Wafer-scale, layer-controlled organic single crystals for high-speed circuit operation.

Two-dimensional (2D) layered semiconductors are a novel class of functional materials that are an ideal platform for electronic applications, where the whole electronic states are directly modified by external stimuli adjacent to their electronic channels. Scale-up of the areal coverage while maintaining homogeneous single crystals has been the relevant challenge. We demonstrate that wafer-size single crystals composed of an organic semiconductor bimolecular layer with an excellent mobility of 10 cm2 V-1 s-1 can be successfully formed via a simple one-shot solution process. The well-controlled process to achieve organic single crystals composed of minimum molecular units realizes unprecedented low contact resistance and results in high-speed transistor operation of 20 MHz, which is twice as high as the common frequency used in near-field wireless communication. The capability of the solution process for scale-up coverage of high-mobility organic semiconductors opens up the way for novel 2D nanomaterials to realize products with large-scale integrated circuits on film-based devices.

INTRODUCTION

Developments in the synthesis and deposition of large-scale two-dimensional (2D) nanomaterials have motivated researchers to realize an ideal building block for future electronic and optoelectronic devices. The key engineering challenge to the integration of these nanomaterials into high-performance devices and applications is the mass production of large-area films without sacrificing the uniform 2D nature (, ). Graphene, group IV elemental sheets (for example, silicene and germanene) (), and transition metal dichalcogenides (for example, MoS2, MoSe2, WS2, and WSe2) () have been emerging as strong contenders in 2D nanomaterials for electronic and optoelectronic applications. Although the two-dimensionality offers unique and excellent electronic properties, the ideal production of uniform, large-area monolayer (1L) sheets with reasonably high efficiency and yield has yet to be realized (). Unlike nanosheets composed of an atomic layer, organic semiconductors (OSCs) form quasi-2D single crystals, where molecules are weakly bonded by van der Waals interaction instead of covalent bonds. This not only offers potential for chemical design but also room temperature solution processing.

Recent intensive studies on the development of OSC compounds have been highlighted by an improvement of the carrier mobility up to 10 cm2 V−1 s−1 (–), which constitutes one step forward toward printable, wearable device applications. Most of these applications, such as radio-frequency identification (RF-ID) tags (, ) and large-scale integrated circuits that drive multiple sensors (–), will require relatively high-speed circuit operation and large-area coverage. Many researchers have attempted to realize layer-controlled, wafer-scale organic single-crystalline semiconductors. For example, organic field-effect transistors (OFETs) using few-layered organic single crystals as an active semiconducting layer have been reported for several materials (–). Few-layer single crystals of pentacene and dioctyl-benzothienobenzothiophene grown epitaxially on atomically flat 2D boron nitride substrates exhibit high mobility (, ). However, these ultrathin crystals, which are normally composed of only a few molecular units (approximately 10 nm), can be fabricated only on a particular substrate with lattice constant matching; thus, their application is considerably limited. Although it has been reported that ultrathin single-crystalline films can also be grown via various solution-coating processes, such as drop casting (), dip coating (), and an air flow–driven assembly technique (), there are some difficulties with respect to the crystal uniformity and/or coverage of the resultant films. Among these techniques, a meniscus-driven solution-processed method can promise scaling-up of uniform organic crystals (–), although state-of-the-art organic single-crystalline films have yet to be demonstrated.

Here, we successfully demonstrate the selective deposition of 1L, bilayer (2L), and trilayer (3L) molecular single crystals with wafer-scale coverage by optimization of the meniscus-driven crystal growth technique. 2L-OFETs show the best performance with a mobility that reaches up to 13 cm2 V−1 s−1, and the channel width–normalized contact resistance extracted by the transmission line method (TLM) is 46.9 ohm·cm, which is, to the best of our knowledge, the lowest value reported for OFETs. In addition, high-frequency operation of the 2L-OFET is confirmed with a channel length of 3 μm. The high mobility and minimized contact resistance of the fabricated 2L-OFET resulted in a cutoff frequency of 20 MHz, and a diode-connected 2L-OFET functioned as a high-speed rectifier that can convert ac signals to dc voltages at frequencies up to 29 MHz.

RESULTS AND DISCUSSION

Fabrication and characterization of ultrathin organic single crystals

Unlike conventional 2D materials, the present organic single-crystalline thin films can be deposited directly onto any given substrate via a simple one-shot solution process. A single crystal of one of our benchmarked materials, 3,11-dioctyldinaphtho[2,3-d:2′,3′-d′]benzo[1,2-b:4,5-b′]dithiophene (C8-DNBDT-NW), which is synthesized and purified in-house (fig. S1), was formed via the meniscus-driven solution method (that is, continuous edge casting; Fig. 1A). The decent solubility of C8-DNBDT-NW contributes to lower the process temperature down to 60° to 70°C (see table S1 and fig. S2 in section S1), which provides ideal fabrication conditions for ultrathin single crystals composed of a few molecular layers. In the present crystal growth technique, solvent evaporation occurs at the edge of the meniscus region of a droplet retained at a shearing blade, which promotes crystallization from a supersaturated zone of the solution. The solute precipitation rate is critically influenced by the amount of solvent evaporation; therefore, the process temperature, shearing rate of the substrate, and the solution concentration (see details in section S2) are adjusted to control the number of layers from 1L to multilayer while achieving large-area coverage. The formation of single-crystalline films was confirmed by cross-polarized optical microscopy observations (Fig. 1, B and C). When the crystal growth direction is parallel or perpendicular to the polarization angle, a completely black image is obtained, which indicates that the crystal axes are highly oriented. The film thickness estimated from atomic force microscopy (AFM) observations was determined to be approximately 7.2 nm, which is equivalent to the height of two molecular layers (Fig. 1D). The single-crystalline domain continues over a few centimeters, which is sufficiently long to cover several hundreds of transistors as confirmed by the scanning electron microscopy (SEM) image shown in Fig. 1E, where brighter and darker regions are 1L and 2L domains, respectively (figs. S13 and S14). The thickness of the single crystal is easily evaluated by the different color contrasts in the polarized optical microscopic images (Fig. 2, A to C). In addition, excellent scalability of the thickness was confirmed by Raman spectroscopy measurements. Figure 2D shows Raman spectra taken for 1L, 2L, and 3L single crystals, where the peak assigned to an intramolecular vibration increases linearly with respect to the number of layers (figs. S15 and S16).

Fig. 1

Characteristics of ultrathin single-crystalline semiconductor films.

(A) Schematic image of the continuous crystal growth method. (B and C) Cross-polarized optical microscopy images of the fabricated ultrathin C8-DNBDT-NW single crystals. (D) AFM image and cross-sectional profile of the 2L single crystal. (E) Combined SEM image of ultrathin 1L and 2L single-crystalline films.

Fig. 2

Comparison of 1L, 2L, and 3L single crystals.

(A to C) Cross-polarized optical microscopy images of 1L-, 2L-, and 3L-OFETs. (D) Raman spectra for 1L, 2L, and 3L single crystals. (E) Two-terminal conductivity as a function of gate voltage VG for 1L-, 2L-, and 3L-OFETs. TEM images and corresponding SAED patterns for (F and G) 1L and (H and I) 2L single crystals. (J) Channel sheet conductivity of the 2L- and 3L-OFETs measured by the gFPP method. Relationship between ΔVcs and the drain current measured by gFPP output measurements for the (K) 2L-OFET and (L) 3L-OFET. ΔVcs is the voltage drop at each source electrode. Dashed lines represent where the drain current is proportional to ΔVcs.

OFET performance of ultrathin single crystals

To evaluate electric properties, OFETs were fabricated with 1L, 2L, and 3L single crystals by subsequent deposition of 2,3,5,6-tetrafluoro-7,7,8,8-tetracyanoquinodimethane (F4-TCNQ) () and gold through a shadow mask. Figure 2E shows the two-terminal conductivity σ2T for the 1L-, 2L-, and 3L-OFETs. The 2L- and 3L-OFETs exhibit textbook-like performance with a near-zero threshold voltage and mobility of more than 10 cm2 V−1 s−1, whereas the mobility of the 1L-OFET was two orders of magnitude lower (figs. S5 to S8). To address this reduction in mobility for the 1L film, the crystal packing structures in 1L and 2L single crystals were investigated using transmission electron microscopy (TEM). Figure 2 (F to I) shows TEM images and selected-area electron diffraction (SAED) patterns of the rectangular unit cells for the 1L and 2L single crystals, which are assigned to a herringbone packing structure in the bulk of the C8-DNBDT-NW crystal (figs. S3 and S4). This indicates that the microscopic crystal structures of the 1L and 2L single crystals are essentially identical. Note that either a broader diffraction spot or a doubly overlapped peak is often found in the SAED patterns of 1L single crystals (figs. S17 to S19), which implies that there is a finite fraction of unavoidable defects and lattice mismatches in 1L single crystals. We consider that these imperfections of the first layer give rise to the significant decrease in mobility. In addition, the lack of disorder patterns in the 2L crystals can be explained as a sufficient amount of organic molecules being supplied during formation of the 2L crystal, which is likely to relax the defects in the first layer.

Contact resistance evaluation

We now turn the discussion to the contact resistance effect. In most OFETs, carrier injection and extraction at the source and drain electrodes occur at a heterojunction between the OSC and metal electrodes. The interfacial contact resistance often dominates the transistor characteristics, so that high-frequency operation is likely to be limited. Both the interfacial and access resistance of the bulk OSC contributes to the net contact resistance in a top-contact, bottom-gate geometry, because charge carriers should travel along the relatively resistive out-of-plane direction of the organic layer from the electrode to the channel. A reduction of the organic layer thickness can significantly reduce the net contact resistance. To address the influence of contact resistance, the gated four-point probe (gFPP) method was used to evaluate the 2L- and 3L-OFETs (fig. S9). Figure 2J shows the four-terminal sheet conductivity σ4T as a function of gate voltage, VG. No apparent difference of σ4T for the 2L and 3L films was observed, and the mobilities of both reached up to 13 cm2 V−1 s−1. This suggests that the present 2L film is sufficiently thick for ideal carrier transport. To further investigate the contact resistance, the gFPP method was adapted to examine the output characteristics. The voltage drop at the source electrodes ΔVcs, shown in the horizontal axis, is extracted from the effective channel potential (see details in section S2). In the 2L-OFET, the drain current, ID, is almost perfectly proportional to ΔVcs over broader ΔVcs ranges, which means that ohmic contact is established (Fig. 2K). On the other hand, a superlinear dependence in ID-ΔVcs is obtained for the 3L-OFET, and the superlinearity is more apparent in the lower ΔVcs region (Fig. 2L), which indicates that the excess layer is likely to behave as a Schottky barrier. To qualitatively analyze the contact resistance, RC, several transistors were fabricated with different channel lengths for 2L and 3L single-crystalline films, and their contact resistance was extracted by the TLM (Fig. 3A and figs. S10 to S12). Figure 3 (B and C) shows the width-normalized total resistance (RC·W) as a function of the gate voltage (VG). The large single-crystalline film has a length of more than a few centimeters, so that the contact resistance can be extrapolated with high accuracy, which is confirmed by a high square of the regression coefficient (R2) of 0.99 over the total VG. A comparison of RC·W for the 2L- and 3L-OFETs with various VG reveals RC·W for the 2L-OFET to be less than that obtained for the 3L device by a factor of 5, as shown in Fig. 3D. This result supports the idea that the contact resistance of top-contact OFETs can be controlled and significantly reduced by a decrease in the film thickness. RC·W for the 2L-OFETs is VG invariant, whereas the contact resistance has VG dependence for typical top-contact OFETs, which is often considered to be caused by the trap states in the access region (–). The observed VG invariant contact resistance can be explained by the presence of an F4-TCNQ layer inserted between the organic layer and the metal electrode. The additional molecular dopant layer is often used to improve charge injection because the dopant layer effectively reduces the trap states. The access region in 2L films is equivalent only to the secondary layer, which is directly attached to acceptor F4-TCNQ molecules; therefore, the interfacial doping becomes more effective compared to 3L single-crystalline films. Thus, it was concluded that F4-TCNQ can fill the trap density of states in the access region of 2L films, which results in VG-independent contact resistance. This is consistent with the ID-ΔVcs characteristics (shown in Fig. 2, K and L). The interface between Au- and F4-TCNQ–doped OSC layers for 2L-OFETs is realized in an ideal ohmic contact, whereas the excess, undoped OSC layer for 3L-OFETs can be the space-charge layer, resulting in the superlinear ID-ΔVcs characteristics. We do not speculate on the mechanism of doping but merely note that understanding charge transport/injection mechanism at the metal/organic interface will be a key issue to further reduce the contact resistance (). RC·W for 2L-OFETs is determined to be 46.9 ohm·cm at a gate voltage of −30 V, which is, to the best of our knowledge, the lowest contact resistance reported for any OFETs except for electrolyte-gated transistors with the contact strongly doped by ionic gel (). In electrolyte-gated transistors, ions penetrate from the dielectric gel into the semiconductor, which results in extremely low contact resistance. However, this electrochemical doping occurs only at low switching frequencies of less than 1 kHz (). For the purpose of transistor application to high-frequency devices, 2L-OFETs, with the adoption of a special doping mechanism, are more favorable in terms of the operation speed.

Fig. 3

Contact resistance evaluation by TLM.

(A) Schematic image of the OFET channel and micrograph of the fabricated 2L-OFETs for TLM measurement. (B) TLM plots for the 2L-OFETs at various gate voltages. Rtotal·W values are extracted from the two-terminal device characteristics. (C) A magnified view at the intercept in (B). (D) Dependence of the contact resistance on the gate voltage for the 2L- and 3L-OFETs.

High-frequency measurement

Both high mobility and very low contact resistance in 2L single-crystalline films allows for an ideal device that will enable high-speed operation. To evaluate the dynamic response, a 2L-OFET with a channel length of a few micrometers was fabricated. Figure 4A shows a schematic image of the short-channel transistor for high-frequency measurements, which includes patterned gate, source, and drain electrodes fabricated via multiple photolithographic processes (). The channel length (L) and the total overlap length between the gate and the source/drain electrodes (LC) were measured to be approximately 3 and 4.5 μm, respectively (Fig. 4B). The transfer and output properties are shown in Fig. 4 (C and D, respectively), where a reasonably high on/off ratio of more than 108 is achieved, even with the relatively short channel length. Thus, no apparent short channel effect is observed. The effective mobility (μeff) extracted from the slope of the transfer curve is estimated to be 2.7 cm2 V−1 s−1 (fig. S20), which is much lower than the intrinsic mobility (μint) obtained from the gFPP measurement. The empirical relation between effective and intrinsic mobility is expressed as ()where Ci is the gate capacitance per unit area of the gate insulator and Vth is the threshold voltage. Note that although 2L single crystals exhibit ultralow contact resistance, the influence of contact resistance is still dominant in a short-channel device with a channel length of a few micrometers, which results in a significant reduction of the effective mobility.

Fig. 4

Short-channel 2L-OFET for cutoff frequency measurement.

(A) Schematic image of the fabricated 2L-OFET. (B) Micrograph of the 2L-OFET with a channel length of 3 μm. (C) Transfer and (D) output properties of the fabricated 2L-OFET with L/W = 3/750 μm. (E) Schematic diagram of the setup for the cutoff frequency measurement. (F) Estimation of the cutoff frequency of the 2L-OFET. Gain is defined as 20 log(ΔID/ΔIG), where ΔIG and ΔID are the amplitudes of the output gate and drain currents, respectively. The cutoff frequency is extracted from the frequency of the gain equal to zero.

The cutoff frequency (fT), which is the maximum operational frequency of an individual transistor, is described by the following equationwhere VD is the applied drain voltage. fT is defined as the frequency at which the current gain is zero, which means the gate current is equal to the drain current. Figure 4E shows a schematic diagram of the setup for the cutoff frequency measurement (). Both the gate and drain dc voltages were set to −10 V.

The cutoff frequency is extracted to be 20 MHz from the frequency when the amplitude of the gate current (ΔIG) is the same as that of the drain current (ΔID), as shown in fig. S21, which is in agreement with the theoretically calculated value of 19 MHz based on Eq. 2. The obtained cutoff frequency is compared with several previous works summarized in table S4. Note that because fT is proportional to the drain voltage, the VD-normalized cutoff frequency is an index for comparison. The cutoff frequency obtained in this study is the highest value for solution-processed devices to our knowledge. In addition, the present 2L-OFET has a slightly longer channel length than that in previous work; therefore, by further reducing the channel length, there should be more room for the improvement of the operational speed.

A high-speed organic rectifier plays a paramount role in achieving low-cost RF-ID tags based on OSCs because the rectifier supplies a dc voltage to the entire circuit on the tags. There are two different approaches to the fabrication of organic rectifiers: a vertical organic diode, where a semiconducting layer is sandwiched between two metal electrodes with different work functions, and, alternatively, a diode-connected OFET. The former approach has a significant geometrical advantage to obtain high-speed response; the effective channel approximates the thickness of the organic layers (, ). However, this configuration is vulnerable to a large voltage produced by electromagnetic induction and to Joule heating (). Therefore, a diode-connected OFET with sufficient insulating capability is more suitable for wireless communication with RF-ID tags. The rectification characteristics of the diode-connected 2L-OFET were measured using the circuit shown in Fig. 5A. The charges passing through the diode are accumulated in the capacitor to produce an output dc voltage (Vout). Figure 5 (B and C) shows the rectifying characteristics at 1 and 25 MHz, where an ac voltage (Vin) with an amplitude of 8 V is applied. The result indicates that the input ac signal can be converted into a constant dc voltage. Figure 5C shows the frequency dependence of the Vout. When the maximum rectifying frequency (frectify) is defined as the frequency at which Vout is decreased by a factor of −3 dB, frectify is extracted to be 29 MHz from Fig. 5D. frectify is expressed as ()where V0 is the amplitude of the input ac voltage and β represents Vout/V0. The theoretical frectify calculated from Eq. 3 is 37 MHz, which is larger than the experimental value. This discrepancy may be due to the voltage dependence of μeff. Equation 3 assumes that μeff is a constant parameter; therefore, the mobility is fixed to 2.1 cm2 V−1 s−1 as the maximum operated value at VG = −4.7 V (see more details in fig. S20). Because μeff is reduced under the influence of its gate dependence, μeff that is averaged over the low-voltage regime is diminished to 1.8 cm2 V−1 s−1 (see more details in section S4 and fig. S20), resulting in a theoretical frectify value of approximately 31 MHz, which is in agreement with experiments. The frectify of 29 MHz is more than twice as high as the commonly used frequency in near-field communication of RF-ID tags (13.56 MHz), although the applied voltage is half of that in our previous report ().

Fig. 5

Rectifying characteristics of a diode-connected 2L-OFET.

(A) Circuit diagram for measurement of the rectifying characteristics. Input ac voltage and output dc voltage of a diode-connected 2L-OFET at frequencies of (B) 1 MHz and (C) 25 MHz. The amplitude of the input signals is 8 V. (D) Frequency dependence of the Vout normalized with respect to Vout obtained at an input frequency of 100 kHz. frectify is extracted from the frequency at which Vout decreases by −3 dB.

In conclusion, we have successfully demonstrated a wafer-scale, layer-controlled organic single-crystalline film composed of only a few molecular layers and achieved a sufficiently high carrier mobility of 13 cm2 V−1 s−1 together with an extremely low contact resistance of 46.9 ohm·cm. Optimization of the meniscus-driven solution crystallization technique developed in-house is expected to scale up the areal coverage of the layer-controlled ultrathin single-crystalline films. The present one-shot deposition method is applicable to any given substrate. These state-of-the-art 2D molecular crystal sheets allow high-frequency operation with a channel length of 3 μm and response at a frequency of 20 MHz with an applied voltage of −10 V. Furthermore, a diode-connected 2L-OFET exhibits a rectifying capability of up to 49 MH2, which is higher than the commonly used frequency in the wireless communication of RF-ID tags. We consider this technique for the fabrication of ultrathin single crystals to contribute to the realization of high-speed organic electronic devices. The two-dimensional molecular crystal sheets demonstrated here could serve as useful scaffolds not only for the integration of functional materials into high-performance devices but also for the realization of groundbreaking functionalities in OSCs.

MATERIALS AND METHODS

Fabrication of devices for contact resistance evaluation

A 100-nm-thick Si substrate with a thermally grown SiO2 layer was cleaned with acetone and 2-propanol and then treated by an ultraviolet-ozone process. A self-assembled monolayer (SAM) of 2-(phenylhexyl)trimethoxysilane was deposited on the surface by vapor deposition at 120°C for 3 hours. Organic crystalline films of C8-DNBDT-NW were then grown from a 0.02–weight % chlorothiophene solution using the continuous edge-casting technique described in our previous work (, ). The substrate was heated up to 60° to 70°C and moved with a shearing rate of 20 μm s−1 while a blade was fixed 100 μm above the substrate. The thickness of the crystalline films was mainly controlled by changing the substrate temperature; 1L, 2L, and 3L single crystals were obtained with substrate temperatures of 58° to 60°C, 60° to 62°C, and 65°C, respectively. After annealing the substrate at 80°C in vacuum to remove residual solvent, F4-TCNQ and Au were subsequently deposited through a metal mask to form source/drain electrodes. Organic layers were patterned by dry-etching processes with an yttrium-aluminum-garnet laser (266 nm).

Fabrication of short-channel devices

Thermally evaporated Ag layers were patterned by conventional photolithography to form gate electrodes. A 100-nm-thick aluminum oxide layer was deposited by an atomic layer deposition technique as a gate dielectric layer. The surface of the aluminum oxide was then treated with a 2-(phenylhexyl)phosphonic acid SAM by immersing the substrate into a 0.2 mM solution for 13 hours. Organic layers were fabricated by almost the same process conditions as those for crystal growth. Source and drain electrodes were patterned by multiple lithographic processes, where OSCoR4001 (Orthogonal Inc.) and AURUM S-50790 (Kanto Chemical Co. Inc.) were used as a photoresist and a gold etchant, respectively.

Electrical measurements

All measurements were performed under ambient conditions. The transistor properties were measured using a semiconductor parameter analyzer (Keithley 4200-SCS). Dynamic measurements to evaluate the cutoff frequency were conducted using the same setup as that reported previously (). ac signals with a peak-to-peak voltage of 1 V [vg(t)] were generated using a function generator (Tektronix AFG3102), and dc gate (VG) and drain (VD) voltages of −10 V were generated using the same semiconductor parameter analyzer. The synthesis voltage [vg(t) + VG] was applied to the gate electrode using a bias tee (ZFBT-4R2GW+, Mini Circuits). The output gate and drain currents were measured using an oscilloscope (Tektronix MDO3014) with current probes (Tektronix CT-6). Measurement of the rectifying characteristics was performed using another function generator (Tabor Electronics WS8102) to apply an input voltage, and an oscilloscope (Teledyne LeCroy HDO4054) was used to read the input and output current signals.