Liquid crystal-like active layer for high-performance organic field-effect transistors

High carrier mobility and uniform device performance are of crucial importance for organic field-effect transistor (OFET)-based device and integrated circuit applications. However, strategies for achieving high device performance with small variations from batch to batch are still desired. Here, we report a thin liquid crystal-like film of 2,8-difluoro-5,11-bis(triethylsilylethynyl) anthradithiophene (dif-TES-ADT) grown on a N,N′-ditridecylperylene-3,4,9,10-tetra-carboxylic diimide (PTCDI-C13) template, confirmed by atomic force microscopy and polarized fluorescence microscopy. The liquid crystal-like films with large crystalline domains are further employed as carrier transport channels for OFETs. As a result, we achieved high-performance OFETs with a saturation carrier mobility of 1.62 ± 0.26 cm2 V−1 s−1 and a small variation of 16% among three batches. This finding provides a new strategy to design materials and device structures to simultaneously achieve high carrier mobility and device uniformity.


INTRODUCTION
Organic field-effect transistors (OFETs) are the basic building blocks of organic electronics to realize functions such as signal amplification and on-off switching by controlling the carrier transport in organic semiconductors near a dielectric surface via an external electric field [1]. Since the first polythiophene OFET was reported in 1987 [2], great progress has been witnessed for organic electronics owing to their intrinsic merits of mechanical flexibility, light weight and potential low material and device processing costs [3][4][5]. Basically, the carrier mobility and device performance uniformity are the greatest concerns for OFET applications in organic electronics. A high carrier mobility up to 43 cm 2 V −1 s −1 has been reported via the optimization of mole-cular design, processing technologies and interface engineering [6][7][8][9]. Despite the progress in the carrier mobility with small variations less than 20% reported in the literature [10][11][12], strategies with clear physical mechanisms that simultaneously achieve a high carrier mobility and a small variation in the device performance from batch to batch are still urgently sought.
To obtain high carrier mobility for high-frequency OFET devices, molecules with extended π systems are favored, with the formation of π-π stacking in crystalline domains during the film growth [13]. For a given small-molecule organic semiconductor and device configuration, the domain orientation, size and boundaries play crucial roles in the performance variation among devices [14]. For inorganic semiconductors such as Si and GaAs [15,16], the devices are processed with single-crystalline films epitaxially grown on single-crystalline substrates, which ensures the same orientation and no grain boundaries, thus achieving small variation in the device performance within one batch. Advanced film preparation techniques such as molecular beam epitaxy (MBE) [17] and metal organic chemical vapor deposition (MOCVD) [18] further guarantee the reproducibility of the devices from batch to batch under well-controlled growth parameters in a vacuum/inert gas environment. Unfortunately, owing to the absence of single-crystalline substrates, organic semiconductors are typically grown on amorphous surfaces with randomly distributed orientation domains. Much effort has been dedicated to controlling the orientation through template-directed growth either in vacuum or in solution as well as by employing external force/magnetic/electrical fields [12,19,20]. However, little progress has been reported on the device performance variation from batch to batch.
Liquid crystal-like films whose molecules behave similarly to a confined two-dimensional (2D) liquid component are also considered good materials for OFETs [21,22]. Hanna's research group [23][24][25] demonstrated that OFETs based on these liquid crystal-like films exhibited highly uniform device performance with a carrier mobility on the order of 10 −2 cm 2 V −1 s −1 .
Recently, a mobility variation of less than 10% was reported using the liquid crystalline molecule 2-decyl-7-phenyl-[1]benzothieno[3,2-b][1]benzothiophene (Ph-BTBT-10) based on a calculation involving 49 devices on 5 substrates [26]. With the ambition of developing a strategy for realizing OFETs with both high carrier mobility and uniform performance, we report here an organic liquid crystal-like interface by depositing 2,8difluoro-5,11-bis(triethylsilylethynyl) anthradithiophene (dif-TES-ADT) on a N,N′-ditridecylperylene-3,4,9,10-tetracarboxylic diimide (PTCDI-C 13 ) template. The idea is to take advantage of the "crystal" nature for efficient carrier transport and the "liquid" nature for isotropicity in thin films. As a result, the transistors achieve a high saturation carrier mobility of 1.62 ± 0.26 cm 2 V −1 s −1 , which is higher than that of the solution-processed dif-TES-ADT single-crystal arrays (1.5 cm 2 V −1 s −1 ) [27]. The saturation carrier mobility is almost 150 times higher than that for a standard OFET with dif-TES-ADT directly grown on a silicon substrate. Benefiting from the transport channel with a liquid crystal-like layer, a small batch-to-batch device variation of 16% is achieved.

Materials
Silicon wafers as substrates with a 300-nm thermal oxide layer (C i = 10 nF cm 2 ) were purchased from Si-Mat company. The wafers were ultrasonically cleaned for 10 min with acetone, ethanol, and deionized water and then dried with nitrogen before being treated with oxygen plasma at 200 W for 5 min. The samples were then transferred to a vacuum chamber for the deposition of organic molecules. The organic semiconductor molecules dif-TES-ADT (>99%) and PTCDI-C 13 (>95%) were purchased from Codex International and Sigma-Aldrich Chemie GmbH and used directly without further purification. Octadecyltrichlorosilane (OTS) was purchased from Sigma-Aldrich Chemie GmbH (≥90%).

Film deposition and device fabrication
For PTCDI-C 13 deposition, the substrate temperature was set at 250°C, and the sublimation temperature of the molecules was set at 260°C, with a deposition rate of 0.1-0.2 nm min −1 . The deposition was performed using a homemade instrument with a vacuum below 10 −5 Pa. The dif-TES-ADT molecules were then deposited on the substrate at room temperature, with a sublimation temperature of 120°C and a deposition rate of 0.05-0.1 nm min −1 .
For OFET fabrication, 40-nm Au electrodes were deposited on top of the dif-TES-ADT films using an Edwards E306 thermal evaporation system through a shadow mask with a channel length of 50 μm at 2 nm min −1 . During this process, tungsten wire was used as the heating source under a vacuum of 10 −4 Pa (the metal materials, which had a purity of 99.9%, were purchased from ChemPur). A bottom-gate top-contact transistor based on the dif-TES-ADT films was obtained.

Film and device characterization
Atomic force microscopy (AFM) measurements were performed in air by a Multimode NanoScope IIIa instrument (Digital Instrument) in tapping mode with an n + -silicon tip. The AFM images were analyzed by WSxM software. The molecule-resolution images of PTCDI-C 13 and dif-TES-ADT were taken in contact mode with a PPP_LFMR sensor (Nano & More) by using a Cypher ES (Oxford Instruments) in ambient conditions. The topography images of submonolayer and phase images were taken in "tapping mode" by using FS1500_AuD sensors (Oxford Instruments).
Fluorescence and polarization images of the films were obtained on a Zeiss Axioplan Fluorescence Microscope under blue and ultraviolet (UV) light.
The electrical performance of transistors was tested with a 7B Tungsten Probe using a Keithley 4200SCS semiconductor parameter analyser integrated with a Micromanipulator 6150 probe station from Micromanipulator Co., Inc.

Growth of the dif-TES-ADT on PTCDI-C 13 monolayer
Dif-TES-ADT (Fig. 1a) is a derivative of anthradithiophene with symmetrical triethylsilylethynyl groups at the peri-positions of fluorinated anthradithiophene [28]. It is a widely used p-type semiconducting molecule whose carrier mobility for solutionprocessed single-crystal arrays reaches approximately 1.5 cm 2 V −1 s −1 [27]. Very recently, a carrier mobility up to 5 cm 2 V −1 s −1 was reported for OFETs fabricated with ultrathin crystalline films [29]. However, the crystalline films were typically processed in solution, and none of the previous studies provides statistics of the batch-to-batch uniformity. Surprisingly, the physical vapor deposition (PVD) method, which bears intrinsic advantage for high uniformity and good reproducibility [30][31][32], is rarely reported for dif-TES-ADT thin film preparation. PTCDI-C 13 (Fig. 1b) is an n-type organic semiconductor material with a large π-electron system. Moreover, the long alkyl chain substituents on both sides make the molecule more flexible, which may promote orderly arrangement of molecules on a high-temperature substrate, thereby forming layered films [33]. OFETs based on PTCDI-C 13 monolayer show no field effect when operating at gate-source voltages from 20 to −50 V (Fig. S1), which indicates the PTCDI-C 13 would play no role in p-type OFET operation. Both dif-TES-ADT and PTCDI-C 13 thin films can be processed by PVD and solution methods. Here, we employed PVD in combination with template induction for the film preparation, achieving good reproducibility over large areas, with dif-TES-ADT as a semiconducting layer and PTCDI-C 13 as a template layer.
We used silicon wafers with 300-nm thermal oxide as substrates with a surface roughness of 0.15 nm, as shown in the AFM image in Fig. 1c. After depositing 2-nm dif-TES-ADT, which was monitored by a microbalance near the substrate, the morphology on the SiO 2 substrate shows discrete islands with lateral sizes and heights ranging from tens of nanometres to submicrometres and from 3 to 25 nm, respectively (Fig. 1d). The islands occupy 26% of the surface area with almost no connection to each other (Fig. S2a). As the film thickness increases to 15 nm, the dif-TES-ADT islands grow, connecting to adjacent islands and forming a network with a surface occupation ratio greater than 99% (shown in the AFM images of Fig. 1e and Fig. S2b). The large number of grain boundaries in the dif-TES-ADT films processed by PVD would result in low device performance and large variation of electronic devices, which is why almost all the dif-TES-ADT OFETs were processed by solution methods to obtain large crystalline domains [34,35].
Alternatively, the weak epitaxy growth method has been proven to be an efficient way to grow high-quality organic crystalline films on molecular templates [36,37]. For example, various semiconducting molecules can be grown on para-hexaphenyl (p-6P) monolayers, which take on a lamellar shape with excellent in-plane orientation within each p-6P domain [14,38]. When deposited on a SiO 2 surface, PTCDI-C 13 can form a smooth monolayer on SiO 2 at a substrate temperature of 250°C, as shown in Fig. 1f. The inserted high-resolution AFM (HR-AFM) image illustrates the excellent in-plane ordering with lattice parameters a = (0.48 ± 0.01) nm, b = (0.85 ± 0.01) nm, and γ = 80°± 1°, which are consistent with the crystal structure of PTCDI-C 13 reported in the literature [33]. The large domain with excellent molecular ordering of PTCDI-C 13 provides a good molecular template candidate for weak epitaxy growth of dif-TES-ADT to eliminate the large number of grain boundaries existing on SiO 2 . With the same amount of 2-nm dif-TES-ADT deposited on the PTCDI-C 13 submonolayer surface, the AFM image of Fig. 1g indicates a completely different morphology, exhibiting a uniform thickness of 1.6 nm and a surface occupation ratio of 59% (Fig. S2c). Further deposition of 15-nm dif-TES-ADT leads to a layered growth mode with a domain size up to several micrometers, as visualized by the AFM image shown in Fig. 1h. The HR-AFM image of multilayer dif-TES-ADT grown on PTCDI-C 13 reveals lattice parameters of a = (0.73 ± 0.01) nm, b = (0.75 ± 0.01) nm, and γ = 72°± 1°(inset in Fig. 1h), which are also in good agreement with the X-ray diffraction (XRD) data in Fig. S3 and the literature [39].

Liquid crystal-like dif-TES-ADT submonolayer grown on PTCDI-C 13
During AFM measurements on the submonolayer dif-TES-ADT grown on PTCDI-C 13 , we found that the contour of the domain changed over time, as shown in Fig. 2a-c. The as-grown sample presents a clear molecular layer of dif-TES-ADT over that of PTCDI-C 13 . In successive scans over 24 min, the rearrangement of dif-TES-ADT into bilayers and multilayers was observed, while the PTCDI-C 13 layer underneath remained unchanged (Fig. 2b, c). More importantly, the boundaries of the submonolayer dif-TES-ADT domains are movable, as indicated by the white arrows in Fig. 2b, c, implying

SCIENCE CHINA Materials
that there may be a softer fusion between adjacent crystalline domains without rigid boundary blocking, thus indicating a liquid-like behavior. To confirm the dynamic behavior of the vacuum-deposited film, we conducted AFM peeling-off experiments. By performing AFM imaging in the repulsive regime (phase slightly below 90°), dif-TES-ADT molecules were removed from the PTCDI-C 13 layer (Fig. S4). When the tipsample force was decreased to measure in the attractive regime, only a few areas of PTCDI-C 13 were found to be covered by dif-TES-ADT (Fig. 2d). Fig. 2g shows a cross section along the red line in Fig. 2d. Both materials are clearly identified based on their thickness. The thickness of the PTCDI-C 13 films on SiO 2 is 2.8 nm, while the dif-TES-ADT monolayer has a thickness of 1.7 nm, which is in agreement with the lattice parameter c of PTCDI-C 13 [33] and dif-TES-ADT crystals (Fig. S3). In addition, the phase images reveal a strong material contrast, with dif-TES-ADT in purple and PTCDI-C 13 in yellow. Successive frame up and frame down images were taken across the same sample area. The time for each frame was 1 min and 58 s. Video S1 displays the complete sequence of images. Fig. 2e, f show the selected topography and phase images. These images show that the dif-TES-ADT submonolayer becomes larger over time. The blue arrows in Fig. 2d, e indicate the direction of growth of the dif-TES-ADT submonolayer. The growth and reassembly behaviors of the dif-TES-ADT submonolayer reveal the high mobility of the dif-TES-ADT molecules on PTCDI-C 13 . In contrast, the monolayer of PTCDI-C 13 is stable and does not suffer any modification during the AFM peeling-off and imaging process.
As presented in Fig. 3a, rearrangement of the dif-TES-ADT submonolayer on PTCDI-C 13 was observed when the sample was directly exposed to air. After storage in air for 2 h, the AFM image shows the coexistence of mono-, bi-and multilayer dif-TES-ADT on the stable PTCDI-C 13 layer. The surface was further characterized by polarized fluorescence microscopy (PFM), as shown in Fig. 3b-i. When excited under blue light, the surface presents strong red emission from PTCDI-C 13 , decorated with weak yellow dif-TES-ADT emission, which can be distinguished in the upright inset of Fig. 3b. The PFM images (Fig. 3c-e) show a clear red light intensity modulation with changing polarization angle from 0°to 90°, confirming the good crystallinity of PTCDI-C 13 domains with random orientation. For better observation of the dif-TES-ADT layers, the excitation light was switched to UV to decrease the emission of PTCDI-C 13 owing to the low light absorption. The fluorescence microscopy image in Fig. 3f illustrates that all the mono-, bi-and multilayers of dif-TES-ADT can be well resolved. As marked in the upright inset of Fig. 3f, the submonolayer dif-TES-ADT emission intensity changes and exhibits birefringence with the angle of the polarizer, confirming the crystallinity of the dif-TES-ADT films. Combining this crystalline property with the mobility of molecules presented in Fig. 2, we conclude that the dif-TES-ADT submonolayer could be a liquid crystal-like film on the PTCDI-

Liquid crystal-like interface layers of dif-TES-ADT on PTCDI-C 13
After five days of exposure of the submonolayer dif-TES-ADT/ PTCDI-C 13 films (see Fig. 2) to ambient conditions, the dif-TES-ADT became multilayer films with thicknesses greater than 12 nm (Fig. 4a). The rearrangement of the dif-TES-ADT molecules from submonolayer to multilayer also indicates the high diffusivity of dif-TES-ADT on PTCDI-C 13 . The phase image in Fig. 4b also reveals the two organic components of the sample. Fig. 4b provides the phase image, giving different values for dif-TES-ADT (purple) and PTCDI-C 13 (yellow). To access the innermost layers, the dif-TES-ADT crystal was peeled off by AFM. Fig. 4c illustrates the peeling-off technique.
The area for peeling off is selected, and the set-point is decreased to operate in the high repulsive regime (step 1). Under this invasive AFM scan, some dif-TES-ADT layers are removed from the crystal. Later, the set-point amplitude is set close to the free oscillation amplitude to work in the attractive regime, and the sample is imaged with a larger scan size to check the new state of the crystal (step 2). Steps 1 and 2 can be done several times to reach more innermost layers. Fig. 4d displays a topography image after a few "peeling-off scans" on the sample shown in Fig. 4a. Further peeling-off scans enable us to create a cavity inside the dif-TES-ADT crystal that reaches the PTCDI-C 13 layer on the substrate surface, as shown in Fig. 4e.
Peeling-off scans were performed for dif-TES-ADT crystals with different thicknesses to investigate the structure of the films at the dif-TES-ADT/PTCDI-C 13 interface. Interestingly, none of the carried-out peeling-off experiments enabled us to visualize the very first layer of dif-TES-ADT on top of the PTCDI-C 13 layer in the repulsive regime with molecular resolution. Fig. 4f shows a high-resolution image of a cavity inside a dif-TES-ADT crystal, which shows the first detectable innermost layer of dif-TES-ADT. As shown in Fig. 4g, the first detectable layer of dif-TES-ADT above PTCDI-C 13 that could be imaged with high molecular resolution has a thickness of 5.3 nm. In both organic film regions, the phase in Fig. 4f shows linear features that seem to be parallel to the "b" vector of PTCDI-C 13 ((0.85 ± 0.02) nm) and either the "a" or "b" vector of dif-TES-ADT ((0.70 ± 0.02) nm). These values match the lattice parameters of the two molecules in Fig. 1f, g, but the values of a and b for dif-TES-ADT are so close that which vector corresponds to 0.70 nm in Fig. 4f is unclear. The distance between lines reasonably agrees with the "bsin(γ)" of PTCDI-C 13 and either the "asin(γ)" or "bsin(γ)" of dif-TES-ADT. The crystal orientation of dif-TES-ADT grown on PTCDI-C 13 was determined from Fig. 4f. In the first detectable layer, one of the vectors of dif-TES-ADT is rotated (10°± 3°) with respect to the vector "b" of PTCDI-C 13 . No notable changes in the lattice parameter or orientation were identified between the 1 st , 2 nd and 3 rd detectable layers of dif-TES-ADT. In addition, the step heights between the 1 st , 2 nd and 3 rd detectable layers are in good agreement with the lattice parameter "c" of the dif-TES-ADT crystals. Notably, the influence of PTCDI-C 13 on the bulk structure of dif-TES-ADT in the c-axis direction appears to be negligible, as shown in Fig. S3b.

SCIENCE CHINA Materials
Therefore, it is concluded that above 5.3 nm, the dif-TES-ADT on PTCDI-C 13 should be crystalline films, while the few layers directly in contact with PTCDI-C 13 are liquid crystal-like films, hampering the HR-AFM imaging.

Device fabrication and characterization
Samples with 15-nm dif-TES-ADT on PTCDI-C 13 were then fabricated for OFETs, whose thickness was monitored by a quartz balance near the sample holder. Bottom-gate top-contact transistors were fabricated to evaluate the electrical performance of the dif-TES-ADT films grown on PTCDI-C 13 . With the conclusion that a thin liquid-like layer exists on PTCDI-C 13 , the device structure is a layered structure of Au electrodes, dif-TES-ADT films in the bulk phase, dif-TES-ADT liquid crystal-like thin films, a PTCDI-C 13 monolayer and a SiO 2 /Si substrate, as schematically depicted in Fig. 5a. Carrier transport mainly occurs in the first several layers of the organic semiconductor films near the dielectric layer in a transistor [40,41]. Therefore, the electrical performance, such as the carrier mobility, is determined by the quality of the films at the interface, which in our device is a dif-TES-ADT liquid crystal-like thin film. Fig. 5b exhibits the typical transfer characteristic of an OFET measured in air. A saturation carrier mobility of 1.76 cm 2 V −1 s −1 is extracted from the transfer curve according to the equation: where I D is the current between the source and drain electrodes, μ is the carrier mobility, C i is the dielectric capacitance per unit area (10 nF cm −2 ), W and L are the width (1000 μm) and length (50 μm) of the device channel, respectively, V GS is the gatesource voltage and V th represents the threshold voltage. A high on/off ratio of 10 6 and a low threshold voltage of 5 V were obtained. The excellent electrical performance is comparable to that of the device based on single-crystal dif-TES-ADT films [27]. Compared with the carrier mobility of the device without PTCDI-C 13 (0.011 cm 2 V −1 s −1 ) shown in Fig. S5, the increased carrier mobility could be attributed to the large crystalline domains, good fusion of the thin liquid crystal-like dif-TES-ADT films at the interface and passivation of the dielectric layer surface by PTCDI-C 13 .
In addition, 150 devices in three batches, with 50 devices in each batch, were fabricated to evaluate the device performance uniformity. Fig. 5c-e show the carrier mobility, on/off ratio and threshold voltage of all 150 devices extracted from the corresponding transfer curves. The results show an average carrier mobility of 1.62 ± 0.26 cm 2 V −1 s −1 . Interestingly, the carrier mobility of the liquid crystal-like films is even above that of solution processed single-crystals with the same semiconducting molecules (1.5 cm 2 V −1 s −1 ) [27]. In the solution process, defects induced by residual organic solvents often deteriorate the carrier transport owing to carrier scattering [42,43]. In addition, similar to that of dielectric surface modification, the molecular template further passivates surface defects of the SiO 2 [44]. We believe both the absence of residual solvent-induced defects and the screening of dielectric surface defects play crucial roles for the improvement of the carrier mobility.
Furthermore, a small variation of 16% from batch to batch was obtained, which is equivalent to a recent report of 13% achieved within one batch [11]. The devices with an on/off ratio on the order of 10 5 and 10 6 account for 89%, and the devices with a threshold voltage less than 10 V account for 92%. To compare the effect of common self-assembled monolayers-treated dielectric surfaces on device performance, dif-TES-ADT OFETs grown on OTS-treated SiO 2 were fabricated as controls for comparison. As shown in Fig. S6, the dif-TES-ADT films grown on OTS-modified surface present poor continuity with an average carrier mobility of 0.99 ± 0.61 cm 2 V −1 s −1 and a variation of 62%. Analogous to the good device-to-device performance uniformity of OFETs with liquid-like crystals, we attribute the good device performance uniformity to the presence of the thin liquid-like dif-TES-ADT layer. The soft fusion between crystal domains with flexible boundaries in the layer may potentially create a carrier transport channel with weak anisotropy, which in turn leads to good device-to-device and batch-to-batch uniformity.

CONCLUSIONS
In summary, the presence of a thin liquid crystal-like interface of dif-TES-ADT films grown on the PTCDI-C 13 template is revealed by AFM and PFM characterizations. As confirmed by the peeling-off technique, the molecules within the liquid crys-tal-like interface are highly movable and can be easily manipulated by the AFM tip. The interfacial region is sandwiched between a rigid dif-TES-ADT film in the bulk phase and the PTCDI-C 13 monolayer, which can both be imaged by HR-AFM. Employing the thin liquid-like semiconducting interface for the active carrier transport layer results in an average saturation carrier mobility of 1.62 cm 2 V −1 s −1 . More importantly, a small variation of 16% among 3 batches was achieved by taking advantage of the liquid crystal-like dif-TES-ADT interlayer. In addition, the PTCDI-C 13 monolayer and dif-TES-ADT film can be deposited successively without breaking the vacuum, which could further fully take the advantage of vacuum deposition for high uniformity over a large area. The findings provide a new way to design materials and device structures, which could simultaneously achieve high carrier mobility and device uniformity from batch to batch. interest.