Skip to main content
NCBI home page
ACS AuthorChoice logoLink to ACS AuthorChoice
. 2022 Jun 28;16(7):10994–11003. doi: 10.1021/acsnano.2c03523

Demonstration of Anti-ambipolar Switch and Its Applications for Extremely Low Power Ternary Logic Circuits

Yongsu Lee 1, Sunmean Kim 1, Ho-In Lee 1, Seung-Mo Kim 1, So-Young Kim 1, Kiyung Kim 1, Heejin Kwon 1, Hae-Won Lee 1, Hyeon Jun Hwang 1, Seokhyeong Kang 1,*, Byoung Hun Lee 1,*
PMCID: PMC9331138  PMID: 35763431

Abstract

graphic file with name nn2c03523_0008.jpg

Anti-ambipolar switch (AAS) devices at a narrow bias region are necessary to solve the intrinsic leakage current problem of ternary logic circuits. In this study, an AAS device with a very high peak-to-valley ratio (∼106) and adjustable operating range characteristics was successfully demonstrated using a ZnO and dinaphtho[2,3-b:2′,3′-f]thieno[3,2-b]thiophene heterojunction structure. The entire device integration was completed at a low thermal budget of less than 200 °C, which makes this AAS device compatible with monolithic 3D integration. A 1-trit ternary full adder designed with this AAS device exhibits excellent power–delay product performance (∼122 aJ) with extremely low power (∼0.15 μW, 7 times lower than the reference circuit) and lower device count than those of other ternary device candidates.

Keywords: anti-ambipolar, heterojunction, low thermal budget, ternary logic, pull-middle network, ternary full adder, low-power system

An increase in computing capabilities and integration density of CMOS devices has been achieved by reducing the minimum feature size of transistors.1−5 However, as the power consumption at the interconnect region now accounts for more than 90% of active power consumption for highly scaled chips, a progressive improvement in transistor structure alone cannot be a proper solution for the significantly increasing power consumption.3,4,6−9 Therefore, a more aggressive change in the chip architecture is necessary to deal with the power crisis for future data processing and other computing needs.

Multivalued logic (MVL) technology has been investigated as an alternative technology to overcome the significant rise in power consumption.10−14 MVL is a computing architecture operating with more than two logic states. Therefore, MVL provides significant benefits in terms of the number of transistors and interconnection length required to accomplish identical functions compared to binary logic. Especially, ternary logic, which consists of three logic states (0, 1, and 2 or −1, 0, and +1), is the most extensively studied among the MVL architectures because it theoretically provides the lowest power consumption and best noise margin.15−20

The major bottleneck in ternary logic is the absence of a unit ternary device that can be integrated with the semiconductor process and operated in the same manner as MOSFETs, e.g., room-temperature operation and high drive current. Thus, early studies on ternary architecture were performed using a combination of binary devices. The carbon nanotube field-effect transistor (CNTFET) has been widely used for ternary circuit research owing to its high mobility, tunable threshold voltage (Vth), and very small physical dimensions. However, because of the difficulty in arranging multiple types of CNTs in an integrated circuit, studies on CNTFETs have only been performed theoretically.17−19 The graphene barristor, the Schottky barrier triode between graphene and a semiconductor, was also suggested as a unit transistor for ternary circuits by Heo etal. because of the easy integration process and tunability of Vth to control the Schottky barrier height.21

In addition to the combination of binary devices, ternary circuit implementation using a unit ternary device has been reported. For quantum dot gate FETs,20 an intermediate state was obtained by dynamic Vth shift owing to the tunneling charges accumulated in the quantum dot gate through a thin gate dielectric. However, it was difficult to fabricate two layers of well-aligned quantum dots, and the size limitation of quantum dots and the space between them restricted further scaling.

Recently, ternary devices with a dual-channel structure using graphene barristors have been suggested. The dual-channel ternary device utilizes two parallel channels with different Vth to obtain stepwise transfer characteristics. Kim etal. and Lee etal. demonstrated n-type and p-type dual-channel ternary devices with ZnO and dinaphtho[2,3-b:2′,3′-f]thieno[3,2-b]thiophene (DNTT), respectively, and complementary ternary circuits were demonstrated.22,23 Furthermore, ternary transistors with a stack-channel structure having two stacked ultrathin semiconductor channel layers with different Vth have been demonstrated.24,25 In this device, the intermediate state was realized using an ultrathin phase composite ZnO layer having a voltage-independent constant current flow mechanism, which is attributed to the finite density of states in the conduction band from the mobility edge quantization phenomenon. Low process temperature and easy fabrication process make this device compatible with monolithic 3D integration.

A major challenge in ternary circuit design is finding a proper method to handle the intermediate state. When both devices in the ternary inverter are simultaneously in the intermediate state, a leakage current path between VDD and VGND is formed, leading to intrinsic power dissipation.17−25 This leakage current mitigates the benefits of ternary logic, which is primarily associated with lower device count and less power consumption.

To avoid this problem, a ternary circuit scheme, called single pole triple throw (SPTT), has been proposed.26 It uses an additional power rail for VDD/2. Although it becomes much more straightforward to obtain three states with this approach, the number of devices to form three states inevitably increases. Usually, two binary switches represent states 0 and 2, and two additional devices are necessary to represent state 1. The addition of power rails was technologically challenging in the 1970s; however, the recent progress in buried power rail technology can alleviate the area burden of additional power rails and contacts.27,28 Thus, the SPTT approach became a promising technology for ternary architecture.

A single anti-ambipolar switch (AAS) device can replace the two devices required to emulate the intermediate state of a standard ternary inverter (STI) designed with an SPTT scheme. AAS devices pass information at a sharply limited gate bias range corresponding to the intermediate state, VDD/2. Thus, an appropriately designed AAS device can be used as a unit device in a pull-middle (PM) network, which is connected with a VDD/2 power rail. Peak-to-valley ratio (PVR), defined as the ratio of on (peak) current and off (valley) current, is one of the important characteristics of AAS devices. To implement the proposed ternary circuit, a high PVR is desirable because a high on current makes state transition fast, and a low off current reduces the power consumption due to the leakage current between networks, similar to the circuits using unipolar devices.

Several prior studies on AAS behavior used bias-dependent modulation of transconductance (gm) curves in Si, Ge, and III–V semiconductor devices.29−35 Unfortunately, the low PVR in these devices limited their practical applications. Recently, various heterojunction devices using transition metal dichalcogenides (TMDCs), such as MoS2, WSe2, SnSeS, and ReS2, were proposed, and high current density and moderate PVR have been reported.36−40 However, the difficulty of large-area integration of TMDC heterojunction devices and their compatibility with CMOS circuits are still a hurdle for practical applications. Thus, manufacturing an AAS device compatible with large-scale CMOS integration and having a reasonable performance at room temperature is one of the most important challenges in ternary logic technology.

Here, we report a ZnO–DNTT AAS device, with a high PVR of above 106, adjustable operation bias range, solid stability and reliability, and CMOS process compatibility, especially with back-on-the-line or 3D integrated circuits. The performance of the ternary full adder designed with the ZnO–DNTT AAS device exhibited ∼7 times lower power–delay product (PDP) than the ideal CNTFET-based ternary circuits in testing.

Results and Discussion

Figure 1(a) shows the schematic device structure of the ZnO–DNTT AAS device. Bulk p++ Si substrate and 90 nm SiO2 are used as back gate and gate dielectric, respectively. On top of the gate dielectric, ZnO and DNTT channels are formed in sequence. Both semiconductors were selected because of the low process temperature required for monolithic 3D integration and air-stable characteristics required for device integration. ZnO was stably deposited using a low-temperature ALD process at 120 °C.41,42 DNTT, a p-type organic thin-film semiconductor, was also deposited at room temperature using thermal evaporation. Both ZnO and DNTT showed reasonable air stability. The heterojunctions of the ZnO and DNTT layers partially overlap with each other, and Au electrodes contact each side of the heterojunction channels. The fabrication process is detailed in the Methods section (and Supporting Information, Figure S1). Figure 1(b) shows the fabricated device, with device dimensions of W = 180 μm and L = 420 μm for the overlapped regions and L = 300 μm for the underlapped regions. Figure 1(c) shows the cross-sectional TEM image of the ZnO–DNTT heterojunction. The XPS analysis of the ZnO channel indicates that ZnO is a bundle of nanorods having a wurtzite structure. For DNTT, a benzene ring structure with sulfur atoms conjugated with carbon atoms is verified by XPS analysis (Supporting Information, Figure S2).

Figure 1.

Figure 1

Open in a new tab

(a) Schematic device structure of the ZnO–DNTT AAS device. (b) Photograph of the fabricated device. Scale bar = 200 μm. (c) Cross-sectional TEM image of the ZnO–DNTT heterojunction region. Scale bar = 20 nm. (d) ID–VG curves of p-type DNTT TFT and n-type ZnO TFT. The thicknesses of each channel are 50 and 29 nm at VD = ∓ 5 V, respectively. (e) ID–VG curves of ZnO–DNTT AAS devices at VD = 5 V. (f) Negative gm property of the AAS device at VD = 5 V. (g) ID–VG curves of ZnO–DNTT AAS devices with VD ranging from −5 and 5–30 V. Symbols indicate experimental data, and lines indicate device modeling fitting results.

Figure 1(d) shows the ID–VG curves of thin-film transistors (TFTs) from individually fabricated DNTT and ZnO TFTs with typical electrical characteristics of depletion mode p-type and n-type semiconductor TFTs, respectively. The p-type DNTT TFT exhibited a subthreshold swing of 0.33 V/dec, a field-effect mobility of 2.3 cm2/V·s, and an on/off ratio of 106. For the ZnO n-type TFT, a subthreshold swing of 1.12 V/dec, a field-effect mobility of 48 cm2/V·s, and an on/off ratio of 107 were obtained. Swing values appear to be high in our devices because the gate dielectric thickness is 90 nm. The on/off ratio and field-effect mobility of our devices are comparable to or better than those of individual n-type ZnO TFT and p-type DNTT TFTs previously reported.41−44 Furthermore, owing to the balance of these electrical characteristics between n- and p-type TFTs, the AAS devices fabricated with the combination of ZnO TFT and DNTT TFT showed symmetrical transfer properties.

The transfer characteristic of ZnO–DNTT AAS devices measured at VD = 5 V is shown in Figure 1(e). A PVR of ∼105 was obtained within a narrow turn-on region from −3 to 8 V. The AAS device showed a negative gm, which reached −20 nS at VDNTT = 5 V, as shown in Figure 1(f). Figure 1(g) shows the drain bias dependence of the AAS device. As VD increased, the peak voltage, peak current, and operation window of the AAS device increased, and a PVR of ∼106 was obtained at VD = 30 V. This is the highest PVR of an AAS device yet reported (102–105).29−40,45,46 On the other hand, when VDNTT is negative (−5 V), the current does not flow because of the reverse bias barrier of the PN junction.

Figure 2 shows the operation mechanism of the AAS device. The electrical behavior of the AAS device can be easily understood if we interpret this device as three devices connected in series that are controlled by the same gate bias: n-type ZnO TFT–ZnO/DNTT PN junction–p-type DNTT TFT. When VG was lower than Vth of the ZnO TFT (Vth,n) (Figure 2(a)) or higher than Vth of the DNTT TFT (Vth,p) (Figure 2(c)), the final conductance of this series connected device was very low; that is, when one of two semiconductors is turned off, the carriers cannot flow across the depleted channel. Forward diode current can flow through the ZnO/DNTT heterojunction region only when the ZnO TFT and DNTT TFT regions are partially turned on. Thus, this device can only be turned on between Vth,n and Vth,p (ΔVON) (Figure 2(b)). As a result, the current flow increases rapidly at a certain bias and dissipates at the other bias, as shown in Figure 2(d). IPEAK position is defined as the highest current value at VPEAK, and the peak-to-valley ratio is defined as IPEAK over the lowest current outside of ΔVON.

Figure 2.

Figure 2

Open in a new tab

Schematic diagram of the cross-sectional AAS device structures and band diagrams with carrier flows when (a) VG < Vth,n, (b) Vth,n < VG < Vth,p, and (c) Vth,p < VG. (d) Typical ID–VG curve of the AAS device.

Physically, the drive current of this device is similar to the leakage current of an inverter consisting of a p-type DNTT TFT and an n-type ZnO TFT, but the current flux can be modulated by individually controlling the properties of an n-type TFT, a p-type TFT, and ZnO/DNTT heterojunctions. Figure 3(a) shows the outcome of device modulation with different ZnO channel thicknesses. As the ZnO channel thickness increases, the electron charge concentration tends to increase, so the Vth of ZnO TFTs decreases.25,47 Accordingly, the turn-on range (ΔVON) of the AAS device can be modulated. By controlling the thickness of ZnO, IPEAK was modulated in proportion to ΔVON. In the DNTT channel, the hole charge concentration can also be varied by changing the thickness of the channels.48Figure 3(b) shows the ID–VG curves of AAS devices with different DNTT channel thicknesses. Because the Vth variability of individual ZnO TFTs was much larger than that of DNTT TFTs (Supporting Information, Figure S4(a–b)), the range of ΔVON modulation was higher for devices with different ZnO thicknesses than for those with different DNTT thicknesses. Meanwhile, even though there was a trade-off between IPEAK and ΔVON, very sharp and narrow regions of ΔVON could be achieved with extremely high PVRs above 103 in both cases.

Figure 3.

Figure 3

Open in a new tab

ID–VG curves of the AAS device with different (a) thicknesses of ZnO (50 nm of DNTT), (b) thickness of DNTT (29 nm of ZnO), and (c) types of doping layer at VD = 5 V (29 and 50 nm of ZnO and DNTT, respectively).

While the thickness change in the n-type or p-type region affected the Vth and current level, the peak current position was not changed significantly. In other words, shifting the range of the turn-on region of the device was limited. To find a way to shift the device turn-on region, an additional polymer-based doping layer was inserted between the ZnO and DNTT layers. Figure 3(c) shows the result of the doping layer insertion. We used polyethylenimine (PEI) (n-doping) and poly(acrylic acid) (PAA) (p-doping) as doping layers, which induced electrons and holes at both surfaces of ZnO and DNTT, respectively. With electron doping (PEI), the Vth of single-channel TFT devices shifts toward the negative bias side. Meanwhile, the Vth shifts toward the positive bias side with hole doping (PAA) (Supporting Information, Figure S4(c–d)). Figure 3(c) clearly shows that the range of the turn-on region of the device can be shifted by inserting the doping layer. In summary, we have identified various parameters that can be used to modulate the peak current level, peak position, and width of the turn-on region. With further optimization, we expect to realize a more sharply defined turn-on region at the specific design position.

The fabrication process appears to be relatively stable and shows promising scale-up feasibility. Figure 4(a) shows consistent ID–VG curves for 72 separate AAS devices in a single chip. To verify the stability and manufacturability of ZnO–DNTT AAS devices, the statistical data were extracted. Figure 4(b) shows the average PVR values measured over 180 days, having a consistent distribution within a 2 × 2 cm chip (inset of Figure 4(b) shows the optical photograph of this chip). Figure 4(c) shows the distributions of PVR values, VPEAK, and ΔVON. The coefficient of variation, CV = s/xm × 100%, where s is the standard deviation and xm is the average value, can be applied to quantify the distribution of data.49 The CVs of the distributions for PVR, VPEAK, and ΔVON are 9.7%, 4.4%, and 3.4%, respectively. CVs of each distribution are within 10% for all devices measured from the same chip, which means the device distributions are reasonably tight even with the manufacturing process at university facilities (clean class 10000). Table 1 compares various AAS devices reported in the literature. Most of the devices fabricated with TMDCs used the exfoliation method, which is not suitable for practical large-area manufacturing processes. Some heterojunction devices were fabricated using organic semiconductors deposited using the evaporation method, but the operation voltages were very high.45,46 Our ZnO–DNTT AAS device exhibited air-stable and uniform device properties with a reasonably tight distribution and can be fabricated with a low process temperature, which is very useful for monolithic or heterogeneous 3D integration.

Figure 4.

Figure 4

Open in a new tab

(a) ID–VG curves for 72 separate AAS devices at VD = 30 V, in which thicknesses of ZnO and DNTT are 29 and 50 nm, respectively. (b) PVR of the AAS device over time (inset: 2 cm × 2 cm wafer with 72 AAS devices) and (c) the distributions of the PVR, VPEAK, and ΔVON from 72 AAS devices.

Table 1. Comparison of AAS Devices.

device JD [nA/μm] PVR fab. method fab. T [°C] EOT [nm] |VD| [V]
ZnO–DNTT (this work) 0.9–5.6 105–106 ALD and evaporation 120 90 5–30
Si (p+–i–n+)29 0.004–2 2–10 CMOS   25 0.05–0.7
Si (n+–p+–n+)30 3.6 2 CMOS   2 0.001
Si (p+–n+–p+)31 20–109 1–5.5 CMOS 800 3 0.01–0.05
Si–Ge nanowire32 0.22–270 20–48 CVD 495 2 0.2–0.8
GaSb–InAsSb nanowire33 2200 1.6 epitaxy 500 10 0.5
MoS2–WSe236 30 103 exfoliation 360 >300 1
SnSeS–BP37 133 102 exfoliation   4.68 1
MoS2–BP38 67 103 exfoliation 250 300 1
ReS2–BP39 4000 105 exfoliation 250 3.9 1
MoS2–pentacene40 450 103 exfoliation 100 300 10
PTCDI-C8-6T45 0.2 5 × 104 evaporation 120 >200 60
PTCDI-C8–DNTT46 0.05–0.6 103–105 evaporation 175 >350 30–60
Open in a new tab

For more practical applications, circuit-level implementation strategies of ternary circuits are examined using the ZnO–DNTT AAS device model. The semiphysical analytical device model is developed using a series connection of n- and p-type TFT models and a PN junction model. The simulation result fits very well with the experimental data shown in Figure 1(g).

Conceptually, the ternary circuit using the AAS device can be simplified as shown in Figure 5(a). Three networks for pull-up (PU), pull-down (PD), and pull-middle (PM) are necessary to represent ternary states, and they are connected to VDD, VGND, and VDD/2 voltage sources, respectively. Compared with previous ternary circuit schemes (Supporting Information, Figure S5), an additional voltage source (VDD/2) and a PM network were added to provide an intermediate state. Recently, an inverter showing ternary characteristics using an organic-based AAS device has been reported.46 In this scheme, the leakage current problem at the intermediate state still exists because both PU and PD are half turned on. In our scheme, devices in PU and PD networks turn off at the intermediate state, which is similar to that in a CMOS inverter circuit. Within the narrow operation window where both PU and PD networks are turned off, the AAS device is turned on to provide the intermediate state while avoiding the leakage current from VDD to VGND. In summary, three networks are used to represent three states in our scheme, and only one network is turned on for each state.

Figure 5.

Figure 5

Open in a new tab

(a) Static gate design methodology for the proposed ternary circuit, (b) their truth table and switching table strategy, (c) device symbol and device structure of an AAS ternary unit cell, and (d) current transfer characteristics of an ideal AAS ternary unit cell model for EOT = 1 nm and W = L = 100 nm.

The static ternary gate design methodology using this approach is shown schematically in Figure 5(b). The truth table is divided into three tables, which correspond to PU, PD, and PM networks. Note that the AAS device works only under a forward drain bias (VD = +5 V). When the drain bias is negative, there is no current flow, as shown in Figure 1(g). In this case, the transition between the PM and PU (or PD) network can be restricted if a reverse bias is applied to the AAS device during the transition. This problem can be overcome by adding an AAS device in the opposite direction as shown in Figure 5(c). Figure 5(d) shows the theoretically modeled I–V curves of an AAS unit cell, which shows bidirectional switching characteristics with VDD less than 2 V. For this modeling, the equivalent oxide thickness (EOT) of the gate dielectric is assumed to be 1 nm, and the channel length and width are assumed to be 100 nm to emulate the scaled device operation.

Using the AAS unit cell shown in Figure 5(c), an STI is designed to demonstrate the circuit level benefits of the ZnO–DNTT AAS device. Figure 6(a) shows the transition from a PU to PM network. With low VIN, the p-type device in the PU network is turned on and the AAS device in the PM network is turned off. As VIN increases, the p-type device is starting to turn off while the AAS device is starting to turn on. When the transition between the PU and PM network occurs, VOUT is changed from 2 V (VDD) to 1 V (VDD/2). The leakage current flows from the VDD node to the VDD/2 node during this state transition only, similar to that in a CMOS inverter. Likewise, Figure 6(b) shows the transition between the PM and PD network when VOUT is changed from 1 V (VDD/2) to 0 V (VGND). The voltage transfer characteristics (VTCs) and the leakage current are shown in Figure 6(c). In both cases, the leakage current flows only during the state transition, in which both networks (PU and PM or PM and PD) are turned on at the same time. As a result, the static power problem at the intermediate state of ternary logic can be minimized. In other STIs, leakage currents at the intermediate state are on the order of μA to nA,19,25 but the leakage current of the STI with an AAS device is at the pA level. Overall, the static power of an STI with an AAS device can be reduced to a sub-nW level, which is ∼1/100 of the typical value for a CNTFET-based ternary STI (Supporting Information, Figure S6).

Figure 6.

Figure 6

Open in a new tab

Proposed STI operation and its VTCs as VOUT changes from (a) 2 to 1 V and (b) 1 to 0 V and (c) whole VTC (symbol) with drive current (line).

Using the above design strategy, various ternary module circuits such as STI, negative ternary inverter (NTI), positive ternary inverter (PTI) (Supporting Information, Figure S7), NMIN, NMAX (Supporting Information, Figure S8), SUM, CONS, and ANY (Supporting Information, Figure S9) were designed using the AAS devices. Integrating these module circuits, a ternary full adder was simulated. Figure 7(a) shows a ternary half adder designed by combining one SUM and one CONS gate. A ternary full adder is designed by combining two half adders and one ANY gate.19,21 The truth tables of SUM, CONS, and ANY gates are shown in Figure 7(b). The transient responses of the ternary full adder obtained at 5 MHz frequency are shown in Figure 7(c). Because these devices are not fully optimized, some glitches remain; however, the overall logic functionality has been successfully confirmed, which indicates that ternary logic circuits of any size can be designed using our approach.

Figure 7.

Figure 7

Open in a new tab

(a) Gate level of the ternary full adder, (b) truth tables of SUM, CONS, and ANY gates, and (c) transient responses of the proposed ternary full adder.

The performance of our ternary full adder implemented with the ZnO–DNTT AAS device was evaluated and compared with other circuits having similar complexity using HSpice (Table 2). In all the circuits, four inverters were connected to the fan-out node, and all scaled physical dimensions were set to 100 nm (for binary CMOS, it was 90 nm). The CNTFET model of Stanford University50 and the predictive technology model for a 90 nm CMOS51 are used to evaluate other full adder circuits. For the binary ternary adder, 2-bit circuits are used for the comparison because the size of data of 1 trit is closest to that of 2 bits.21,52

Table 2. Comparison of Full Adder Circuits.

logic architecture VDD [V] # Tr power [μW] delay [ps] PDP [aJ]
1-trit ternary this work (L = W = 100 nm) 2 88 0.15 799 122
CNTFET (L = 100 nm) 1 110 1.07 160 172
2-bits binary 90 nm CMOS 2 56 7.15 120 858
1 56 0.096 200 19
Open in a new tab

Compared to the CNTFET-based 1-trit full adder, the proposed full adder has great advantages in terms of the number of transistors, power, and power–delay product, even at larger VDD. In particular, the power consumption is ∼7 times lower than that of the CNTFET-based full adder. Furthermore, the power consumption and PDP of the 1-trit full adder are superior to the 2-bit binary full adder; especially, the power consumption is ∼47 times lower than that of the binary full adder. The power consumption of the ternary circuit is even comparable to that of the optimal CMOS circuit at VDD = 1 V. Although the PDP of the ternary circuit is comparable to that of binary CMOS due to significant power saving, the delay needs to be improved further, by potentially optimizing parasitic components such as contact resistances or capacitances. Overall, we have confirmed that the proposed ternary circuits utilizing the AAS device can have technical advantages for extremely low power applications if the appropriate scaling of the AAS device can be achieved.

Conclusion

A ZnO–DNTT AAS device and ternary unit cell design to overcome the well-known leakage current problem of ternary logic in the intermediate state has been successfully demonstrated with extremely low power operation capability. Using this device concept, the advantages of the ternary full adder over other ternary devices or even binary CMOS devices, especially in terms of power consumption, have been confirmed. Even though the operation voltage of the AAS device should be scaled down further for more scaled devices, this approach will expand the capabilities of ternary logic architecture, especially for low-power ternary logic systems monolithically integrated with binary logic systems.

Methods

Fabrication Process and Electrical Characterization of ZnO–DNTT AAS Device

The 90 nm SiO2/p++ Si substrates (used as a back gate) were cleaned with acetone, isopropyl alcohol, and distilled water (DI) for 5 min in sequence by sonication. Then, ZnO layers were deposited using the atomic layer deposition (ALD) process at 120 °C using a diethyl zinc (DEZ) precursor and H2O oxidant. The thicknesses of ZnO varied from 19 to 50 nm by controlling the cycle of the ALD process. Then, the ZnO layers were patterned using photolithography and a wet etch process with 1% HCl solution diluted by DI. To perform chemical doping, the surface of active ZnO was submerged into a 0.2 wt % ethanol-diluted branched PEI (Sigma-Aldrich) or PAA (Sigma-Aldrich) solution for 3 h. Subsequently, the devices were cleaned with pure ethanol for a few seconds to prevent the excessive doping effect. For the p-type semiconductor, a DNTT (Sigma-Aldrich) layer was deposited to form a heterojunction structure using thermal evaporation at 25 °C. Patterning was performed using a shadow mask. The thicknesses of DNTT varied from 5 to 50 nm. Then, 70 nm thermally evaporated Au electrodes were deposited using the shadow mask process. While fabricating the heterojunction devices, ZnO and DNTT single-channel devices were also fabricated as references. Electrical characterization was performed using a semiconductor parameter analyzer (Keithley 4200) at 25 °C in ambient air.

Device and Logic Circuit Modeling

Device modeling and circuit synthesis were performed using HSpice (Synopsys). A semiphysical analytical TFT model was used to model the device characteristics of the ZnO–DNTT AAS device.53 Transient response, power dissipation, and delay of ternary circuits were extracted using the developed ZnO–DNTT AAS device model and compared with the modeling results of the ternary circuits designed with the CNTFET model50 and the predictive technology model for 90 nm CMOS.51

Acknowledgments

This work was partially supported by Creative Materials Discovery Program on Creative Multilevel Research Center (2017M3D1A1040834) and FEOL platform development program (2020M3F3A2A02082436) through the National Research Foundation of Korea (NRF), funded by the Ministry of Science and ICT, Korea.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsnano.2c03523.

  • Fabrication process flow of the ZnO–DNTT AAS device, XPS analysis of ZnO and DNTT, transfer curves and transconductance of ZnO–DNTT AAS device dependence on voltage bias, transfer curves of ZnO and DNTT single TFT dependence on channel thickness and doping type, various STI schemes, comparison of STI operation, various ternary circuit designs and their operation (Figures S1–S9) and PM ternary device switching table (Table S1) (PDF)

The authors declare no competing financial interest.

Supplementary Material

nn2c03523_si_001.pdf (881.4KB, pdf)

References

  1. Moore G. E. Cramming More Components onto Integrated Circuits. Proc. IEEE 1998, 86 (1), 4. 10.1109/JPROC.1998.658762. [DOI] [Google Scholar]
  2. Hiramoto T. Five Nanometre CMOS Technology. Nat. Electron 2019, 2 (12), 557–558. 10.1038/s41928-019-0343-x. [DOI] [Google Scholar]
  3. Khan H. N.; Hounshell D. A.; Fuchs E. R. H. Science and Research Policy at the End of Moore’s Law. Nat. Electron 2018, 1 (1), 14–21. 10.1038/s41928-017-0005-9. [DOI] [Google Scholar]
  4. Salahuddin S.; Ni K.; Datta S. The Era of Hyper-Scaling in Electronics. Nat. Electron 2018, 1 (8), 442–450. 10.1038/s41928-018-0117-x. [DOI] [Google Scholar]
  5. Akinwande D.; Huyghebaert C.; Wang C.-H.; Serna M. I.; Goossens S.; Li L.-J.; Wong H.-S. P.; Koppens F. H. L. Graphene and Two-Dimensional Materials for Silicon Technology. Nature 2019, 573 (7775), 507–518. 10.1038/s41586-019-1573-9. [DOI] [PubMed] [Google Scholar]
  6. Son Y.; Frost B.; Zhao Y.; Peterson R. L. Monolithic Integration of High-Voltage Thin-Film Electronics on Low-Voltage Integrated Circuits Using a Solution Process. Nat. Electron 2019, 2 (11), 540–548. 10.1038/s41928-019-0316-0. [DOI] [Google Scholar]
  7. Choi J.; Han J. S.; Hong K.; Kim S. Y.; Jang H. W. Organic–Inorganic Hybrid Halide Perovskites for Memories, Transistors, and Artificial Synapses. Adv. Mater. 2018, 30 (42), 1704002. 10.1002/adma.201704002. [DOI] [PubMed] [Google Scholar]
  8. Andrae A. S. G.; Edler T. On Global Electricity Usage of Communication Technology: Trends to 2030. Challenges 2015, 6 (1), 117–157. 10.3390/challe6010117. [DOI] [Google Scholar]
  9. Shibata T.; Ohmi T. A Functional MOS Transistor Featuring Gate-Level Weighted Sum and Threshold Operations. IEEE Trans. Electron Devices 1992, 39 (6), 1444–1455. 10.1109/16.137325. [DOI] [Google Scholar]
  10. Rine D. C.Computer Science and Multiple-Valued Logic: Theory and Applications; Elsevier: Amsterdam, 1977. [Google Scholar]
  11. Gaudet V. A Survey and Tutorial on Contemporary Aspects of Multiple-Valued Logic and Its Application to Microelectronic Circuits. IEEE Journal on Emerging and Selected Topics in Circuits and Systems 2016, 6 (1), 5–12. 10.1109/JETCAS.2016.2528041. [DOI] [Google Scholar]
  12. Smith The Prospects for Multivalued Logic: A Technology and Applications View. IEEE Transactions on Computers 1981, C–30 (9), 619–634. 10.1109/TC.1981.1675860. [DOI] [Google Scholar]
  13. Chen W.-H.; Dou C.; Li K.-X.; Lin W.-Y.; Li P.-Y.; Huang J.-H.; Wang J.-H.; Wei W.-C.; Xue C.-X.; Chiu Y.-C.; King Y.-C.; Lin C.-J.; Liu R.-S.; Hsieh C.-C.; Tang K.-T.; Yang J. J.; Ho M.-S.; Chang M.-F. CMOS-Integrated Memristive Non-Volatile Computing-in-Memory for AI Edge Processors. Nat. Electron 2019, 2 (9), 420–428. 10.1038/s41928-019-0288-0. [DOI] [Google Scholar]
  14. Yang R.; Li H.; Smithe K. K. H.; Kim T. R.; Okabe K.; Pop E.; Fan J. A.; Wong H.-S. P. Ternary Content-Addressable Memory with MoS2 Transistors for Massively Parallel Data Search. Nat. Electron 2019, 2 (3), 108–114. 10.1038/s41928-019-0220-7. [DOI] [Google Scholar]
  15. Kim K.; Kim S.; Lee Y.; Kim D.; Kim S.-Y.; Kang S.; Lee B. H.. Extreme Low Power Technology Using Ternary Arithmetic Logic Circuits via Drastic Interconnect Length Reduction. In 2020 IEEE 50th International Symposium on Multiple-Valued Logic (ISMVL); 2020; pp 155–158, 10.1109/ISMVL49045.2020.00-13. [DOI]
  16. Kim Y. J.; Kim S.-Y.; Noh J.; Shim C. H.; Jung U.; Lee S. K.; Chang K. E.; Cho C.; Lee B. H. Demonstration of Complementary Ternary Graphene Field-Effect Transistors. Sci. Rep 2016, 6 (1), 39353. 10.1038/srep39353. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Raychowdhury A.; Roy K. Carbon-Nanotube-Based Voltage-Mode Multiple-Valued Logic Design. IEEE Transactions on Nanotechnology 2005, 4 (2), 168–179. 10.1109/TNANO.2004.842068. [DOI] [Google Scholar]
  18. Lin S.; Kim Y.-B.; Lombardi F. CNTFET-Based Design of Ternary Logic Gates and Arithmetic Circuits. IEEE Transactions on Nanotechnology 2011, 10 (2), 217–225. 10.1109/TNANO.2009.2036845. [DOI] [Google Scholar]
  19. Kim S.; Lim T.; Kang S.. An Optimal Gate Design for the Synthesis of Ternary Logic Circuits. In 2018 23rd Asia and South Pacific Design Automation Conference (ASP-DAC); 2018; pp 476–481, 10.1109/ASPDAC.2018.8297369. [DOI]
  20. Karmakar S.; Chandy J. A.; Jain F. C. Design of Ternary Logic Combinational Circuits Based on Quantum Dot Gate FETs. IEEE Transactions on Very Large Scale Integration (VLSI) Systems 2013, 21 (5), 793–806. 10.1109/TVLSI.2012.2198248. [DOI] [Google Scholar]
  21. Heo S.; Kim S.; Kim K.; Lee H.; Kim S.-Y.; Kim Y. J.; Kim S. M.; Lee H.-I.; Lee S.; Kim K. R.; Kang S.; Lee B. H. Ternary Full Adder Using Multi-Threshold Voltage Graphene Barristors. IEEE Electron Device Lett. 2018, 39 (12), 1948–1951. 10.1109/LED.2018.2874055. [DOI] [Google Scholar]
  22. Kim S.-Y.; Heo S.; Kim K.; Son M.; Kim S.-M.; Lee H.-I.; Lee Y.; Hwang H. J.; Ham M.-H.; Lee B. H. Demonstration of Ternary Devices and Circuits Using Dual Channel Graphene Barristors. In 2019 IEEE 49th International Symposium on Multiple-Valued Logic (ISMVL); IEEE: Fredericton, NB, Canada, 2019; pp 25–30, 10.1109/ISMVL.2019.00013. [DOI]
  23. Lee Y.; Kim S.-M.; KIM K.; Kim S.-Y.; Lee H.-I.; Kwon H.; Lee H.-W.; Kim C.; Some S.; Hwang H. J.; Lee B. H.. Dual-Channel P-Type Ternary DNTT–Graphene Barristor. Submitted to Sci. Rep. 2022. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Lee L.; Hwang J.; Jung J. W.; Kim J.; Lee H.-I.; Heo S.; Yoon M.; Choi S.; Van Long N.; Park J.; Jeong J. W.; Kim J.; Kim K. R.; Kim D. H.; Im S.; Lee B. H.; Cho K.; Sung M. M. ZnO Composite Nanolayer with Mobility Edge Quantization for Multi-Value Logic Transistors. Nat. Commun. 2019, 10 (1), 1998. 10.1038/s41467-019-09998-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Kim S.; Kim K.; Kim A. R.; Lee H.; Lee Y.; Kim S.; Yu S. H.; Lee H.; Hwang H. J.; Sung M. M.; Lee B. H. Operation Principles of ZnO/Al2O3-AlDMP/ZnO Stacked-Channel Ternary Thin-Film Transistor. Adv. Electron. Mater. 2021, 7 (6), 2100247. 10.1002/aelm.202100247. [DOI] [Google Scholar]
  26. Shim C.-H.; Heo S.; Noh J.; Kim Y. J.; Kim S.-Y.; Khan A. K.; Lee B. H.. Design of Ratioless Ternary Inverter Using Graphene Barristor. In 2016 IEEE 46th International Symposium on Multiple-Valued Logic (ISMVL); 2016; pp 23–30, 10.1109/ISMVL.2016.51. [DOI]
  27. Gupta A.; Kundu S.; Teugels L.; Bommels J.; Adelmann C.; Heylen N.; Jamieson G.; Pedreira O. V.; Ciofi I.; Chava B.; Wilson C. J.; Tokei Z.. High-Aspect-Ratio Ruthenium Lines for Buried Power Rail. In 2018 IEEE International Interconnect Technology Conference (IITC); 2018; pp 4–6, 10.1109/IITC.2018.8430415. [DOI]
  28. Gupta A.; Pedreira O. V.; Arutchelvan G.; Zahedmanesh H.; Devriendt K.; Mertens H.; Tao Z.; Ritzenthaler R.; Wang S.; Radisic D.; Kenis K.; Teugels L.; Sebai F.; Lorant C.; Jourdan N.; Chan B. T.; Subramanian S.; Schleicher F.; Hopf T.; Peter A. P.; et al. Buried Power Rail Integration With FinFETs for Ultimate CMOS Scaling. IEEE Trans. Electron Devices 2020, 67 (12), 5349–5354. 10.1109/TED.2020.3033510. [DOI] [Google Scholar]
  29. Kim C.; Lee Y.; Lee S. Systematic Modulation of Negative-Differential Transconductance Effects for Gated P+-i-N+ Silicon Ultra-Thin Body Transistor. J. Appl. Phys. 2017, 121 (12), 124504. 10.1063/1.4979213. [DOI] [Google Scholar]
  30. Naquin C.; Lee M.; Edwards H.; Mathur G.; Chatterjee T.; Maggio K. Gate Length and Temperature Dependence of Negative Differential Transconductance in Silicon Quantum Well Metal-Oxide-Semiconductor Field-Effect Transistors. J. Appl. Phys. 2015, 118 (12), 124505. 10.1063/1.4931662. [DOI] [Google Scholar]
  31. Kim K. R.; Kim H. H.; Song K.-W.; Huh J. I.; Lee J. D.; Park B.-G. Field-Induced Interband Tunneling Effect Transistor (FITET) with Negative-Differential Transconductance and Negative-Differential Conductance. IEEE Transactions on Nanotechnology 2005, 4 (3), 317–321. 10.1109/TNANO.2005.847008. [DOI] [Google Scholar]
  32. Zhang P.; Le S. T.; Hou X.; Zaslavsky A.; Perea D. E.; Dayeh S. A.; Picraux S. T. Strong Room-Temperature Negative Transconductance in an Axial Si/Ge Hetero-Nanowire Tunneling Field-Effect Transistor. Appl. Phys. Lett. 2014, 105 (6), 062106. 10.1063/1.4892950. [DOI] [Google Scholar]
  33. Dey A. W.; Svensson J.; Ek M.; Lind E.; Thelander C.; Wernersson L.-E. Combining Axial and Radial Nanowire Heterostructures: Radial Esaki Diodes and Tunnel Field-Effect Transistors. Nano Lett. 2013, 13 (12), 5919–5924. 10.1021/nl4029494. [DOI] [PubMed] [Google Scholar]
  34. Ganjipour B.; Dey A. W.; Borg B. M.; Ek M.; Pistol M.-E.; Dick K. A.; Wernersson L.-E.; Thelander C. High Current Density Esaki Tunnel Diodes Based on GaSb-InAsSb Heterostructure Nanowires. Nano Lett. 2011, 11 (10), 4222–4226. 10.1021/nl202180b. [DOI] [PubMed] [Google Scholar]
  35. Carnevale S. D.; Marginean C.; Phillips P. J.; Kent T. F.; Sarwar A. T. M. G.; Mills M. J.; Myers R. C. Coaxial Nanowire Resonant Tunneling Diodes from Non-Polar AlN/GaN on Silicon. Appl. Phys. Lett. 2012, 100 (14), 142115. 10.1063/1.3701586. [DOI] [Google Scholar]
  36. Nourbakhsh A.; Zubair A.; Dresselhaus M. S.; Palacios T. Transport Properties of a MoS2/WSe2 Heterojunction Transistor and Its Potential for Application. Nano Lett. 2016, 16 (2), 1359–1366. 10.1021/acs.nanolett.5b04791. [DOI] [PubMed] [Google Scholar]
  37. Huang M.; Li S.; Zhang Z.; Xiong X.; Li X.; Wu Y. Multifunctional High-Performance van Der Waals Heterostructures. Nat. Nanotechnol. 2017, 12 (12), 1148–1154. 10.1038/nnano.2017.208. [DOI] [PubMed] [Google Scholar]
  38. Lv W.; Fu X.; Luo X.; Lv W.; Cai J.; Zhang B.; Wei Z.; Liu Z.; Zeng Z. Multistate Logic Inverter Based on Black Phosphorus/SnSeS Heterostructure. Adv. Electron. Mater. 2019, 5 (1), 1800416. 10.1002/aelm.201800416. [DOI] [Google Scholar]
  39. Xiong X.; Kang J.; Hu Q.; Gu C.; Gao T.; Li X.; Wu Y. Reconfigurable Logic-in-Memory and Multilingual Artificial Synapses Based on 2D Heterostructures. Adv. Funct. Mater. 2020, 30 (11), 1909645. 10.1002/adfm.201909645. [DOI] [Google Scholar]
  40. Dong J.; Liu F.; Wang F.; Wang J.; Li M.; Wen Y.; Wang L.; Wang G.; He J.; Jiang C. Configuration-Dependent Anti-Ambipolar van Der Waals p–n Heterostructures Based on Pentacene Single Crystal and MoS2. Nanoscale 2017, 9 (22), 7519–7525. 10.1039/C7NR01822C. [DOI] [PubMed] [Google Scholar]
  41. Li S.; Chen X.; Liu L.; Zeng Z.; Chang S.; Wang H.; Wu H.; Long S.; Liu C. Micron Channel Length ZnO Thin Film Transistors Using Bilayer Electrodes. J. Colloid Interface Sci. 2022, 622, 769–779. 10.1016/j.jcis.2022.04.016. [DOI] [PubMed] [Google Scholar]
  42. Li H.; Han D.; Dong J.; Yi Z.; Zhou X.; Zhang S.; Zhang X.; Wang Y. Enhanced Performance of Atomic Layer Deposited Thin-Film Transistors With High-Quality ZnO/Al2O3 Interface. IEEE Trans. Electron Devices 2020, 67 (2), 518–523. 10.1109/TED.2019.2957048. [DOI] [Google Scholar]
  43. Acharya R.; Günder D.; Breuer T.; Schmitz G.; Klauk H.; Witte G. Stability of Organic Thin-Film Transistors Based on Ultrathin Films of Dinaphtho[2,3-b:2′,3′-f]Thieno[3,2- b]Thiophene (DNTT). Journal of Materials Chemistry C 2021, 9 (1), 270–280. 10.1039/D0TC04554C. [DOI] [Google Scholar]
  44. Ibrahim G. H.; Zschieschang U.; Klauk H.; Reindl L. High-Frequency Rectifiers Based on Organic Thin-Film Transistors on Flexible Substrates. IEEE Trans. Electron Devices 2020, 67 (6), 2365–2371. 10.1109/TED.2020.2989730. [DOI] [Google Scholar]
  45. Kobashi K.; Hayakawa R.; Chikyow T.; Wakayama Y. Negative Differential Resistance Transistor with Organic P-n Heterojunction. Adv. Electron. Mater. 2017, 3 (8), 1700106. 10.1002/aelm.201700106. [DOI] [Google Scholar]
  46. On S.; Kim Y.-J.; Lee H.-K.; Yoo H. Ambipolar and Anti-Ambipolar Thin-Film Transistors from Edge-on Small-Molecule Heterostructures. Appl. Surf. Sci. 2021, 542, 148616. 10.1016/j.apsusc.2020.148616. [DOI] [Google Scholar]
  47. Nakata M.; Tsuji H.; Sato H.; Nakajima Y.; Fujisaki Y.; Takei T.; Yamamoto T.; Fujikake H.. Influence of Oxide Semiconductor Thickness on TFT Characteristics. 2012 19th International Workshop on Active-Matrix Flatpanel Displays and Devices (AM-FPD); 2012.
  48. Chang J.-F.; Shie H.-S.; Yang Y.-W.; Wang C.-H. Study on Correlation between Structural and Electronic Properties of Fluorinated Oligothiophenes Transistors by Controlling Film Thickness. Crystals 2019, 9 (3), 144. 10.3390/cryst9030144. [DOI] [Google Scholar]
  49. Lovie P.Coefficient of Variation. Encyclopedia of Statistics in Behavioral Science; John Wiley & Sons, Ltd.: Chichester, 2005; Vol. 1, pp 317–318, 10.1002/0470013192.bsa107. [DOI] [Google Scholar]
  50. Stanford University CNTFET Model , https://nano.stanford.edu/downloads/stanford-cnfet-model (accessed Feb 14, 2022).
  51. Predictive Technology Model (PTM) , http://ptm.asu.edu (accessed Feb 14, 2022).
  52. Etiemble D.Comparing Ternary and Binary Adders and Multipliers. 2019, arXiv:1908.07299, https://arxiv.org/abs/1908.07299 (accessed Feb 8, 2022).
  53. Lee S.; Striakhilev D.; Jeon S.; Nathan A. Unified Analytic Model for Current–Voltage Behavior in Amorphous Oxide Semiconductor TFTs. IEEE Electron Device Lett. 2014, 35 (1), 84–86. 10.1109/LED.2013.2290532. [DOI] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

nn2c03523_si_001.pdf (881.4KB, pdf)

Articles from ACS Nano are provided here courtesy of American Chemical Society

RESOURCES