Contact engineering in organic field-effect transistors

Materials Today  Volume 00, Number 00  September 2014

RESEARCH: Review

RESEARCH

Contact engineering in organic field-effect transistors Chuan Liu1,2, Yong Xu1 and Yong-Young Noh* Department of Energy and Materials Engineering, Dongguk University, 26 Pil-dong, 3 ga, Jung-gu, Seoul 100-715, Republic of Korea

Organic field-effect transistors (OFETs) are promising for numerous potential applications but suffer from poor charge injection, such that their performance is severely limited. Recent efforts in lowering contact resistance have led to significantly improved field-effect mobility of OFETs, up to 100 times higher, as the results of careful choice of contact materials and/or chemical treatment of contact electrodes. Here we review the innovative developments of contact engineering and focus on the mechanisms behind them. Further improvement toward Ohmic contact can be expected along with the rapid advance in material research, which will also benefit other organic and electronic devices. Introduction of contact injection The injection problem Newly synthesized organic semiconductors (OSCs) have demonstrated carrier mobilities of over 10 cm2/V s for the crystalline state and 2 cm2/V s for the amorphous state in organic field-effect transistors (OFETs) [1–5], approaching those of polycrystalline silicon metal–oxide–semiconductor FETs (Si MOSFETs). Such dramatic progress alleviates the bottleneck of the charge transport in OSCs but the need for improvement in contact injection properties remains. Conventional single-crystal Si MOSFETs can exhibit excellent contact properties (Ohmic contacts) since the charge injection is by tunneling from a metal contact to heavily doped silicon bulk, after injection the charge carriers can easily transport from the contact wells to the channel (bulk) as they are the same material, that is, silicon with different doping concentrations [6]. However, for new types of FETs, such as organic, metal oxide, and carbon-based FETs, the contact resistance (Rc) is not so good because the electrode–semiconductor junction consists of heterogeneous materials, that is, metal and semiconducting materials, which shows higher Rc values. For example, OFETs usually have an Rc value of 100 V cm as compared to that of Si MOSFETs where Rc is less than 0.1 V cm [7,8]. Therefore, the device performance is *Corresponding author:. Noh, Y.-Y. ([email protected]) 1

These authors contributed equally to this work. Current address: State Key Laboratory of Optoelectronic Materials and Technologies, School of Physics and Engineering, Sun Yat-Sen University, Guangzhou 510274, China. 2

significantly limited by the poor charge injection and therefore a small enhancement of injection would lead to a considerable improvement of OFET performance. Some examples are shown in Table 1. Note that comparison between Rc values is only valid in the same example, since in different examples semiconductors, processing techniques, and device structure vary. For clarity, in this context we use ‘field-effect mobility’ (mFET) to denote the mobility extracted from transfer characteristics measuring the OFET device performance (including device factors such as contact resistance), and use ‘transport mobility’ (m) to denote the mobility associated with transport property in semiconductor materials.

Contact injection barrier The charge injection from metal/OSC junction has been described in terms of thermionic emission or tunneling mechanism. For both models, the current density in a metal/semiconductor diode is described as     qV a J ¼ J 0 exp 1 ; (1) kT where Va is the applied voltage on the semiconductor and J0 is closely related to the Schottky barrier, wb, and depends on the properties of the metal/semiconductor junction. If not taking into account the image-lowering effect, the thermionic emission model (Richardson–Schottky model, Fig. 1a, left) gives  q’  J 0 ¼ J RS ¼ A T 2 exp  b ; kT

(2)

1369-7021/ß 2014 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.mattod.2014.08.037

1 Please cite this article in press as: C. Liu, et al., Mater. Today (2014), http://dx.doi.org/10.1016/j.mattod.2014.08.037

MATTOD-405; No of Pages 18 RESEARCH

Materials Today  Volume 00, Number 00  September 2014

TABLE 1

Examples of contact engineering for OFETs. For contact resistance (Rc), field-effect mobility (mFET ) and threshold voltage (Vth), the data inside and outside brackets are from unmodified electrodes and modified electrodes, respectively [extracted at the same gate voltage (Vg) and drain voltage (Vd) as in the references].

RESEARCH: Review

Contact engineering

Contact materials

Semiconductor (type of carriers)

wm (Au)

Device structure

Rc (kV cm)

mFET (cm2/(V s))

Vth (V)

Ref.

SAM (p-type)

NOTP/Au

TIPS-Pentacene (p)

4.84 (4.75)

TGBC

291 (998)

0.103 (0.047)

6 (9)

[9]

SAM (p-type)

BTFMBT/Au

P3HT (p)

5.8 (5.0)

TGBC

180 (610)

0.26 (0.16)

–

[10]

SAM (n-type)

4-Chlorobenzenemethanethiol/Au

PDI-8CN2 (n)

–

BGBC

32 (130)

0.10 (0.05)

10.6 (5.5)

[11]

SAM (ambipolar)

1DT/Au

F8BT (ambipolar)

4.0–4.2 (4.7–4.9)

TGBC

330 MV (560 MV) (p)/670 MV (2050 MV) (n)

0.0007 (0.0004) (p)/0.001 (0.0005) (n)

24 (32)/20 (30)

[12]

CIL (metal oxide, p-type)

MoO3/Au

C8-BTBT (p)

–

BGTC

10 (5000)a

2.3 (0.87)

5 (12)

[13]

CIL (metal oxide, n-type)

TiOx/Au

PC61BM (n)

–

BGTC

300 (15,000)

0.028 (0.001)

0 (5)

[14]

CIL (salts, p-type)

FeCl3/Au

C8-BTBT (p)

–

BGTC

8.8 (200)b

7.0 (3.4)

10 (24)

[15]

CIL (salts, n-type)

CsF/Au

PTVPhI-Eh (n)

4.1 to 4.2 (4.7)

TGBC

5000 (1,200,000)c

0.26 (0.022)

44.27 (61.1)

[16]

CIL (polymer, n-type)

PEIE/Au

P(NDI2OD-T2) (n) (N2200)

3.90 (5.10)

TGBC

–

0.04 (0.1)

0.4 (4.5)

[17]

Carbon-based

Graphene

Pentacene (p)

4.71 (4.46)

BGBC

560 (850)

0.4–1.01 (0.16–0.28)

–

[18]

a b c

Read from Fig. 2 (at Vg = 80 V, Vd = 3 V) in the reference. Read from Fig. 2 (at Vg = 40 V, Vd = 1 V) in the reference. Read from Fig. 5 (L = 10 mm, Vg = 60 V, Eh = Ethylhexyl) in the reference.

2 Please cite this article in press as: C. Liu, et al., Mater. Today (2014), http://dx.doi.org/10.1016/j.mattod.2014.08.037

MATTOD-405; No of Pages 18 RESEARCH

RESEARCH: Review

Materials Today  Volume 00, Number 00  September 2014

FIGURE 1

Origin of contact resistance in OFETs. (a) Metal/OSC interface injection governed by thermionic emission (left) or tunneling (right). (b) Injection barrier follows Schottky–Mott limit (left), Fermi-level pinning (middle), or the integer charge transfer (right) model. Reproduced with permission from [20–22] (2009, 2009, 2013).

where A* is the Richardson constant, T is the absolute temperature, wb is the barrier height, and k is the Boltzmann constant. If the applied voltage is large or the semiconductor side is heavily doped (dopant density is ND) such that the barrier width becomes very thin (Fig. 1a, right), the current flow in the diode then becomes a tunneling process described by rffiffiffiffiffiffiffiffiffiffiffiffi  2’ ee0 m : (3) J 0 ¼ J t  exp  b h ND

according to the above models Rc from the interfacial injection at zero bias is

Note that wb depends only weakly on the doping density through lowering of the image-force barrier. Besides, another possible injection process would happen if there are massive gap-states near the interface inside the band-gap (which may come from defects), wherein charges tunnel from the metal to these gapstates and then hop to the transporting levels in the OSC (Fig. 1a, right, ‘injection via gap-states’) [19]. As one can estimate Rc at zero bias by   dJ 1 kT 1 J ; ¼ (4) Rc ¼ dV a q 0

for the classical thermionic emission process and tunneling process with rich dopants, respectively. In either case, the current level would ideally increase and Rc will decrease exponentially with wb, making it the most critical parameter for contact engineering. With vacuum-level alignment, wb is simply the difference between the work functions of the metal electrode (wm) and the highest occupied molecular orbit (HOMO) or lowest unoccupied molecular orbit (LUMO) levels of the OSCs (the Schottky–Mott limit). In fact, wb is more related to certain gap-states (e.g., EICT+ and EICT) induced by polarons rather than HOMO and LUMO

V a ¼0

 Rc ¼ Rc;RS ¼

 q’  k b ; and exp  qA T kT

(5)

rffiffiffiffiffiffiffiffiffiffiffiffi ee0 m ; ND

(6)



2’b Rc ¼ Rc;t  exp h 

3 Please cite this article in press as: C. Liu, et al., Mater. Today (2014), http://dx.doi.org/10.1016/j.mattod.2014.08.037

MATTOD-405; No of Pages 18 RESEARCH

RESEARCH: Review

levels. But for simplicity, we approximately take the latter values to calculate wb for discussion. Many non-interactive interfaces formed by spin-coating polymer OSCs on metals, as well as some interfaces formed by evaporating small-molecule OSCs on contaminated metals, follow the Schottky–Mott limit (Fig. 1b, left) [21]. Yet most metal-on-OSC interfaces made by thermal evaporation and small molecule-on-clean metal interfaces depart from this vacuum-level alignment [21,23,24]. If the contact metal has a large number of surface states or a surface modification layer, Fermilevel pinning can happen such that the work function after forming the contact ð’0m Þ is pinned, and thus wb becomes almost constant on the contact species. For example, the Fermi level at metal/C60/OSC interfaces (or the effective work function of the metal, ’0m ) was found to be pinned to the charge neutrality level of C60 4.7 eV for different metals with varying values of wm (Fig. 1b, middle) [20]. The surface modification layer of C60 (3 nm) induces a large interfacial dipole at the metal/OSC interfaces and thus reduces wb to be a constant for hole injection. The above two limits can be unified in the integer charge transfer (ICT) model, which has recently gained more experimental confirmation from the energy-level alignment at metal/OSC interfaces [22]. The injection barrier will follow the Schottky–Mott limit when there is negligible electron transfer between the contact and the organic material in equilibrium (i.e., ’0m ¼ ’m ). This happens when wm falls between two energy levels, EICT and EICT+, which are determined by the OSC. On the other hand, wm will be pinned at certain levels if it falls below EICT or above EICT+ ð’0m ¼ constant 6¼ ’m Þ, therein leading to a constant wb. Such behavior has been observed when depositing 1,4,5,8-naphthalenetetracarboxylic (NTCDA) and dianhydride tris-(8-hydroxyquinoline) aluminum (Alq3) on different substrates, as shown in Fig. 1b (right). Here, we have classified the interfaces simply from the viewpoint of work-function change. For a detailed classification by the strength of the interaction (physisorption or chemisorption), readers can refer to the review by Braun et al. [25]. The work function wm can be measured by Kelvinprobe force microscopy (KPM), whereas wb can be determined by fitting current–voltage characteristics to the aforementioned thermionic emission or tunneling model [26], or it can be directly detected by photoelectron spectroscopy (PES) [21,25].

Evaluation of contact injection in OFETs The injection barriers, together with all the other factors that limit the injection (e.g., tunneling barriers, geometrical factors, and trapping sites), are sensitively reflected in the contact resistance Rc of OFETs, which is usually extracted by a two-terminal or multiple-terminal method. We list only the simplest methods in Fig. 2, along with device structure, measured parameters, extraction methods, and properties. For a full review of extraction methods readers can refer to the review by Natali and Caironi [8]. The most common two-terminal methods are the Y-function and transfer-length methods (TLM, sometimes also referred as ‘transmission-line method’), both of which measure the transfer characteristics (the drain current (Id) against the gate voltage (Vg)). The Y-function method requires only one transfer scan (Id–Vg) of an individual device in linear regime with applying a small sourcedrain voltage (Vd  Vg) (Fig. 2a) [27]. If the transconductance ðg m ¼ ð@I d =@V g ÞÞ starts to decrease from a certain Vg (usually for large values of Vg and small channel lengths L), then the contact

Materials Today  Volume 00, Number 00  September 2014

effect starts to dominate the mFET attenuation and Rc as well as a contact-free transistor mobility can be extracted from the Y-funcpffiffiffiffiffiffiffi tion ðY ¼ ðI d = g m ÞÞ [27]. Because the Y-function method assumes a constant Rc with Vg, one should turn to TLM to extract Rc at arbitrary values of Vg (especially at large values of Vg). Several contacts having unequal spacing between them or, equivalently, OFETs with different values of L are scanned in the linear regime of Id–Vg to extract the intercept of TLM plot, that is, Rc (Fig. 2b) [28]. The group of Rtot–L linear-regression lines often converge at a common point, called the convergence point (CP), which can be used to qualify the charge injection and charge transport [29]. When there is considerable scatter in m and Vth (e.g., OFETs on bare SiO2 with significant trapping effect), a modified TLM (M-TLM) is suggested to improve extraction accuracy [30,31]. Furthermore, to extract Rc in a saturated regime, power TLM (P-TLM) should be applied by measuring the output characteristics (Id–Vd), which are derived from the dissipated power in the whole device regardless of the linear or saturated source-drain electric field [32]. The two-terminal methods are valued for their simplicity and generality and can be used regardless of device structure and size, but they rely on idealized assumptions and cannot be used to separate the source and drain resistance (Rs, Rd). In contrast, multiple-terminal methods require complicated patterning of the electrodes and a relatively large device size, but they are valued for their ability to measure Rs and Rd accurately. The gated fourpoint-probe (gFFP) measurement employs two additional terminals that detect the local potential in the channel, V1 and V2 (Fig. 2c) [33,34]. This method generally necessitates the gradual channel assumption and may suffer from channel nonuniformity. Using KPM, one can measure the electrostatic potential along the channel and thus detect drops at the source/drain electrode (Vs, Vd) [35,36]. No matter whether or not a gradual channel is assumed, Rs and Rd are calculated by dividing Vs and Vd by Id. Apparently, these methods are much more sophisticated for fabrication and equipment set-up, and they are also more suitable for bottom-gate devices as compared to the two-terminal methods performing conventional I–V scanning. In addition to Rc, the surface states at the OSC/dielectric interface are also affected by the fabricated metal/OSC junction, and they become an important factor for characterizing charge injection. The density of surface states (NST) increases in low-quality contacts (i.e., in the case of strong diffusion or thermal damage), and it can be roughly estimated by using the inverse sub-threshold swing (SS) in the Id–Vg curves [13,38]. NST can be extracted much more precisely in the low-frequency noise measurement (LFN) by detecting the noise of Id originating from the fluctuations in carrier number or transport mobility m [39,40], a method that has recently demonstrated its power in carbon- or nanomaterial-based FETs [41,42].

Device factors Role of device structure Note that none of the above methods is capable of measuring the specific resistivity at the metal/semiconductor interface alone. The measured Rc includes all the resistances except the channel resistance (Rch), which is the contribution from just the linear field in the accumulation channel, and Rc includes the part at the interface of contact/OSC (Rc,int) and the part from adjacent to the contact to

4 Please cite this article in press as: C. Liu, et al., Mater. Today (2014), http://dx.doi.org/10.1016/j.mattod.2014.08.037

MATTOD-405; No of Pages 18 RESEARCH

RESEARCH: Review

Materials Today  Volume 00, Number 00  September 2014

FIGURE 2

Main methods used to evaluate contact resistance in OFETs: (a) Y-function method, (b) transfer-line method (TLM), and (c) gated four-point probe method (gFFP). Reproduced from [27,33,37]. Copyrights (2010, 2007, 2005).

near the channel (Rc,bulk). The latter is always affected by geometrical factors and transport properties. The advantage of staggered geometry as compared to coplanar structure is based on the horizontally extended injection area gapped between the contact electrodes and gate dielectrics [43]. As Vg increases in OFETs, the channel becomes more conductive and the injection extends further away from the channel under the contact area, thereby reducing Rc and boosting the injection (‘current crowding’ effect). Such compensation alleviates the dependence of injection and Rc on the Schottky barrier height in the staggered configuration, dependence that is much more dominant in coplanar devices with vertical and smaller injection areas. This leads to that some OSCs in particular, for example, parylene-, thiophene-, or diketopyrrolopyrrole (DPP)-based OSCs, exhibit unipolar transport in coplanar structures but can exhibit ambipolar operation in staggered structures with the same Au electrodes [44].

Role of transport properties Injection and Rc are affected by transport mobility and the hopping mechanism of OSC mainly because: (a) a low transport

mobility near contact impedes metal/OSC injection [45] as well as the bulk transport mobility to reach the channel [46], enlarging Rc,int and Rc,bulk simultaneously; (b) more localized states or a stronger hopping mechanism causes stronger dependence of m and Rc on Vg [47]. In OSC m can be described as m / jV g  V th jb

(7)

Here b correlates to the density of localized states and the hopping mechanism [47,48]. Due to (a) and Eqn 7, it has been observed that Rc also exhibits a power dependence on Vg [31,49] ðRc  Rc0 Þ / jV g  V th jg ;

(8)

where Rc0 denotes Rc at infinite Vg. In particular, if in TLM analysis the convergence point (Lconv, Rconv) appears [29], then simply Rg1  Rconv and g  (b + 1). Hence, for a certain OSC, different morphologies (such as single crystalline, poly-crystalline and amorphous) and trap densities induced by depositions can result in different values and Vg-dependence of Rc [50]. Even insulating dielectric layers can also affect Rc by tuning morphology, defect states and energetic disorder in OSC layer [31]. 5

Please cite this article in press as: C. Liu, et al., Mater. Today (2014), http://dx.doi.org/10.1016/j.mattod.2014.08.037

MATTOD-405; No of Pages 18 RESEARCH

Contact improvement

RESEARCH: Review

For a more detailed review about characterizations and device physics, readers can refer to the review by Natali and Caironi [8]. Here, we focus on the use of contact engineering to reduce Rc, improve injection, and enhance mFET of OFETs. Researchers have intensively adapted tuning wb as a primary method, which can be achieved by changing the work function of the contact electrodes, inducing surface states in order to control Fermi-level pinning, or generating gap states to assist tunneling. Raw metals of whatever species (Au, Cu, Ti, Pt, Cr, Mo, etc.) exhibit large values of Rc in OFETs and may exhibit lowered Rc values for hole injection after oxygen plasma etching or washing with Piranha solution [49,51]. In recent studies, fine tuning of hole injection and electron injection has been achieved by further modification of metals using self-assembled monolayers (SAMs), insertion layers between the metal and the OSC, doping OSC with effective dopants, or nonmetal electrodes. We will briefly discuss the methods and mechanisms of these approaches in the following sections.

Improving injection via self-assembled monolayer Dipole formation of SAM and work function Modification of the work function of the clean metal surface by SAM was accomplished with contributions from two dipole layers (Fig. 3a): the bond dipole (BD) formed at the immediate metal– sulfur (M–S) interface and the aligned dipole moments of the molecules within the SAM, causing vacuum shifting (DUvac)

Materials Today  Volume 00, Number 00  September 2014

[52]. According to classical electrostatics, the potential drop at the thiol-SAM/metal interface can be calculated by the dipole from the integral of the electric fields, and the corresponding change in work function and wb is [53]   mMS m D’ ¼ BD þ DU vac ¼ N þ SAM ; (9) e0 kMS e0 kSAM where N is the grafting density of the SAM, mM–S and mSAM are the dipole moments of the M–S bond dipole and the SAM molecule perpendicular to the metal surface, respectively, e0 is the permittivity of free space, and kM–S and kSAM are the dielectric constants of the M–S bond dipole and SAM molecule, respectively. The first term is strongly dependent on the metal species (e.g., the Ag–S dipole moment is larger than the Au–S dipole moment) [54], and the second part is determined by the SAM molecular structure [52,55]. Although the first term has an almost fixed sign and magnitude depending on the metal (the sign can be changed when using dockings other than thiols) [56], the latter can be positive or negative and its magnitude is more dominant in the total dipole moment [57]. Therefore, engineering the substituents could finely tune the work function of the metals. The commonly used thiol-SAMs on Au or Ag for tuning charge injection of OFETs are listed in Fig. 2b. Some of the work functions are shown in Fig. 2c. It can be seen that generally alkane terminals [–(CH2)n– CH3, –S–CH3, –O–CH3, and phenyl, –C6H5] decrease the work function because the dipole moments are opposite that of the

FIGURE 3

Using SAMs to modify the injection barrier. (a) Formation of dipole and wm tuning. (b) Commonly used SAMs in OFETs. (c) Corresponding wm values of modified Au. Inset in (c) shows using a mixture of two SAMs to continuously tune the work function of Ag. (d) Tuning wm on Au(1 1 1) by designing a terpyrimidine SAM with the same dockings (X = S) and different tail groups (Y = NH2, H, F, and CN) or with different dockings (X = O, S, Se, and Te) and the same tail group (Y = H). Reproduced with permission from [9,26,52,66]. Copyrights (2008, 2010, 2009, 2010). 6 Please cite this article in press as: C. Liu, et al., Mater. Today (2014), http://dx.doi.org/10.1016/j.mattod.2014.08.037

MATTOD-405; No of Pages 18

effective M–S dipole (positive sign, pointing out) [57–60]. In contrast, a terminal with halogen atoms having a large electronegativity (–CF3, –Cl, –Br, –COOH) increase the work function because their dipole moments are in the same direction as that of the M–S dipole (negative sign) [12,54,57,59,60]. Hence, alkanethiols are regarded as suitable for reducing the electron injection barrier (wm decreases to 4.1 eV), whereas perfluorinated thiols are useful for improving hole injection (wm increases to 5.8 eV) [26,58– 65]. In designing SAMs, modifying the tail groups are more effective than changing the docking for tuning wm (Fig. 3d), for example, –NH2 decreases wm, whereas –F and –CN increase wm [66]. Note that the effects of increasing/decreasing wm are similar when casting the same SAM on different metals, for example, Au, Ag [54,67,68], Cu [69], Pt [70], ITO [71], and so on, as well as on nonmetal contacts (e.g., graphene) [72].

Fine tuning of the work function The work function can be flexibly tuned by controlling the coverage of a single SAM [56,73] or, alternatively, by mixing two SAM

RESEARCH

molecules [26,54,64]. Such an approach is straightforward thanks to the use of solution-processing for SAM deposition. Several studies have proven, both theoretically and experimentally, that the change of work function is almost linearly correlated to the coverage percentage of the SAM on Au or Cu, that is, from zero shift to maximum shift [56,73]. However, 100% coverage of SAM is not the best condition for achieving maximum shift of wm due to the depolarization effect. Cornil et al. [74] reported that the dipole moment of SAM molecules is strongly reduced by as much as 30% when going from the isolated state to the aggregate state via a quantum-chemical calculation. To further enlarge the tuning region of the work function, two SAMs can be mixed on metals [26,64]. For example, by mixing the aforementioned 1-decanethiol (DT) and 1H,1H,2H,2H-perfluorodecanethiol (PFDT) in different ratios in a methanolic solution and immersing the Ag surface (4.67 eV) in the solution, the DT and PFDT molecularly mixed monolayer covers the homogeneous Ag surface and the work function then varies between 4.1 and 5.8 eV [26]. The work function increases almost linearly with the increasing ratio of FDT

FIGURE 4

Photoactive azobenzene SAM on Au. (a) H-azobenzene SAM and the time evolution of changes in the work function of the modified gold electrode as monitored by UPS during UV irradiation and thermal recovery. (b) Perfluorinated azobenzene SAM and its work function as monitored by KPFM during the same process. The opposite tuning of wm by irradiation between (a) and (b) is due to different end groups. Reproduced with permission from [75,77]. Copyrights (2008, 2013). 7 Please cite this article in press as: C. Liu, et al., Mater. Today (2014), http://dx.doi.org/10.1016/j.mattod.2014.08.037

RESEARCH: Review

Materials Today  Volume 00, Number 00  September 2014

MATTOD-405; No of Pages 18 RESEARCH

Materials Today  Volume 00, Number 00  September 2014

RESEARCH: Review

composition (Fig. 3c, inset), suggesting that fine control of the injection barrier height can be achieved by controlling the composition of the thiol solution. Recent advances have demonstrated that the injection barrier with azobenzene-derived SAMs can be finely modulated by photoisomerization. Ah Qune et al. [75] reported that the molecular dipole moments within H– and CN– azobenzene SAMs alter during photoisomerization from the trans form to the cis form and change back upon thermal recovery (i.e., back from the cis form to the trans form) (Fig. 4a). Changing the substituted functional group in the p0 position of the azobenzene from electron-donating to electron-withdrawing resulted in opposite responses of wm (and thus wb) of the SAM-covered metal, as revealed by ultraviolet photoemission spectroscopy (UPS) measurement. The change of wm during the cis to trans thermal recovery was monitored in real time and showed a reversible tuning up to 0.2 eV. Further work by Crivillers et al. [76,77] demonstrated that attaching perfluorinated substituents to azobenzene SAMs increases the wm of Au to 5.40 eV. The photochemical modification-induced modulation of the wm of Au(1 1 1) of 220 meV predicted by the theory corresponds well to experimental measurements by scanning Kelvin-probe microscopy (SKPM). The control of the work function of gold electrodes by photoisomerization of SAMs could lead to more effective and controlled tuning of their electrical characteristics in, for example, bio-sensing device applications.

Injection barrier with a SAM Similar to the metal/OSC case, the injection barrier in an OSC/ SAM/metal structure can deviate from the Schottky–Mott limit because of the difference between the wm of the SAM/metal interface and the HOMO or LUMO of the OSC. This difference is caused either by the extra interface dipole formation upon deposition of OSC or the aforementioned Fermi-level pinning. For simplicity, we again take the aforementioned DT and PFDT as examples [12]. The work function of DT/Au and PFDT/Au was measured by KPM to be 0.7 eV and +0.5 eV, respectively, relative to the UV/ozone-treated Au (Fig. 5a). The hole-injection barrier for the OSC is expected to change by +0.7 eV for DT and 0.5 eV for PFDT with respect to bare Au. However, upon spin coating the OSC 9,9-dioctylfluoreneco-benzothiadiazole (F8BT), UPS revealed that the hole-injection barrier was 0.6 eV for bare Au, 1.5 eV for DT/Au, and 0.7 eV for PFDT (Fig. 5b). That is to say, the barrier changed by +0.9 eV for DT and +0.1 eV for PFDT relative to bare Au. The unexpected large hole-injection barrier in PFDT/Au is most likely caused by pinning of the Fermi level at the interface at the polaron level of F8BT for substrate work-function values greater than 5.2 eV. The band alignment is depicted in Fig. 5c. This example indicates well the violation of vacuum-level alignment and the importance of the extra interface dipoles and the pinning of the substrate Fermi level. Similar to the aforementioned charge transfer in metal/OSC junction, the interactions between SAM/metals and OSCs can be strong enough to cause remarkable charge transfer, which could dramatically alter the injection barrier [78,79]. For instance, a benzenethiol SAM, which has strong electron-withdrawing trifluoromethyl (CF3) groups, induces electron transfer from the deposited copper(II) phthalocyanine (CuPc) and leads to electron accumulation in SAM/Au, causing a work-function shift from 5.2 eV to 4.1 eV [79]. Such a charge transfer was not observed

FIGURE 5

Comparison between metal work function of SAM/Au and the injection barrier at the OSC/SAM/Au interface. (a) Histogram of work-function change measured by SKPM taking the O2-plasma treated Au as the reference. (b) UPS spectra in the valence band region of F8BT films on different SAMs/Au with corresponding hole injection barrier (HIB). (c) The energy diagrams showing band alignment and the injection barriers. Reproduced with permission from [12]. Copyrights (2008).

when using a methyl (CH3)-terminated benzenethiol SAM. Hence, the injection barrier is better confirmed by measuring the binding energy of the OSC/metal surface by photoemission spectra (PES) or other techniques [25,80].

Tunneling through a SAM The injection in a OSC/SAM/metal structure is, in fact, not only a matter of the injection barrier discussed above; there are also two other steps, as shown in Fig. 6a. A tunneling process across the SAM molecules and a transport process through the bulk OSC layer to reach the channel take place before and after injection, respectively, processes that also contribute to the contact resistance. In

8 Please cite this article in press as: C. Liu, et al., Mater. Today (2014), http://dx.doi.org/10.1016/j.mattod.2014.08.037

MATTOD-405; No of Pages 18 RESEARCH

RESEARCH: Review

Materials Today  Volume 00, Number 00  September 2014

FIGURE 6

Impact of SAM layer on practical charge injection. (a) A schematic of the three steps of charge injection from the metal layer to the OSC layer. (b) Impact of chain length of SAM in tunneling: (b1) tunneling current as a function of chain length of the alkanethiols (odd-even effect), (b2) impact on injection length scale, (b3) surface potential, and (b4) field-effect mobility mFET. (c) Impact of SAM on morphology of OSCs: (c1) pentacene grown on bare Au, (c2) pentacene grown on p-terphenylmethanethiol/Au; (c3) dif-TESADT grown on PFBT/Au with h0 0 1i- and h1 1 1i-textured crystals in the channel, and (c4) dif-TESADT grown on PFBT/Au with only the h0 0 1i-textured crystals in the channel. Reproduced with permission from [83,84,90,91]. Copyrights (2011, 2007, 2005, 2012).

Step 1, the SAM molecule structure and chain length mainly determine the injection efficiency. The intensively studied alkanethiol SAMs have a single C–C bond alkane chain that is insulating and adds to resistance [81,82]. The tunneling current decreases with increasing carbon number and the current of even numbers is higher than that of odd numbers (‘odd-even effect,’ Fig. 6b1) [83]. In total, three competing effects need to be considered with OSC/ alkanethiol/metal structures [84,85]: (1) the injection barrier generally decreases with increasing n, although the details are complicated [86], (2) the tunneling distance increases with n [87], and (3) the length scale of interfacial disorder z decreases with n. Factors (1) and (3) dominate for short chain lengths, whereas (2) starts to have a noticeable impact for long chain lengths. Thus, the overall length scale for charge injection first decreases to a minimum and then increases (Fig. 6b2, triangles). Consequently, in OFET alkanethiols with small values of n, the voltage drop is eliminated near the source contact, creating a linear electrostatic

potential in the channel as compared to a large voltage drop at the source on bare Au or OFET alkanethiols with larger values of n (Fig. 6b3). Hence, the mFET of OFETs increases by a factor of ten as n increases from 0 to 8 and then decays exponentially at larger values of n (Fig. 6b4), which roughly obeys the classic Simmons’s law on tunneling [83]. In addition to the general trend, there are also fluctuations in mFET and current with even values of n exceeding those of odd values of n due to the aforementioned ‘odd-even effect’ [83,86,88]. Hence, to attain high performance, SAMs with an even number of carbons less than 10 could be optimum in terms of tunneling. As compared to alkanethiols, aromatic SAMs have phenyl rings that have more delocalized p–p stacking and are regarded to cause less resistance during tunneling. Demonstration using aromatic thiols also exhibits a trend that mFET decreases with an increasing number of phenylene units, probably indicating higher injection resistances [76]. Moreover, recently Minari et al. fabricated gold nanoparticles that possess a metal core 9

Please cite this article in press as: C. Liu, et al., Mater. Today (2014), http://dx.doi.org/10.1016/j.mattod.2014.08.037

MATTOD-405; No of Pages 18 RESEARCH

surrounded by aromatic molecules as the conductive ligand. The aromatic ligands enable good solubility and charge injection simultaneously, avoiding the thermal annealing process to remove the insulating ligands that is necessary for conventional nanoparticles. The resulting mFET is 7.9 cm2/V s after all roomtemperature printing of electrodes, semiconductors and dielectrics [89], demonstrating the promising future of applying aromatic units in contact materials well. RESEARCH: Review

Morphological evolution of OSCs induced by SAM Step 3, transport through the OSC, relies on the morphology of the OSC layer adjacent to the SAM-covered electrodes, especially for crystalline OSCs. In thermal evaporation or solution-casting processes such as spin-coating, deposition of the first few nanometers of the OSC layers are strongly affected by the chemical composition or surface affinity of the SAMs, which in turn affects the upper OSC film. The chemical composition mainly affects the interaction and induces extra nuclei, whereas the surface affinity affects the drying/crystallization time and thickness of the film. Both characteristics, that is, chemical composition and surface affinity, affect the orientation of the OSC molecules [38,90,92,93]. A classic example is pentacene, in which the orientation of the molecules and the size of the crystals strongly influences the substrate properties [90]. On bare Cu(1 1 0) and Au(1 1 1), the pentacene monolayer lies flatly due to the electronic interaction between the p-orbitals of pentacene and the empty d-orbitals of the metal atoms [94,95]. This is manifested as the strong out-of-plane bending vibrations of the C–H aromatic ring, which can be detected by reflection–absorption IR spectroscopy. In sharp contrast, on the Au surfaces coated by alkanethiol and aromatic SAMs, the pentacene molecules are almost perpendicular to the metal electrode, as indicated by the same but significantly reduced vibration mode. Interestingly, this vertical orientation occurs regardless of the terminal group of the SAM, whether a –CN, –CH3, –COOH, NH2, –OH, or C60 group. Furthermore, pentacene molecules showed generally larger crystal sizes with a lamellar-like surface morphology on the p-terphenylmethanethiol-modified Au (Fig. 6c1, c2) [90]. This is definitely different from pentacene films on bare Au having rod-shaped smaller grains. The improved crystallinity then reduces the density of the grain boundaries from the electrode to the channel, and is thus supposed to enhance local transport mobility m near contact in the injection path [96]. Remarkably, the improvement in crystallization can be especially significant when the grain size is comparable to the channel length of OFETs. In the work by Gundlach et al. [92], pentafluorobenzene thiol (PFBT)-covered Au electrodes induce the nucleation of fluorinated 5,11-bis(triethylsilylethynyl) anthradithiophene (diFTESADT) in the plane of the film, enabling the formation of plate-like crystals. Crystals near the contact edges were found to grow 10 mm or more into the channel. Consequently, mFET in bottom-gate, bottom-contact OFETs is 3–5 times higher for the OFETs with L < 25 mm as compared to nontreated Au contacts. The mechanism was further revealed by a recent study by Li et al. [91], who used microbeam grazing incidence wide-angle X-ray scattering (mGIWAXS) to precisely map the polycrystalline textures in the channel. The pure h0 0 1i texture of diF-TES-ADT extended 25 mm away from the PFBT/Au electrodes into the channel and was capable of bridging channels up to 50 mm. Beyond this channel

Materials Today  Volume 00, Number 00  September 2014

length or using fast spin-coating, formation of the h1 1 1i grains in the central part of the channel acts as a significant bottleneck to carrier transport (Fig. 6c3, see the peak intensity of h1 1 1i). By waiting an additional 5 s before spin coating, growth of h0 0 1itextured domains from the PFBT/Au electrodes extended such that nucleation and growth of the h1 1 1i-textured grains was inhibited (Fig. 6c4, see the absence of peak h1 1 1i), resulting in enhanced performance. The contact-induced crystallization is probably because of the interaction between the sulfur atoms in the thiophene rings of the diF-TESADT and the PFBT-treated Au rather than the surface energy of the SAM [92]. Such crystallization has also been observed in TIPS-pentacene deposited on Cu contacts with the same PFBT SAM [69]. For crystalline polymers, Noh et al. [61] also reported the improvement of a thiophene-based conjugated polymer morphology on metal electrodes and even in the channel with PFDTmodified Au electrodes. Besides morphology, the thickness of the OSC film also plays an important role in injection, especially in a staggered configuration, as carriers are injected perpendicular to the channel. Most OSCs have isotropic charge transport, and the transport mobility perpendicular to the channel is much lower than that along the channel [97,98] which adds significantly to the contact resistance [99,100]. Therefore, the surface affinity of the SAM layer can affect the thickness of the OSC film, thus tuning the contact resistance. For instance, on a 1-DT-modified Au surface, the thickness of a spin-coated F8BT film is 20–30 nm, reduced from the 40 to 50 nm thickness on an oxygen-plasma-etched Au surface. This leads to the improvement of both electron and hole injection and reduces the contact resistance to nearly half the original value [12].

Insertion layer The insertion of a charge injection layer at metal electrode/OSC interface is another commonly used approach to improve contact properties not only in OFETs but also in organic light-emitting diodes (OLEDs) and organic photovoltaics (OPVs). Various materials that have been used as insertion layers to date are primarily metal oxides, inorganic salts and organic compounds (small molecules or polymers). Here, we concentrate on the metal oxides and discuss the common mechanisms for improving injection, which are also applicable to inorganic salts and organic compounds as discussed later on.

Metal oxides Metal oxides, in particular transition-metal oxides, have been extensively studied as contact interlayers to alleviate the contact limitation in organic devices [13,101–107]. Metal oxides show good processing flexibility and great capability in charge injection tuning. They can be deposited by solution-based coatings (e.g., V2O5 [108,109], MoO3 [110]) or by conventional routines such as vacuum evaporation (e.g., MoO3 [13], Al2O3 [105]) and oxygen plasma treatment (e.g., CuO [107]). Metal oxides cover a wide spectrum of electrical properties from insulating to semiconducting and metallic with a large range of work functions [111]. These unique features make them useful for tuning the injection/extraction properties of both electrons and holes in different devices. Two examples of using an oxide interlayer to improve the performance of OFETs are shown in Fig. 7. For p-type devices, Kano et al. [13] reported that inserting a thermally evaporated MoOx layer

10 Please cite this article in press as: C. Liu, et al., Mater. Today (2014), http://dx.doi.org/10.1016/j.mattod.2014.08.037

MATTOD-405; No of Pages 18 RESEARCH

RESEARCH: Review

Materials Today  Volume 00, Number 00  September 2014

FIGURE 7

Using an insertion layer to improve OFET injection: (a)–(d) MoOx for p-type and (e)–(h) TiOx for n-type. (a) Schematic representation, (b) transfer characteristics, (c) contact and channel resistance, and (d) inverse subthreshold slope. (e) Schematic representation, (f ) transfer characteristics, (g) contact resistance as a function of wm, and (h) FET mobility mFET as a function of wm. The dashed lines indicate the effect of insertion of oxide layer. Reproduced with permission from [13,14]. Copyrights (2009, 2009).

between Au electrode and C8-BTBT semiconductor resulted in enhanced current level, lowered contact resistance (Rc), and improved inverse subthreshold swing (Fig. 7a–d). For n-type devices, Cho et al. spin coated TiOx layers onto [6,6]-phenyl-C61 butyric acid methyl ester (PC61BM) and [6,6]-phenyl-C71 butyric acid methyl ester (PC71BM) semiconductors before depositing the metal electrodes. They observed much improved output current, reduced Rc and higher mFET in the OFETs with several contacting metals of Ag, Al, Cr and Au (Fig. 7e–h) [14]. Despite these encouraging results, the origin of these improvements by incorporating metal oxide interlayers, especially in OFETs [105,112], is still not well understood. However, some clues can be recognized in considering the mechanism of contact transport presented above.

Work-function tuning Metal oxide interlayers have shown great capability to tune the wm of metallic electrodes (e.g., Au, Ag, Al, Cu, Mo, Ti, ITO) to more than 2 eV [111]. In fact, the wm value of a conductor is the result of both body and surface effects, and the surface effects strongly depend on its surface properties, which in turn can be influenced by the contacting material. By choosing the parent metal and properly controlling the formation conditions, metal oxides can behave as insulators (e.g., CuO, NiO), n-type (e.g., MoO3, TiO2, V2O5) and p-type (e.g., Cu2O, Cr2O3) semiconductors, and conductors (e.g., MoO2, WO2, TiO). When metal oxides come into intimate contact with a metal, the surface contribution to the metal’s wm is altered by adsorption, charge transfer and reaction, such that the electrode’s wm is changed. Note that one cannot attribute this alteration simply to one or a few specific causes since various effects may be concomitantly involved in contact formation when adding a thin layer of metal oxide. Moreover, metal

oxides have a very wide range of work functions spanning from 3.0 eV for ZrO2 to 7.0 eV for V2O5 [113]. This wide range would be one of the reasons for the excellent tunability of the electrode’s wm by means of charge exchange upon adsorption. In general, the low-wm metal oxides are used as a charge-injection interlayer (or modifier) in n-type OFETs (e.g., TiO2 [114], ZnO [115,116], ZrO2 [117]), whereas high-wm metal oxides are applied to p-type OFETs (e.g., MoO3 [118,119] WO3 [120], CuO [103], V2O5 [121,122]). The magnitude of such wm tunability is found to sometimes depend on the thickness of the oxide interlayer [111] (Fig. 8), probably due to the charge transfer and reaction between the electrode and oxide (and/or the surrounding atmosphere) that leads to a gradual change of contact properties. In Fig. 8, chemical interactions between the electrode metals (Ni, Mo, V) and the oxide buffer layer leads to an oxide cation reduction and decreased wm within the first few nanometers of the metal/oxide interface. There is no reaction between Au/MoO3, so that the observed wm variation would be due to a charge-transfer interaction in which Mo5+ cations are stabilized by screening from the Au Fermi level. The very different behaviors seen for Cu are because it forms an oxide alloy with MoO3. The performance of such OFETs (e.g., mFET and Rc) thus depends on the thickness of the oxide interlayer. An optimal thickness is often observed in the range of 1–10 nm [115,123]. At very small thicknesses (e.g., <0.5 nm), it is difficult to form a conformal and continuous interlayer, and its influences on charge injection cannot take effect completely. At greater thicknesses (e.g., >10 nm), the negative influences arising from greater roughness, high trap density, low conductivity and a very different band structure start to play greater roles; also, injection by tunneling is eliminated. If the oxide is insulating (e.g., Al2O3), the injection barrier will increase rather rapidly as the interlayer is 11

Please cite this article in press as: C. Liu, et al., Mater. Today (2014), http://dx.doi.org/10.1016/j.mattod.2014.08.037

MATTOD-405; No of Pages 18 RESEARCH

Materials Today  Volume 00, Number 00  September 2014

RESEARCH: Review FIGURE 8

(a) Oxidation state profiles measured by XPS and (b) work-function profiles measured by UPS for MoO3 films grown on various metal electrodes. Reproduced with permission from [111]. Copyrights (2013).

getting thickened, due to formation of metal–insulator–semiconductor structure resulted from the very wide band gap of those oxides.

OFETs [15]. Once the tunneling injection is dominant, the barrier predicted by the Schottky–Mott rule does not matter anymore. As one result, the contact resistance is significantly reduced and found to be no longer strongly gate-voltage dependent but rather

Doping mechanism Doping is an important mechanism that can improve charge injection by increasing direct charge tunneling. As has been presented above, a space charge region is usually created within the OSC film when metal is brought into contact with an organic semiconductor because of their different Fermi levels. This region depletes any mobile charges so that the charges are injected only by thermal activation through an energetic barrier (by thermionic emission) if this depletion zone is thick (e.g., >10 nm) because the tunneling rate decreases exponentially with increasing barrier thickness [124]. It turns out that the width of the space charge 1=2 region can be reduced by doping OSCs ð  ND Þ to enhance injection by tunneling. In Si MOSFETs, the bulk silicon in the contact area is degenerately doped and tunneling injection from contact metal into silicon is essential. However, in OFETs, most OSCs are intrinsic semiconductors and are used without further doping. This means that they have a very low charge concentration at thermal equilibrium due to their wide band gaps. Such low charge concentration results in a rather thick depletion layer close to the metal/semiconductor interface and injection by tunneling is suppressed [37]. Including a hole- or electron-rich interlayer of ptype (e.g., CuO [125,126]) and n-type semiconducting (e.g., TiO2) metal oxides between contact metal and OSC can locally dope the contact area, thus decreasing the depletion-layer thickness to be as thin as a few nanometers, which is a reachable distance for tunneling (field emission) [112]. In addition, local contact doping improves the contact transport in the bulk OSC by screening the trap states present there (Fig. 9a, b). This mechanism is similar to that of other contact dopants of metal salts (e.g., FeCl3) in

FIGURE 9

(a) Schematic energy diagram of the metal–organic semiconductor contact without doping, where the depletion layer is very thick and hole injection is by thermal activation through a high-energy barrier. (b) Energy diagram of contact in a p-type OFET with acceptor doping, where the depletion thickness decreases, the tunneling injection becomes predominant, and trap states are screened. (c) TLM plots for a group of FeCl3-doped OFETs for contact resistance extraction. (d) Comparison of contact resistance for undoped and doped OFETs. Reproduced with permission from [15]. Copyrights (2012).

12 Please cite this article in press as: C. Liu, et al., Mater. Today (2014), http://dx.doi.org/10.1016/j.mattod.2014.08.037

MATTOD-405; No of Pages 18 Materials Today  Volume 00, Number 00  September 2014

Fermi-level depinning and buffering effect Interfacial gap states should be taken into account in discussing the roles of metal oxides as contact interlayers. In general, the metal/OSC interfaces comprise many defects that could produce a high density of interface states distributed over a large energy range within the band gap of OSC. It is possible that high-density interface states (or generation–recombination centers) pin the Fermi level of the contact electrode and the energy misalignment by unmatched work function is not important. In such a case, a thin metal oxide interlayer can depin the Fermi level and reduce the injection barrier [123]. Note that with this mechanism, contact resistance initially decreases and then increases with increasing thickness of the oxide layer, the same as the effect of work-function tuning mentioned before. Moreover, with an insertion layer, the interface reaction and charge transfer can also generate a large number of interfacial gap states, such as MoO3/Al [119], which could assist charge carriers in hopping via these gap states. For an ideal contact without any gap state the charge carriers have to overcome the original energy barriers in order to be injected from the metal into the OSC. However, this scenario is changed by the presence of available energy states within the band gap, and the charge carriers can hop (just once or several times) and finally attain their transport band via these energy ‘ladders’ [33,124,128] (Fig. 1b, ‘injection via gap-states’). Finally, one needs to consider another effect of the oxide interlayer: buffering. For top-contact OFETs, the oxide interlayer is deposited prior to contact metallization. The thermally evaporated metal tends to diffuse into the underlying organic semiconductor film at slow deposition rates and accumulate close to the interface at high deposition rates. Metal diffusion introduces impurities and defects into the organic semiconductor, and metal accumulation damages the upper surface of organic semiconductor film. These two mechanisms adversely affect the contact transport in the bulk semiconductor and at the interface [129,130]. Putting a thin layer of metal oxide or naturally oxidization of the contact metal can protect the fragile organic semiconductor from these two negative impacts during contact metallization, thereby obtaining better contact for more efficient charge injection [125].

Inorganic salts and organic compounds Beside metal oxides, many other materials have been tried as contact interlayers, including inorganic salts (e.g., CsCO3 [16,131], CsF [127], NaF [132]), organic compounds of small molecules (e.g., tetrakis(dimethylamino)ethylene, TDAE [133]) and polymers (e.g., polyethylenimine ethoxylated (PEIE), branched polyethylenimine (PEI) [17], and some polyelectrolytes [134]). The roles of these interlayers fall into the same scope as the metal oxides described previously. For instance, CsF produces a dipole of 0.3 eV in its nanoclusters, which equivalently reduces the electrode’s work function. Meanwhile, CsF acts as an n-type dopant and the electron injection is further improved with more tunneling [127]. In the scope of organic compounds, work-function tuning by polymers has been well demonstrated in the case of PEIE and PEI reported by Zhou et al. [17]. Owing to the inclusion of aliphatic amine groups, PEIE and PEI induce large molecular dipoles upon physisorption onto a conductor surface, which decreases the work function of a wide variety of conductors covering metals, conductive metal oxides, and conducting polymers, as well as graphene. Various devices (organic solar cells, OFETs, OLEDs, and even IGZO-based oxide TFTs) incorporating such PEIE and PEI interlayers showed greatly enhanced electron injection/extraction and much improved device performance in ntype operating regimes (e.g., Vth and electron mFET for n-type OFETs) (Fig. 10). This is believed to be an effective and universal

FIGURE 10

(a) Chemical structure of PEIE and PEI. (b) Photoemission cutoff obtained via UPS for PEDOT:PSS PH1000, ITO, and Au samples with and without PEIE. The dashed line indicates the shift of Fermi level. (c) Device structure of P(NDI2OD-T2) (i.e., N2200) OFETs where a 1.5-nm-thick contact interlayer of PEIE is highlighted. (d) Transfer characteristics of two N2200 OFETs with and without PEIE interlayer. Reproduced with permission from [17]. Copyrights (2012). 13

Please cite this article in press as: C. Liu, et al., Mater. Today (2014), http://dx.doi.org/10.1016/j.mattod.2014.08.037

RESEARCH: Review

nearly constant with the gate voltage (Fig. 9c, d) [13,15], indicating an injection process almost free from hopping [29]. Yet note that local doping in unipolar OFETs may degrade on-off ratio in longterm running due to the uncontrollable dopant diffusion. In conventional Si MOSFETs with significant wb, Ohmic contact is achieved by local ion-implantation of the contact area with dopants opposite to the channel. The contact area (e.g., p-type) and the silicon bulk (n-type) form a natural p–n junction, and the built-in potential and the post-implantation annealing help to stabilize the dopants and thus the formed Ohmic contacts. Yet in OFETs, typically when using intrinsic semiconductors, the contacts would not be protected by such mechanisms, possibly leading to contact degradation and off current enhancement upon long-term storage or operation. On the other hand, for ambipolar OSCs which originally have a low on-off ratio, directly doping the bulk OSC with effective dopants such as CsF (n-type) and F4-TCNQ (p-type) can improve injection properties and shift Vth, without significantly changing the on-off ratio [16,127].

RESEARCH

MATTOD-405; No of Pages 18 RESEARCH

way to obtain low-work-function electrodes for n-type OFETs. Another interesting organic interlayer is DNA, which is a natural polyelectrolyte and has been successfully applied to n-type PC70BM OFETs as an electron-injection layer [135]. Also, in ambipolar OFETs, both electron and hole injection is improved by more than three times due to combined dipole layers within DNA, which alters its orientation upon applying different gate biases. As far, examples of using organic compounds for enhancing injection via gap-states (as the case of MoO3 insertion) are still rare. RESEARCH: Review

Contacts made from graphene and carbon nanotubes Graphene-based electrodes Graphene (GR) features high mechanical strength and flexibility [136], low electric sheet resistance (down to tens of V/& upon chemical doping) [137], and high optical transmittance (up to 97.7%) to visible light [138], which makes it exceptionally suitable as an electrode or insertion layer between metals and OSCs. For more information on the synthesis of GR or graphene oxide (GO) and their application in general electronics, readers can refer to the

Materials Today  Volume 00, Number 00  September 2014

report by Pang et al. [138]. Here, we focus on graphene-related techniques and mechanisms for enhancing charge injection in OFETs. Different from thermally evaporated metals, chemical vapor deposition (CVD)-grown GR films need to be mechanically transferred onto insulating substrates, followed by patterning. The transfer by adhesion layer, including poly(methyl methacrylate) (PMMA), polydimethylsiloxane (PDMS), and tapes, and patterning via photolithography without degrading the electrical conductivity of the GR is crucial for OFET charge injection (Fig. 11). As shown in Fig. 11a top [139], fabrication of monolayer GR electrodes includes the following steps: (i) CVD-grown GR on Cu foil was spin-coated with PMMA and then floated in an ammonium persulfate solution to etch the Cu; (ii) the GR with PMMA was laminated onto insulating substrates (e.g., SiO2/Si) and washed by acetone to remove the PMMA (sometimes followed by annealing); (iii) the GR on the substrates was spin-coated with photoresist and patterned by photolithography to define the channel and electrodes; (iv) the exposed GR was removed by reactive ion

FIGURE 11

Schematic illustration of electrode fabrication by (a) transferring and patterning graphene [18,139] and (b) reduced graphene oxide (GO) or graphite oxide [140,142]. (c) Using self-release layer on PMDS stamp for transferring graphene onto soft surfaces [148]. Reproduced with permission from [18,139,140,142,148]. Copyrights (2011, 2011, 2011, 2009, 2013). 14 Please cite this article in press as: C. Liu, et al., Mater. Today (2014), http://dx.doi.org/10.1016/j.mattod.2014.08.037

MATTOD-405; No of Pages 18 RESEARCH

RESEARCH: Review

Materials Today  Volume 00, Number 00  September 2014

FIGURE 12

Effect of graphene electrode on the contact injection of OFETs. (a) Graphene reduces the injection barrier to pentacene as compared to Au electrodes. (b) Graphene electrodes with/without residual PMMA affect the morphology of pentacene growth and thus injection. Reproduced with permission from [18,149]. Copyrights (2011, 2011).

etching (RIE) plasma, the substrate treated with a hydrophobic SAM, and the photoresist lifted off; (v) OSCs are deposited onto the substrates (e.g., pentacene). Alternatively, as shown in Fig. 11a bottom [18], Ni is used as a supporting substrate in step (i) and used as a protection layer in step (iii). All the procedures described above maintain high transparency (>95%) and good conductivity for OFET applications. The same transferring and patterning techniques are also applicable to reduced graphene oxide (GO) electrodes (Fig. 11b, top, ¨ bkenberg et al. [140]). The difference is that the GO layers were Wo deposited by Langmuir–Blodgett (LB) assembly on a water surface, transferred onto pre-exposed photoresist, and then lifted off by a developer. Thus, the photoresist remains under the GO layers as a supporting layer. Furthermore, reduced GO is compatible with the spin-coating process, thus allowing low-cost, large-scale fabrication [141,142]. Reduced GO needs to be annealed at 10008C in order to render conductivity (Fig. 11b, bottom, Pang et al. [142]). Although reduced GO usually suffers from large numbers of electron traps due to sp3 carbons and vacancies [138,143,144], the solution process used to form electrodes highly encourages the exploration of new methods such as printing [145,146]. The same applies to printable liquid-phase exfoliated GR flakes [147], which have a higher resistance than monolayer GR. In addition, as an important forward step, in order to allow transfer of GR or GO onto arbitrary, soft substrates, Song et al. [148] reported a process using a self-release layer between GR and PDMS stamps (Fig. 11c). The polymeric self-release layer not only protects the GR from

contamination from the PDMS stamp, but it also enables the use of a variety of solvents orthogonal to the destination substrates. Hence, the GR layer can be easily transferred onto any surface, including organic thin films, two-dimensional materials, and prefabricated structures, in order to form electrical contact or multilayer films. These techniques have great potential for fabricating OFETs. The use of GR electrodes for enhancing contact injection in OFETs has mainly been attributed to tuning the injection barrier. Pentacene OFETs with monolayer GR electrodes exhibit higher current than those with Au electrodes (Fig. 12a1, Lee et al. [18]). Although the expected barrier height at vacuum alignment is 0.38 eV for pentacene/graphene (Pen/GR) and 0.1 eV for pentacene/Au (Pen/Au) interfaces, the interface dipoles (Dw) measured by Kelvin probe are in fact 0.09 eV for Pen/GR and 0.44 eV for Pen/Au. The vacuum shift leads to a hole-injection barrier (wb) of 0.38  0.09 = 0.29 eV for Pen/GR and 0.1 + 0.44 = 0.54 eV for Pen/ Au, as shown in Fig. 12a3. The values of wb measured by fitting current–voltage (Id–Vd) at different temperatures to the thermionic emission model are 0.24 eV for Pen/GR and 0.45 eV for Pen/Au (Fig. 12a4). It is explained that the different values obtained using Id–Vd and Kelvin probe measurements as being the result of the barrier-lowering effects of transport charges in pentacene OFETs, such as image potential lowering and rounding-off of the barrier corners. In addition, similar to the mechanism of using SAMs on metal, morphological changes can also play critical roles in using GR electrodes, especially when there are significant residual 15

Please cite this article in press as: C. Liu, et al., Mater. Today (2014), http://dx.doi.org/10.1016/j.mattod.2014.08.037

MATTOD-405; No of Pages 18 RESEARCH

Materials Today  Volume 00, Number 00  September 2014

RESEARCH: Review

polymers remaining after the transferring and patterning process. In the work by Lee et al. [149], these residues separate the GR and pentacene rings and induce a vertical molecular orientation when growing pentacene film (Fig. 12b1). However, on a clean GR surface (obtained by annealing at 5008C), because the molecular structures of pentacene and GR are very similar to that of benzene rings, the p–p interaction between them facilitates the assembly of pentacene molecules in a horizontal orientation and the epitaxial growth of crystals (Fig. 12b2). Because the edge-on orientation is favored by lateral hole transport in pentacene films, the GR with residues results in lower Rc, higher current, and higher mFET (Fig. 12b3, b4). This demonstration also indicates the value of using graphene electrodes as templates in growing organic crystals. Besides the study of p-type OSCs (pentacene [18,139,142], P3HT [150], CuPc [151,152]), electrodes made from reduced GO have been proven to be suitable for both electron and hole injections. Becerril et al. [153] demonstrated that in bottom contact devices with reduced GO electrodes (wm = 5.01 eV), both p- and n-channel OFETs displayed lower contact resistance and nearly 10 times higher mobilities than similar OFETs with Au electrodes. The ptype semiconductors were pentacene and poly(didodecylquaterthiophenealt-didode-cylbithiazole) (PQTBTz-C12), whereas the n-type semiconductor was copper(II) hexadecafluoro-phthalocya¨ bkenberg et al. [140] used reduced GO as the nine (F16CuPc). Wo electrodes (wm = 4.5 eV) and the semiconductor methanofullerene ([6,6]-phenyl-C61-butyric acid ester, PC61BM) to construct ambipolar OFETs, and they demonstrated a complementary inverter circuit. Both studies were performed using solution processing and both demonstrated the potential for low-cost, flexible electronics.

Carbon nanotube electrodes Because Valitova et al. [154] provided a decent review on the recent application of carbon nanotubes (CNTs) in OFETs, here we only briefly compare CNTs and GR as electrodes for OFETs. Similar to GR electrodes, CNTs have a medium work function (4.8 eV) and

thus CNT arrays are suitable for both hole and electron injection [155,156]. Also, in the growth of pentacene films, CNT substrates affect the molecular orientation of pentacene, so that edge-on pentacene is favorable for carrier transport [157]. Thus, graphene and CNTs are both suitable as an electrode or as insertion layers between OSCs and metal electrodes. The main difference is that GR films are planar, whereas CNT electrodes have a large length/ diameter ratio. As a result, the electric field at the tip of CNTs can be a hundred times higher than that of planar structures, which is expected to improve charge injection by assisting the tunneling process from CNTs into OSCs [154]. This mechanism occurs regardless of carrier type and should promote tunneling of both electrons and holes. In addition, recent progress in separating conducting and semiconducting CNTs by polymer wrapping techniques opens further possibilities for optimizing CNT electrodes for OFETs.

Summary and outlook We briefly compare all the contact engineering methods discussed in Table 2. From the viewpoint of fabrication process, SAMs provide easy fabrication at room temperature and a wide range of work-function tuning, making them suitable for both n- and ptype semiconductors. For high-temperature processed semiconducting materials or good thermal stability, for example, for OSCs that need annealing and for oxide semiconductors, inserting an oxide layer between the contact metal electrodes and the semiconductors would be preferable. Graphene, GO, and CNTs are light and thermally stable, but they require complicated fabrication techniques for coating and patterning. Their carbon-based features may be well suited for OSCs, but further study is needed to understand the interaction between them and OSCs. The use of FETs for different applications, such as electronic, biological, or chemical, put distinct requirements on FET contact engineering. In addition to fine-tuning contact materials to improve injection, injection can also be significantly improved by blending OSCs with high-conductivity materials such as CNTs [158] or by

TABLE 2

Comparison of contact engineering materials and methods. Contact

SAM/metal

Oxide/metal

Salt or polymer/metal

GR or GO

CNT

Fabrication

Simple solution process (spin-coating/vapor deposition/soaking)

Simple solution process or thermal deposition

Simple solution process or thermal deposition

CVD or solution-phase exfoliation

CVD

Patterning

No need, grown on metals only

Evaporation through masks

Evaporation through masks or photolithography

Transfer and photolithography

Spin-coat and photolithography

Device structure

Bottom contact (BC)

BC or top contact (TC)

BC (possible for TC)

BC (possible for TC)

BC

Work-function

3.5–5.7 eV (on Au)

3.0–7.0 eV (on Au)

4.5–5.0 eV

4.5–5.0 eV

4.8 eV

Injection barrier-tuning window

Medium (2 eV)

Wide (4 eV)

Medium (2 eV)

Narrow

Narrow

Conductivity

High

High (conductor) Low (insulator)

High

High (GR) Low (GO)

High

Tunneling resistance

Low (tunable)

Medium (tunable)

Low

–

Low

Flexibility

High

Medium

High

High

High

Thermal stability

Low

High

Low

High

High

Transparency

High

Medium

High

High

High

Weight

Light

Medium or heavy

Light

Light

Light

16 Please cite this article in press as: C. Liu, et al., Mater. Today (2014), http://dx.doi.org/10.1016/j.mattod.2014.08.037

MATTOD-405; No of Pages 18

using high-k insulators, for example, PVDF-TrFE [159] and ionic liquid gels [160,161] (Rc of only several V cm) as a way of enhancing carrier density and conductivity. Together with the techniques discussed above, contact engineering is broadly applicable for FETs with semiconductors made from other semiconductors, such as two-dimensional materials (e.g., graphene, MoS2), oxides, SAMs, nanowires, nanoparticles, and quantum dots. Also, the techniques are useful for improving other electronic devices such as lightemitting diodes (LEDs) and photovoltaic (PV) devices. Determining suitable materials and mechanisms remain challenging and an attractive topic for future research.

Acknowledgements This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korean Government (MSIP) (NRF2014R1A2A2A01007159), by the Center for Advanced SoftElectronics (2013M3A6A5073183) funded by the Ministry of Science, ICT & Future Planning and the Dongguk University Research Fund of 2014. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44]

H.N. Tsao, et al. J. Am. Chem. Soc. 133 (2011) 2605–2612. C. Liu, et al. Adv. Mater. 23 (2011) 523–527. I. Kang, et al. J. Am. Chem. Soc. 135 (2013) 14896–14899. Y. Yuan, et al. Nat. Commun. 5 (2014) 4005. H.-R. Tseng, et al. Adv. Mater. 26 (2014) 2993–2998. S.M. Sze, Physics of Semiconductor Devices, 2nd ed., John Wiley & Sons, Inc., New York, 1981. J.C. Scott, G.G. Malliaras, Chem. Phys. Lett. 299 (1999) 115–119. D. Natali, M. Caironi, Adv. Mater. 24 (2012) 1357–1387. D. Boudinet, et al. Org. Electron. 11 (2010) 227–237. K. Singh, et al. ACS Appl. Mater. Interface 3 (2011) 2973–2978. J. Youn, et al. Adv. Funct. Mater. 22 (2012) 1856–1869. X. Cheng, et al. Adv. Funct. Mater. 19 (2009) 2407–2415. M. Kano, et al. Appl. Phys. Lett. 94 (2009) 143304. S. Cho, et al. Adv. Funct. Mater. 19 (2009) 1459–1464. T. Minari, et al. Appl. Phys. Lett. 100 (2012) 093303. K.J. Baeg, et al. ACS Appl. Mater. Interfaces 3 (2011) 3205–3214. Y.H. Zhou, et al. Science 336 (2012) 327–332. S. Lee, et al. Adv. Mater. 23 (2011) 100–105. Z. Bao, J. Locklin, Organic Field-effect Transistors, CRC Press, Boca Raton, FL, USA, 2007. Z. Wang, et al. Appl. Phys. Lett. 95 (2009) 043302–043303. J. Hwang, et al. Mater. Sci. Eng. R 64 (2009) 1–31. L. Lindell, et al. Appl. Phys. Lett. 102 (2013) 223301. X. Crispin, et al. J. Am. Chem. Soc. 124 (2002) 8131–8141. A. Kahn, et al. J. Polym. Sci., Part B: Polym. Phys. 41 (2003) 2529–2548. S. Braun, et al. Adv. Mater. 21 (2009) 1450–1472. K.-Y. Wu, et al. Langmuir 25 (2009) 6232–6238. Y. Xu, et al. J. Appl. Phys. 107 (2010) 114507. T. Minari, et al. Appl. Phys. Lett. 88 (2006) 083513–083514. C. Liu, et al. Appl. Phys. Lett. 104 (2014) 013301. Y. Xu, et al. Appl. Phys. Lett. 97 (2010) 063302. C. Liu, et al. J. Phys. Chem. C 117 (2013) 12337–12345. Y. Xu, et al. Org. Electron. 12 (2011) 2019–2024. P.V. Pesavento, et al. J. Appl. Phys. 96 (2004) 7312–7324. T.J. Richards, H. Sirringhaus, J. Appl. Phys. 102 (2007) 094510. X.R. Li, et al. Phys. Rev. Lett. 108 (2012) 066601. I. Lange, et al. Phys. Rev. Lett. 106 (2011) 216402. T. Minari, et al. Appl. Phys. Lett. 91 (2007) 053508. O. Acton, et al. Adv. Funct. Mater. 21 (2011) 1476–1488. Y. Xu, et al. Appl. Phys. Lett. 97 (2010) 033503. Y. Xu, et al. IEEE Electron Dev. Lett. 34 (2013) 1298–1300. A.A. Balandin, Nat. Nanotechnol. 8 (2013) 549–555. V.K. Sangwan, et al. Nano Lett. 13 (2013) 4351–4355. C.H. Kim, et al. IEEE Electron Dev. Lett. 32 (2011) 1302–1304. C. Liu, et al. Chem. Mater. 22 (2010) 2120–2124.

RESEARCH

[45] [46] [47] [48] [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59] [60] [61] [62] [63] [64] [65] [66] [67] [68] [69] [70] [71] [72] [73] [74] [75] [76] [77] [78] [79] [80] [81] [82] [83] [84] [85] [86] [87] [88] [89] [90] [91] [92] [93] [94] [95] [96] [97] [98] [99] [100] [101] [102] [103] [104] [105] [106] [107] [108] [109] [110] [111] [112] [113] [114] [115] [116]

Y.L. Shen, et al. Phys. Rev. Lett. 86 (2001) 3867–3870. T. Richards, H. Sirringhaus, Appl. Phys. Lett. 92 (2008) 023512–023513. A.J. Kronemeijer, et al. Adv. Mater. 26 (2013) 728–733. G. Horowitz, et al. J. Appl. Phys. 87 (2000) 4456–4463. B. Stadlober, et al. Adv. Funct. Mater. 17 (2007) 2687–2692. S.D. Wang, et al. Appl. Phys. Lett. 93 (2008) 043311. Q.J. Cai, et al. Org. Electron. 9 (2008) 14–20. G. Heimel, et al. Acc. Chem. Res. 41 (2008) 721–729. B. de Boer, et al. Adv. Mater. 17 (2005) 621–625. D.M. Alloway, et al. J. Phys. Chem. C 113 (2009) 20328–20334. K. Demirkan, et al. J. Chem. Phys. 128 (2008) 074705. N. Gozlan, et al. J. Phys. Chem. C 112 (2008) 12988–12992. Q. Sun, A. Selloni, J. Phys. Chem. A 110 (2006) 11396–11400. P. Marmont, et al. Org. Electron. 9 (2008) 419–424. K. Asadi, et al. J. Mater. Chem. 17 (2007) 1947–1953. Q.J. Cai, et al. Org. Electron. 9 (2008) 936–943. Y.Y. Noh, et al. Semicond. Sci. Technol. 26 (2011) 034003. F. Gholamrezaie, et al. Synth. Met. 161 (2011) 2226–2229. K. Asadi, et al. Adv. Mater 21 (2009) 4109–4114. C.-Y. Chen, et al. Org. Electron. 12 (2011) 148–153. Y.Y. Noh, et al. Nat. Nanotechnol. 2 (2007) 784–789. O.T. Hofmann, et al. Nano Lett. 10 (2010) 4369–4374. J.-P. Hong, et al. Appl. Phys. Lett. 92 (2008) 143311. D. Ota´lvaro, et al. J. Phys. Chem. C 116 (2012) 7826–7837. C.H. Kim, et al. ACS Appl. Mater. Interface 5 (2013) 3716–3721. C.D. Zangmeister, et al. Chem. Phys. Lett. 442 (2007) 390–393. R.A. Hatton, et al. Thin Solid Films 394 (2001) 291–296. J. Park, et al. J. Phys. Chem. Lett. 2 (2011) 841–845. L. Romaner, et al. Phys. Rev. B 77 (2008) 045113. D. Cornil, et al. Adv. Funct. Mater. 17 (2007) 1143–1148. L.F. Ah Qune, et al. Appl. Phys. Lett. 93 (2008) 083103–083109. N. Crivillers, et al. Adv. Mater. 23 (2011) 1447–1452. N. Crivillers, et al. Adv. Mater. 25 (2013) 432–436. W. Chen, et al. J. Phys. Chem. B 110 (2006) 26075–26080. W. Chen, et al. Adv. Funct. Mater. 17 (2007) 1339–1344. M. Zhou, et al. Appl. Phys. Lett. 101 (2012) 013501–013504. V.B. Engelkes, et al. J. Am. Chem. Soc. 126 (2004) 14287–14296. C.-C. Kaun, H. Guo, Nano Lett. 3 (2003) 1521–1525. M.M. Thuo, et al. J. Am. Chem. Soc. 133 (2011) 2962–2975. P. Stoliar, et al. J. Am. Chem. Soc. 129 (2007) 6477–6484. S. Casalini, et al. Org. Electron. 13 (2012) 789–795. Y.-T. Tao, et al. Org. Electron. 12 (2011) 602–608. A. Salomon, et al. Adv. Mater. 15 (2003) 1881–1890. F. Tao, S.L. Bernasek, Chem. Rev. 107 (2007) 1408–1453. T. Minari, et al. Adv. Funct. Mater. (2014), http://dx.doi.org/10.1002/ adfm.201400169. W.S. Hu, et al. Langmuir 21 (2005) 2260–2266. R. Li, et al. Adv. Mater. 24 (2012) 5553–5558. D.J. Gundlach, et al. Nat. Mater. 7 (2008) 216–221. H.S. Lee, et al. J. Am. Chem. Soc. 130 (2008) 10556–10564. S. Lukas, et al. Phys. Rev. Lett. 88 (2002) 028301. J.H. Kang, X.-Y. Zhu, Appl. Phys. Lett. 82 (2003) 3248–3250. J. Rivnay, et al. Nat. Mater. 8 (2009) 952–958. J. Northrup, Appl. Phys. Lett. 99 (2011) 062111–062113. A. Troisi, J. Chem. Phys. 134 (2011) 034702. K. Sakamoto, et al. Appl. Phys. Lett. 100 (2012) 123301. Y. Xu, et al. J. Appl. Phys. 113 (2013) 064505–064507. C.W. Chu, et al. Appl. Phys. Lett. 87 (2005) 193508. D. Kumaki, et al. Appl. Phys. Lett. 92 (2008) 013301. G.B. Murdoch, et al. Appl. Phys. Lett. 93 (2008) 083309. G.Z. Nie, et al. Org. Electron. 12 (2011) 429–434. P. Darmawan, et al. Appl. Phys. Lett. 100 (2012) 013303. X. Yu, et al. Synth. Met. 162 (2012) 936–940. W. Hu, et al. Appl. Phys. Lett. 80 (2002) 2640–2641. K. Zilberberg, et al. Adv. Energy Mater. 1 (2011) 377–381. Z.A. Tan, et al. Phys. Chem. Chem. Phys. 14 (2012) 14589–14595. J. Meyer, et al. Adv. Mater. 23 (2011) 70–73. M.T. Greiner, et al. Adv. Funct. Mater. 23 (2013) 215–226. A. Kumatani, et al. Sci. Rep. 3 (2013) 1026. M.T. Greiner, et al. Adv. Funct. Mater. 22 (2012) 4557–4568. H.J. Bolink, et al. Adv. Funct. Mater. 18 (2008) 145–150. H.J. Bolink, et al. Appl. Phys. Lett. 91 (2007) 223501. M.C. Gwinner, et al. Adv. Funct. Mater. 20 (2010) 3457–3465.

17 Please cite this article in press as: C. Liu, et al., Mater. Today (2014), http://dx.doi.org/10.1016/j.mattod.2014.08.037

RESEARCH: Review

Materials Today  Volume 00, Number 00  September 2014

MATTOD-405; No of Pages 18 RESEARCH

RESEARCH: Review

[117] [118] [119] [120] [121] [122] [123] [124] [125] [126] [127] [128] [129] [130] [131] [132] [133] [134] [135] [136] [137] [138] [139]

M. Vasilopoulou, et al. Org. Electron. 14 (2013) 312–319. T. Matsushima, et al. Appl. Phys. Lett. 91 (2007) 253504. Y. Yi, et al. J. Chem. Phys. 130 (2009) 094704. Z.B. Wang, et al. Phys. Rev. B 80 (2009) 235325. J. Meyer, et al. J. Appl. Phys. 110 (2011) 033710. K.J. Baeg, et al. ACS Appl. Mater. Interfaces 5 (2013) 5804–5810. Z.H. Liu, et al. Phys. Rev. B 82 (2010) 035311. J.C. Scott, J. Vac. Sci. Technol. A 21 (2003) 521–531. C.A. Di, et al. Adv. Mater. 20 (2008) 1286–1290. X.B. Lu, et al. Appl. Phys. Lett. 98 (2011) 243301. D. Khim, et al. J. Mater. Chem. 22 (2012) 16979–16985. Y. Hirose, et al. Phys. Rev. B 54 (1996) 13748–13758. F. Faupel, et al. Mater. Sci. Eng. R: Rep. 22 (1998) 1–55. ¨ rr, et al. Adv. Mater. 14 (2002) 961–963. A.C. Du M.H. Chen, C.I. Wu, J. Appl. Phys. 104 (2008). J. Lee, et al. Appl. Phys. Lett. 82 (2003) 173–175. W. Osikowicz, et al. Appl. Phys. Lett. 85 (2004) 1616–1618. J.H. Seo, et al. J. Am. Chem. Soc. 131 (2009) 18220–18221. Y. Zhang, et al. Adv. Mater. 24 (2012) 4255–4260. C. Lee, et al. Science 321 (2008) 385–388. A.C. Neto, et al. Rev. Mod. Phys. 81 (2009) 109. S. Pang, et al. Adv. Mater. 23 (2011) 2779–2795. W.H. Lee, et al. Adv. Mater. 23 (2011) 1752–1756.

Materials Today  Volume 00, Number 00  September 2014

[140] [141] [142] [143] [144] [145] [146] [147] [148] [149] [150] [151] [152] [153] [154] [155] [156] [157] [158] [159] [160] [161]

¨ bkenberg, et al. Adv. Mater. 23 (2011) 1558–1562. P.H. Wo S. Lim, et al. J. Phys. Chem. C 116 (2012) 7520–7525. S. Pang, et al. Adv. Mater. 21 (2009) 3488–3491. K. Erickson, et al. Adv. Mater. 22 (2010) 4467–4472. C. Go´mez-Navarro, et al. Nano Lett. 10 (2010) 1144–1148. S. Wang, et al. Nano Lett. 10 (2009) 92–98. K.-Y. Shin, et al. Adv. Mater. 23 (2011) 2113–2118. E.B. Secor, et al. J. Phys. Chem. Lett. 4 (2013) 1347–1351. J. Song, et al. Nat. Nanotechnol. 8 (2013) 356–362. W.H. Lee, et al. J. Am. Chem. Soc. 133 (2011) 4447–4454. Y. Cao, et al. Adv. Funct. Mater. 19 (2009) 2743–2748. W. Liu, et al. ACS Nano 4 (2010) 3927–3932. Y. Cao, et al. Angew. Chem. Int. Ed. 122 (2010) 6463–6467. H.A. Becerril, et al. ACS Nano 4 (2010) 6343–6352. I. Valitova, et al. Nanoscale 5 (2013) 4638–4646. F. Cicoira, et al. Appl. Phys. Lett. 98 (2011) 183303. F. Cicoira, et al. ACS Nano 5 (2010) 283–290. M.A. McCarthy, et al. ACS Nano 5 (2010) 291–298. M. Wang, et al. Org. Electron. 15 (2014) 809–817. K.-J. Baeg, et al. Adv. Mater. 24 (2012) 5433–5439. D. Braga, et al. Appl. Phys. Lett. 97 (2010) 193311. J.H. Cho, et al. Nat. Mater. 7 (2008) 900–906.

18 Please cite this article in press as: C. Liu, et al., Mater. Today (2014), http://dx.doi.org/10.1016/j.mattod.2014.08.037