Figure 2a presents an idealized device structure of the doping-free homojunction with a gapped channel near the double-acting electrode, which served as both source-drain and gate. Monolayer MoS2 grown by CVD is chosen as the channel material to fabricate devices. The thickness of MoS2 is ~ 0.8 nm measured by atomic force microscopy (AFM), indicating the MoS2 flake is a single layer (Supplementary Fig. 1c). The Raman spectra and PL spectrum of MoS2 are shown in Supplementary Fig. 1d-e, and the conclusions are consistent with the results of the AFM. To confirm the unique device structure of this homojunction, a representative sample after fabrication is characterized by high-resolution TEM, showing the profile of the core region. An energy-dispersive spectrometer (EDS) is also used to confirm the material composition of the core region (Fig. 2a and Supplementary Fig. 2). In our demonstration, the Cr/Au (10/50 nm) as the source contact is made on the MoS2, and hexagonal boron nitride (h-BN) flakes are used as dielectric layers. The extended electrode serves as both the drain and gate contact to complete the device fabrication (Fig. 2b). Noteworthy, an eye-catching gap is left under the monolayer MoS2 channel near the extended drain, creating a homojunction that can benefit from the self-biased effect. The signature structure feature of this homojunction is that the drain not only provides the biased voltage but also works as the gate to realize a barrier-tunable current path. The detailed fabrication process can be seen in the Methods and Supplementary Fig. 3.
To demonstrate the self-biased effect of this homojunction, we perform in-situ KPFM experiments to map the potential distribution of the dashed square (Fig. 2b). The measurement configuration enables us to spatially map the potential distribution of the core region under different bias conditions. Figure 2c and Supplementary Fig. 4 show the KPFM mapping (upper part) and the corresponding averaged contact potential difference (VCPD) profile (lower part) of the selected area at the bias voltages Vbias = -1 V, 0 V, and 1 V, respectively. A sharp potential mutation occurs at the gapped channel position after a bias voltage is applied, and the potential difference becomes more obvious as the bias voltage increases (Supplementary Fig. 5). It should be noted here that electrons, as the majority carriers in the channel, dominate the conductance of the device. Under positive bias voltage, a positive electrostatic field from the extended drain will implement electron doping to make the MoS2 channel highly conductive, thus promoting the flow of electrons from the source to the drain. Under negative bias voltage, electrons are depleted in the channel due to a negative electrostatic doping from the drain, blocking the flow of electrons. As a result, the extended drain will simultaneously apply an electrostatic doping caused by this potential difference to tune the channel carrier concentration and form a bias-tunable homojunction at the gapped channel position.
Subsequently, we perform monolayer MoS2 in-situ PL spectrum characterization experiments under different bias conditions to further analyze the self-biased effect (Fig. 2d and Supplementary Fig. 6). Here, we need to know that the PL intensity of the A exciton is highly dependent on electrostatic doping, while the AT trion is independent on electrostatic doping due to a largely trion binding energy28, and the A exciton will convert into the AT trion as the electron doping increases. Compared with Vbias=0 V, the peak intensity of A exciton is significantly lower than that of AT trion at Vbias= 1 V, indicating that there is an electron doping to further promote the transition from A exciton to AT trion. The peak intensity of A exciton is higher than that of AT trion at Vbias= -1 V, indicating that there is a hole doping to deplete the excess electrons and thus inhibit the transition from A exciton to AT trion. Since the PL intensity of monolayer MoS2 is dominated by the prominent A exciton, a PL intensity diminishment and a peak energy red-shift are observed at Vbias= 1 V, while a significant PL intensity enhancement and a peak energy blue-shift are observed at Vbias= -1 V (Supplementary Fig. 6b).
The result is that this homojunction exhibits an obvious asymmetrical electrical transport behavior when the applied bias voltage is changed from positive to negative, as shown in Fig. 2e. The I-V curve can be roughly divided into three regions: Ⅰ) reverse current saturation region; Ⅱ) forward ideal linear diode region; Ⅲ) forward series resistance-dominant region. An ideality factor (n) for evaluating the rectification performance of the homojunction can be calculated by fitting the Shockley diode Eq. 29,30.
$${I}_{ds}={I}_{s}\left[exp\left(\frac{{V}_{ds}}{n{V}_{T}}\right)-1\right]$$
where the VT and Is are defined as the thermal voltage and reverse saturation current, respectively. A near-unity ideality factor of ~ 1.1 (n = 1 is ideal) can be obtained due to the rapid increase of current in region Ⅱ under forward bias (0.02–0.25 V), indicating excellent rectification performance. Then, we explore the electrostatic self-biased effect of the homojunction under special bias mode and a higher bias voltage (Supplementary Fig. 7). The results show that the device is on when the extended drain has a positive potential relative to the source, which is defined as forward bias. Conversely, the device is off when the extended drain has a negative potential relative to the source, which is defined as reverse bias. The excellent rectification characteristics of the homojunction have almost no degradation after 30 days in the atmospheric environment. And an ultrahigh rectification ratio of ~ 109 with an ultralow reverse saturation current below 1 pA is achieved with a bias range of ± 10 V. As a comparison, this value is far higher than previous reports based on 2D materials that have been developed29–33. This unique asymmetrical electrical behavior of the homojunction is attributed to the partially gapped MoS2 channel and the electrostatic self-biased effect caused by the double-use extended electrode.
Furthermore, we chose HfO2 prepared by ALD instead of h-BN to demonstrate that the homojunction (SBT) is also suitable for other dielectric materials. The influence of HfO2 thickness and channel length on the electrical transport properties of the SBT is systematically investigated, as shown in Supplementary Fig. 8. The size of the gap is mainly determined by the thickness of HfO2 and the rectification characteristics of SBT have no degradation with the downscaling of the gap (THfO2 downscales to ≈ 15 nm). Even an improved rectification ratio is achieved, which may be attributed to the enhanced electrostatic doping as the thickness of HfO2 decreases. Note that the off-current and on-current of SBT increase synchronously with the downscaling of the channel (Lch downscales to < 300 nm). And an improved rectification characteristic is the fact that on-current increases more, which is attributed to lower gap effect on the channel resistance in the on-state. More importantly, the SBT exhibits very high stability, reliability, and endurance, which can be attributed to maintaining the integrity of the channel materials with an ingenious gap design, and the high stability of the grown MoS2 itself. The SBT exhibits excellent rectification characteristics with negligible degradation at a square pulse bias signal for 10 Hz, even after 1,000 switching (on/off) cycles. Moreover, a retention time of about 10 years for steady operation is projected, and further details can be found in Supplementary Fig. 9. The robust rectification characteristics of HfO2-based SBT pave the way for fabricating low-power and high-performance logic circuits, as discussed in detail in the following section.