Flexible, high mobility short-channel organic thin film transistors and logic circuits based on 4H–21DNTT

The low mobility and large contact resistance in organic thin-film transistors (OTFTs) are the two major limiting factors in the development of high-performance organic logic circuits. Here, solution-processed high-performance OTFTs and circuits are reported with a polymeric gate dielectric and 6,6 bis (trans-4-butylcyclohexyl)-dinaphtho[2,1-b:2,1-f]thieno[3,2-b]thiophene (4H–21DNTT) for the organic semiconducting layer. By optimizing and controlling the fabrication conditions, a high saturation mobility of 8.8 cm2 V−1 s−1 was demonstrated as well as large on/off ratios (> 106) for relatively short channel lengths of 15 μm and an average carrier mobility of 10.5 cm2 V−1 s−1 for long channel length OTFTs (> 50 μm). The pseudo-CMOS inverter circuit with a channel length of 15 μm exhibited sharp switching characteristics with a high signal gain of 31.5 at a supply voltage of 20 V. In addition to the inverter circuit, NAND logic circuits were further investigated, which also exhibited remarkable logic characteristics, with a high gain, an operating frequency of 5 kHz, and a short propagation delay of 22.1 μs. The uniform and reproducible performance of 4H–21DNTT OTFTs show potential for large-area, low-cost real-world applications on industry-compatible bottom-contact substrates.


Results
OTFT devices with solution processed 4H-21DNTT films. For this work, as illustrated in Fig. 1c, bottom-gate bottom-contact (BG-BC) OTFTs were fabricated on both rigid glass and 125-μm thick flexible polyethylene naphthalate (PEN) substrates. Patterned aluminum (50 nm) gate electrodes were deposited on cleaned substrates through a shadow mask. Parylene C was selected as the organic gate dielectric layer because it is well known for its environmental stability, low dielectric constant, low defect density, and optimal surface energy for solution shearing with many solvents, and hence is an appropriate selection for bottom gate OTFT configuration 13,31,32 . Parylene C was deposited using chemical vapor deposition (CVD) on substrates with patterned gate electrodes. Independently, the relative dielectric constant for Parylene C was estimated to be ε r = 3.1 from capacitance measurements on a metal insulator metal (MIM) device structure. Au electrodes were deposited through a shadow mask with a channel width (W) of 380-400 μm and various channel lengths (L) ranging from 10 to 150 μm. In order to facilitate charge injection by modification of the contact work function the gold electrodes were treated with 2,3,4,5,6-pentafluorothiophenol (PFBT) thiol-SAM prior to semiconductor deposi- www.nature.com/scientificreports/ tion. A thin film of 4H-21DNTT semiconductor (Fig. 1b) was deposited using a solution shearing process with a thin glass blade (the schematic of the OSC deposition process is shown in Fig. 1d). The glass blade was coated with a thin Teflon layer to make the surface hydrophobic; this prevents OSC from adhering to the blade during solution shearing. In solution shearing technique the deposition parameters such as substrate temperature, concentration of organic semiconductor (OSC) and blade speed have a major influence on the crystalline film formation. Previous reports on optimization of process parameters for meniscus-guided coating techniques have shown that to achieve highly crystalline film the optimal coating speed depends on the substrate temperature and the choice of OSC/solvent system [33][34][35][36][37][38] . In this work, for 4H-21DNTT film optimization we used two concentrations of 4H-21DNTT (0.1 wt% and 0.09 wt%) in o-dichlorobenzene (o-DCB) solvent. The small amount of 4H-21DNTT solution (6 μL) was inserted in the small gap (∼ 100-200 μm) between blade and Parylene C coated substrates (shearing setup in Fig. 1d). 4H-21DNTT film was deposited at 90 ℃ substrate temperature for different blade scanning speeds. Figure 2 shows the polarized microscope images of 4H-21DNTT film under different processing conditions. All images were taken under same light intensity therefore the film brightness suggests the thickness of the film. At higher coating speeds voids are observed in the film due to faster scanning speed than the rate of solvent evaporation 37 . The voids size is reduced as the coating speed decreases and matches closely with solvent evaporation rate, at a blade speed of 12.5 μm/s highly ordered 4H-21DNTT film was obtained. For higher solute concentration the 4H-21DNTT film thickness was increased and thick aggregates were formed in the film (Fig. S3). The thick aggregates in the semiconductor film are undesirable as it can lower the performance of OTFT device 34 . The OTFT characteristics for various OSC deposition condition (Fig. S2) reflects the OSC morphology dependence. The OTFT devices for films with thick aggregates and voids showed poor performance and lower mobility. The solution shearing speed was found to be very critical in achieving the good OTFT performance and void free crystalline 4H-21DNTT film. The best optimized condition (0.09 wt% concentration and 12.5 μm/s blade speed) was selected to achieve a uniform single-crystal 4H-21DNTT thin film with a thickness of a few molecular layers to further investigate the OTFT performance from here on. The solution-sheared 4H-21DNTT thin film on Parylene C-coated glass substrate was investigated by X-ray diffraction (XRD) to better understand its structure and crystallinity (Fig. 1a). The intense and sharp peaks in out-of-plane measurements show the high crystallinity of 4H-21DNTT thin films on the glass/Parylene C substrate. The XRD peaks for 4H-21DNTT can be assigned to (00l) reflections. The crystal growth direction corresponds to the crystallographic c-axis of 4H-21DNTT molecules, i.e., nearly perpendicular to the substrate. The interlayer distance (d-spacing) estimated from the XRD data was 30.1 Å for 4H-21DNTT. The thickness of the mono-layer of 4H-21DNTT estimated from the cross-sectional AFM profile (Fig. 1a, inset) was approximately 2.9 nm, which is consistent with the single molecular length estimated from XRD. The 4H-21DNTT exhibited similar properties, such as the XRD profile and absorption spectra (Fig. S1), to those of the previously reported 5H-21DNTT organic semiconductor. www.nature.com/scientificreports/ OTFT device characterization. Figure 3a shows a cross-polarized optical microscope image of 4H-21DNTT on BG-BC OTFT on glass substrate, showing that the large single-crystal morphology of 4H-21DNTT. On rotating the polarizer and analyzer angle by 45°, completely black images were obtained, indicating that the 4H-21DNTT films had good crystallinity and were uniformly oriented. The electrical performance characteristics of 4H-21DNTT BG-BC OTFT fabricated on a glass substrate are summarized in Fig. 3. The surface around the channel region was scratched mechanically to reduce the possibility of fringe currents 39 , and all the measurements were performed at room temperature (25 ℃ to 27 ℃), in air, and under dark conditions. Figure 3b shows the output characteristics (I D vs. V D ) for different gate voltages of the 4H-21DNTT OTFT with channel length (L) 80 μm and width (W) 400 μm. Figure 3c shows the transfer characteristics of the 4H-21DNTT OTFT (W/L = 400/80) in the saturation regime (V D = − 20 V) and linear regime (V D = − 1 V); negligible hysteresis was observed in both the output and transfer curves. The saturation (μ sat ) and linear (μ lin ) mobilities were calculated from the slopes of the fitted lines for the plots of (I D ) 1/2 vs. V G and I D vs. V G , respectively. For the 4H-21DNTT OTFT device (Fig. 3c), the on/off ratio was greater than 10 7 and in the saturation region (V DS = − 20 V) the mobility (μ sat ) is estimated to be 13.0 cm 2 V −1 s -1 with 99.6% reliability factor (r) 40 and a near zero threshold voltage (V TH ) of 0.4 V. In the linear region (V DS = − 1 V) mobility (μ lin ) was estimated to be 13.6 cm 2 V −1 s -1 with 98.2% reliability factor (r) and a threshold voltage (V TH ) of 0.2 V. A plot of mobility (in the saturation and linear regions) vs. gate voltage (Fig. 3d) shows a broad flat region over a large gate voltage range, which indicates hardly any nonlinear contact issues. Furthermore, the gate-to-source leakage current (I GS ) was less than 10 −10 A. The sub-threshold swing (SS) was determined to be 158.5 ± 3 mV/decade by fitting the exponential region of the drain current in the sub-threshold regime (Fig. 3b). Figure S8 shows the electrical characteristics of the 4H-21DNTT OTFT transistor after the substrate had been exposed to ambient air and room temperature for 30 days. The aged device has shown only 14.1% decrease in mobility and a threshold voltage shift of 0.08 V after 30 days of continuous exposure to air and light. Figure   www.nature.com/scientificreports/ with high on-off ratios (> 10 7 ), onset voltages close to 0 V and nearly gate-independent saturation mobility of 9.4 cm 2 V −1 s -1 for long channel (85 μm) and 6.33 cm 2 V −1 s -1 for short channel (15 μm) devices as shown in Fig. 4e.
High mobility short channel OTFT and contact resistance analysis. The output characteristics do not show any apparent nonlinearity at small V DS voltages, which suggests that the contact resistance of the 4H-21DNTT OTFT device type is low. However, the contact effects become more pronounced and critical for reducing the channel length. The contact resistance (R C ) of our bottom-contact OTFTs was evaluated in the linear regime using the transmission line method (TLM). For this purpose, we fabricated 4H-21DNTT OTFT with varying channel lengths ranging from 10 to 150 μm. The total resistance (R Total ) between the source and drain electrodes in the TLM is defined as the sum of the channel resistance (R ch ) and contact resistance (R C ) at the source and drain electrodes 15,16 .
The contact resistance was estimated from the intercept of the linear fit of the channel width normalized total resistance (R Total W) vs. channel length (Fig. 5a) and calculated using the following equation 15,16 : where μ o is the intrinsic mobility and C i is the insulator capacitance. From the slope of the above equation, the intrinsic mobility (i.e., without the influence of contact resistance) of 4H-21DNTT was estimated. The evolution of the channel width normalized contact resistance (R C W) as a function of gate overdrive voltage is plotted in Fig. 5b. The channel width normalized contact resistance (R C W) for the 4H-21DNTT device is a minimum of ∼1.01 kΩcm and 2.3 kΩcm at V G of − 20 V for glass and flexible substrates, respectively, and then typically increases with smaller |V G |. Despite R C W values below 100 Ωcm have been demonstrated for BG-BC OTFT with ultra-thin gate dielectric layer 41 , our R C W values are quite low compared with previously reported for a solution www.nature.com/scientificreports/ processed BG-BC OTFT with a thick gate dielectric layer 42 . The literature of R C W values for coplanar OTFT is shown in Figure S6a. In OTFTs, the energy level misalignment between contacts and OSC is major contributor for higher contact resistance 13,15,43 . The work function of untreated Au contacts was measured to be 4.8 eV which created a hole injection barrier with the HOMO level of 4H-21DNTT (5.31 eV) and resulted in larger contact resistance of 9.4 kΩcm at − 20 V (Fig. S7). For PFBT treated Au contacts the work function was measured to be 5.58 eV, which suggested an efficient charge injection and nearly one order lowering of contact resistance. We concluded that the larger contact resistance with untreated Au contacts is owing to injection barrier due to energetic mismatch, since no major differences were observed in the morphology of the film as a function of PFBT treatment. Another important factor influencing the contact resistance is the morphology of OSC near the contacts as charge injection occurs at the edge of the channel. Several studies have demonstrated that when OSC have non ideal morphology such as large number of small grains and grain boundaries the trap density is significantly increases leading to a higher contact resistance 44 . At the S/D and OSC interface a smooth crystalline morphology of 4H-21DNTT was observed from polarized microscope images (Fig. 3a) and AFM topography images (Fig. S2a); therefore, we speculate that the contact resistance values were not influenced by the traps originating from OSC microstructure. This indicates that the cumulative effect of an improved charge injection due to PFBT-treated S/D electrodes and fewer charge trapping sites at the semiconductor and gate dielectric interface due to highly crystalline and smooth 4H-21DNTT films, contribute to a reduction in the R C value and nearly ideal OTFT characteristics. Further reduction of the contact resistance for bottom-contact devices might be possible with very thin gate dielectric layers 12,41,45 , however, discussion on gate dielectric thickness is beyond the scope of this report. The maximum and average intrinsic mobility (μ o ) for 4H-21DNTT BG-BC OTFT was 15.1 cm 2 V −1 s −1 and 12.5 cm 2 V −1 s −1 respectively. The μ o for 4H-21DNTT is close to the saturation and linear mobility extracted from the slope of the transfer characteristics because of a relatively low contact resistance in our devices. In order to address the reproducibility and uniformity of the solution-processed 4H-21DNTT OTFTs, the statistical distribution (standard error) of the extracted mobilities on glass and flexible PEN substrates for each channel length (10-150 μm) are compared in Fig. 5e. For both flexible and glass substrates, the statistics were www.nature.com/scientificreports/ calculated from more than 10 devices at each channel length. For all channel lengths, the average mobility of 4H-21DNTT devices was greater than 5 cm 2 V −1 s −1 . The statistical variation of 4H-21DNTT OTFTs with L ranging from 50 to 150 μm for over 50 transistors is shown in Fig. S5. The mean mobility achieved for long channel length devices (L ≥ 50 μm) was 11.1 cm 2 V −1 s −1 and a standard deviation of 1.7. The standard deviation in OTFT performance can be further reduced by using fully automated shearing system with precise control over deposition parameters such as flow rate, shape of the meniscus for each device which is not possible with our current available facilities. The drop in mobility with decreasing channel length indicates that the contact resistance effect becomes more pronounced for short channel length devices. The transfer characteristics of a short channel OTFT with a 15 μm channel length is shown in Fig. 5c (transfer characteristics for each channel length are shown in Fig. S4). The mobility (in the saturation and linear regions) vs. gate voltage plot (Fig. 5d) shows a broad flat region with slight gate voltage dependence, which indicates hardly any nonlinear contact issues even for 10 μm channel length. The highest extracted field-effect mobilities for 15 μm and 10 μm channel lengths (L) were 8.8 and 6.9 cm 2 V −1 s −1 respectively, the comparison of the 4H-21DNTT field effect mobility for short channel OTFTs (L ≤ 50 μm) with the previously reported results is shown in Fig. 1e. Even with the contact resistance of 1.01 kΩ the mobility value achieved for 4H-21DNTT is comparable to that of the top-performing p-type OTFTs reported in the previous literature [12][13][14][15][16][17][18][19][20][41][42][43] (as shown in the graph in Fig. 1e). Furthermore, as the transition or cut-off frequency (f T ) which determines the highest frequency that can be amplified by a single OTFT scales approximately with μ o /L 212 , Fig. S6b shows the literature over view of μ o /L 2 values. We believe that the operational frequency performance of 4H-21DNTT devices can be further increased by lowering the R C W value, operational voltage and scale the device dimensions which is a challenge for coplanar solution processed OTFTs with organic gate dielectric. However, despite high contact resistance the 4H-21DNTT demonstrated competent short channel OTFT performance.

Inverter logic circuit performance based on 4H-21DNTT OTFT.
To further assess the robustness and applicability of our 4H-21DNTT OTFTs in complex integrated circuits, we first evaluated the inverter logic gate, which is the basic component of digital circuits. The modern CMOS design style is unsuitable for OTFTbased circuits because of the large difference in the performance and stability of p-type versus n-type OTFTs. Therefore, for this work, we employed a pseudo-CMOS configuration as it uses unipolar TFTs [28][29][30]66 . A schematic of the pseudo-CMOS inverter with zero-VGS load logic (pseudo-D) is shown in Fig. 6a. The zero-VGS load logic requires depletion mode TFTs (i.e., normally ON TFT devices at V GS = 0 V or a positive on-set voltage for p-type OTFT). Its advantage is its high noise margin, which allows more circuit robustness against process variations compared to diode-load logic design. The pseudo-CMOS inverter is composed of four p-type OTFTs (M1, M2, M3, and M4) with varied W/L ratios (Fig. 6a). The channel length of each OTFT was l5 μm, the channel widths for transistors M1, M2, M3, and M4 were 300, 300, 600, and 600 μm, respectively, and the gate-tosource and gate-to-drain overlap (L OV ) was 30 μm. An image of the fabricated pseudo-CMOS inverter is shown in Fig. 6b.  Figure 6d shows the transient characteristics of the inverter on the glass substrate using a 30 V square wave input signal (V in ) with a frequency of 5 kHz and a bias voltage (V DD = V SS ) of 20 V (flexible inverter Fig. S11). The average high-to-low and low-to-high propagation delays (τ pd ) (Figs. S10, S13) of our inverter circuits were 6.9 μs and 18.8 μs, respectively. The rise (τ r ) and fall (τ f ) time constants were determined by fitting exponential functions to the measured transient output voltage signal.

NAND logic circuit performance based on 4H-21DNTT OTFT.
To further demonstrate the potential of 4H-21DNTT OTFT and assess the reproducibility of our fabrication process, we studied the performance of a pseudo-CMOS based NAND circuit, which is a basic building block to configure D-FF, shift registers, and counters. A schematic of the pseudo-CMOS logic-based NAND circuit is shown in Fig. 7a. The NAND circuit is composed of six p-type OTFTs (T1, T2, T3, T4, T5, and T6) with various W/L ratios. The channel length of each OTFT was l5 μm, the channel widths for transistors T1, T2, T3, T4, T5, and T6 were 300, 300, 600, 600, 1200, and 1200 μm, respectively, and the gate-to-source and gate-to-drain overlap (L OV ) was 30 μm. An image of the fabricated NAND circuit is shown in Fig. 7b. Figure 7c  www.nature.com/scientificreports/ high-and low-input V B (i.e., V B = 0-30 V), the output voltage is always logic high (V OUT > 0.2 V), demonstrating good NAND device characteristics. The dynamic performance of the NAND logic circuit was evaluated by applying two square-wave input signals (V A and V B ) of amplitude 30 V; V A with a frequency of 0.5 kHz and V B with a frequency of 5 kHz (Fig. 7d, S16b). The average high-to-low and low-to-high propagation delays (τ pd ) of the NAND circuits were measured to be 26.2 μs and 22.1 μs, respectively (Figs. S15, S18). Rise and fall times of 46.6 μs and 56.8 μs for glass substrate were observed for a supply voltage of 20 V (Figs. S14, S17). The combination of further reductions in contact resistance and an enhancement in mobility for short-channel OTFT devices is expected to yield even higher dynamic logic circuit performance.

Discussion
In the presented work, we have successfully demonstrated 4H-21DNTT bottom-contact OTFTs fabricated on flexible (PEN) and glass substrates with a low contact resistance of 1 kΩcm, and we recorded a high mobility of 8.8 cm 2 V −1 s −1 for a short channel length of 15 μm. High-quality 4H-21DNTT thin films with long single crystal domains on both flexible and glass substrates were obtained by temperature and speed-controlled solution shearing. We confirmed that high-quality OSC films in conjunction with an improved charge injection due to PFBTtreated Au electrodes resulted in lowering the contact resistance of the bottom-contact TFTs and enabled higher frequency operation in OTFT logic circuits. Furthermore, we fabricated a pseudo-CMOS architecture-based inverter and NAND logic circuit and obtained signal-propagation delays of 32.7 μs and 70.1 μs, respectively, at a bias voltage (V DD = − V SS ) of 20 V, for bottom-contact 4H-21DNTT TFTs fabricated on flexible PEN substrates. We demonstrated an improvement in short-channel 4H-21DNTT OTFT mobility along with a reduction in contact resistance, which is essential for the operation of high-frequency logic circuits. The demonstration of such performance and its achievement using a bottom contact solution processed OTFT with thick gate dielectric presents an achievable pathway towards all-printed, low-cost, industrial production of high resolution OTFTs for driving flexible displays, sensor chips, and RFID applications. In conclusion, we used a very simple solution shearing fabrication method for achieving crystalline thin films of 4H-21DNTT organic semiconductor. Highperformance OTFTs were demonstrated using these crystalline films with an intrinsic mobility of 15.1 cm 2 V −1 s −1 and good environmental stability of their electrical performance. The short channel performance of 4H-21DNTT www.nature.com/scientificreports/ OTFTs on both rigid and flexible substrates is comparable to the best results obtained in literature. This basic study of 4H-21DNTT OTFT properties and its successful application in inverter and NAND integrated circuits demonstrates a promising potential of 4H-21DNTT OSC for applications in flexible, low-cost electronics with further improvement in the contact resistance of OTFTs.
OTFT and logic circuit fabrication. The glass substrates were sequentially cleaned prior to the use by ultrasonication with detergent, deionized water, acetone, and isopropanol for 10 min each and blown dried with nitrogen (99.99%) and the substrates were further cleaned with UV ozone for 10 min. The same method was used to fabricate all OTFTs and inverter and NAND logic circuits on glass and PEN substrates presented in this work. Aluminium (Al) gate electrodes (50 nm) were vacuum deposited continuously through a shadow mask onto the substrates. Parylene C gate dielectric (≈ 550 nm) was deposited on Al gate electrodes using chemical vapor deposition (CVD) system. Gold (Au) source-drain (S/D) electrodes (  www.nature.com/scientificreports/ speed were optimized such that the solution drying speed is approximately the same as scanning speed (Fig. 2). During the solution shearing process, the same volume of solution was injected in the gap for each transistor in order to keep the meniscus shape constant which is important in achieving good reproducibility and less variation in the OTFT performance parameters.
Thin film characterization and OTFT electrical measurements. Film thickness and surface morphology were measured using AFM (SPI3800N SPA 500 AFM probe station). The out-of-plane XRD analysis of the 4H-21DNTT films was performed on a Rigaku Smartlab XRD with 9 kW X-ray power and the sample was scanned for 2θ values ranging from 0° to 25°. The Au work function and 4H-21DNTT HOMO level were measured using Riken Keiki AC3 photoelectron spectrophotometer in air (PESA). Crystal orientation and uniformity were characterized by a polarized optical microscope (Nikon Eclipse LV100ND). The OTFT and logic circuit device fabrication, storage and measurement were performed at ambient condition (relative humidity 65% and with controlled temperature ranging from 25 to 27 ℃). The transfer and output characteristics of the OTFT and logic circuit devices were measured at room temperature in air with a Keithley 4200 semiconducting parameter analyzer. Field-effect mobility was calculated in the saturation (V D = − 20 V) and linear regime (V D = − 1 V) from the linear fitting to plots of √ I D versus V G and I D versus V G, respectively using following equations: The reliability factor was calculated using following equation 40 , A Tektronix AFG1022 function generator, a Tektronix TBS1104 digital oscilloscope, and an Agilent 33502A amplifier were used to assess the transient performance of the inverters and NAND logic circuits.