 242} Since the first application in small molecule OPVs that had an ${\eta }_P$ only $<0.5\%$ , the overall performance of metal NW-based devices has been steadily advancing. Generally, a buffer layer (e.g., PEDOT:PSS, ZnO, or ${\mathrm{TiO}}_x$ ) is required to flatten the surface of metal NW networks and to improve the contact with the active material so that they can serve as either anodes (collect/inject holes) or cathodes (collect/inject electrons) with proper buffer layers. ^{237 , 243 – 245} For instance, Leem et al. fabricated P3HT:PCBM cells with Ag NWs to replace the bottom ITO TE. With PEDOT:PSS and ${\mathrm{TiO}}_x$ as buffer layers, normal and inverted device architectures were achieved, respectively, which had comparable ${\eta }_P$ of 2.0 and 3.5% compared to those of ITO-based reference cells. ³⁹ Gaynor et al. successfully fabricated ITO-free white OLEDs with Ag NWs/poly (methyl methacrylate) composites as TEs, which showed a luminous efficiency $>30\mathrm{lm}/\mathrm{W}$ and were close to that of the ITO-based devices ( $35.8\mathrm{lm}/\mathrm{W}$ ). ²⁴⁶ The solution-processibility of metal NWs also offers a great opportunity to make them as transparent top electrodes for organic optoelectronic devices. For example, Chen et al. successfully applied Ag NW electrodes in a high-performance polymer solar cell to achieve a visibly transparent device, where the ITO and ITO-nanoparticles-fused Ag NW layers serve as the bottom and top TEs, respectively. Such transparent devices showed efficiencies of ${\eta }_P=4.0$ and 3.8% when illuminated from ITO and Ag NW electrodes, respectively, and a maximum transmittance of $\sim 60\%$ in the visible light range. ²⁴⁷ Great mechanical properties are also the advantages of metal NW networks compared to ITO substrates. ^{40 , 88 , 248 , 249} As shown in Figs. 10(c) to 10(e), Yang et al. demonstrated the flexibility of Ag NW electrode based OPV devices on flexible substrates, which exhibited a recoverable efficiency under a bending angle of 120 deg. ⁴⁰ Akter and Kim also studied the stretchability of Ag NW films, where the conductivity of the electrode remained unchanged with $\sim 20\%$ elongation. ²⁴⁸

Fig. 10

(a) SEM image of a layer of Ag nanowires (NWs). (b) Transmittance and sheet resistance (insert) of Ag NW films with various thicknesses (different spin-speeds). Reprinted with permission from Ref. 39. Copyright 2011 John Wiley and Sons. (c) and (d) Experimental setup for measuring the flexibility of OPV devices with Ag NWs. (e)  $\mathrm{J}-\mathrm{V}$ characteristics of a flexible OPV device under different bending conditions. Reprinted with permission from Ref. 40. Copyright 2011 American Chemical Society.

The above great device performance and excellent mechanical stability suggest the strong potential of using random metal NW networks as transparent conductive electrodes. However, several challenges still remain to be solved. For instance, the poor adhesion between the metal NWs and substrates limits their compatibility with different substrates. Adding some adhesive layers at the NW/substrate interface or treating metal NW surface with proper ligands to make them better anchored on substrates are some possible approaches to promote the adhesion. ²²⁸ The long-term stability of metal NW electrodes is another concern for their application in electronic devices. There has been some evidence suggesting that electromigration of metal atoms along the NWs causes failure of the wires and the entire network. This is especially important for devices that require high operation current, in which case the film conductivity keeps decreasing until their full breakdown. ^{250 , 251} Moreover, some metal NWs (e.g., Ag NWs) can be corroded through chemical reactions, such as oxidation and suffixation, which also cause the degradation of metal NWs and their related devices.