Main

Transistors can be divided into three groups according to Ng and Sze1,2,3: field-effect transistors, potential-effect transistors and hot-carrier transistors. The first two groups are represented by the metal–oxide–semiconductor field-effect transistor (MOSFET) and the bipolar junction transistor (BJT), respectively, which have achieved great success in modern integrated circuits, whereas the third group has advantages in speed and multifunction based on the excess kinetic energy of hot carriers4,5, mainly including the hot-electron transistor (HET) and the real-space-transfer transistor (RSTT). A HET uses a metal or a semiconductor as the base, and when electrons are injected into the base from the emitter, they become hot and fast because their energy is higher than those in the base, producing a short base transit time and a high-speed device6,7. Two-dimensional materials such as graphene and molybdenum disulfide (MoS2) have been used as the base to further reduce the base transit time because of their low atomic thickness, providing a potential terahertz operation, which is promising in the next 6G technologies8,9,10,11. In contrast, an RSTT uses an electrical field to accelerate carriers, which, when they become hot enough, will transfer from one route to another, resulting in high-speed operation and a negative differential resistance (NDR)12,13, which is highly needed in various fields such as high-frequency oscillators14,15.

HETs and RSTTs provide potential high performance. However, the mechanisms of the hot-carrier generation are either carrier injection or acceleration, which may limit the device performance and function7,8,9,13. Neither of these devices can provide an ultralow subthreshold swing less than 60 mV dec−1 beyond the Boltzmann limit, which is highly needed for modern low-power applications16,17. In addition, for an RSTT, the NDR is limited when the device is fabricated using silicon (Si) and germanium (Ge) technology, which is compatible with mainstream semiconductor production. A novel mechanism of hot-carrier generation is needed to improve the power consumption and NDR of hot-carrier devices.

Mixed-dimensional electronic devices fabricated by combining bulk and low-dimensional materials can utilize the advantages of different dimensional materials in terms of geometric scale, and electrical and optical performance18,19,20,21, and may combine these advantages to provide a novel mechanism of hot-carrier generation. For low-dimensional materials such as graphene and carbon nanotubes, carrier mobility is high, which can be used to heat carriers using an electrical field. Meanwhile, various potential barriers can be formed using different energy-band combinations of bulk and low-dimensional materials, which can be used to emit high-energy carriers. Here we report a mixed-dimensional hot-emitter transistor (HOET) based on double graphene/germanium Schottky junctions. Using stimulated emission of heated carriers, the transistor achieves a subthreshold swing lower than 1 mV dec−1 and an NDR with a peak-to-valley current ratio greater than 100 at room temperature. Multi-valued logic applications with a high inverter gain and reconfigurable logic states are further demonstrated based on these characteristics.

Device structure and characteristics

The transistor is essentially composed of a monolayer graphene (Gr) with a gap cut in it and a p-type Ge substrate. Gr contacts Ge through the hafnium dioxide (HfO2) window. The two separated Gr layers were used as the emitter (emitter-Gr) and the base (base-Gr), and the Ge substrate was used as the collector (Fig. 1a,b). Devices were fabricated using Gr transfer and standard semiconductor processing. The high-quality monolayer Gr was grown by chemical vapour deposition and transferred onto the Ge substrate22 (Extended Data Fig. 1). The gap in the Gr was fabricated using photolithography with a gap length from 2 μm to 75 μm (Fig. 1c and Methods).

Fig. 1: Device structure and basic characteristics.
figure 1

a, Illustration of the transistor structure, where monolayer Gr with a gap cut in it is on a p-Ge substrate (resistivity 1–10 Ω cm). The two separate Gr layers were used as the emitter (e) and base (b) contacting the Ge substrate through the HfO2 window, and the Ge substrate as the collector (c), with the electrodes being Ti/Au. The channel length and width of the transistor are defined as those of the HfO2 window. b, A diagram of the cross-section of the transistor showing monolayer Gr with a gap on a Ge substrate. c, Transistors with gap lengths of 2 μm to 75 μm were fabricated using photolithography and 9 of them are shown in the optical image. Scale bar, 100 μm. d, Transfer characteristics Ic–Vb show a negative abruptly changing collector current Ic with an SS less than 1 mV dec−1 beyond the Boltzmann limit at room temperature. e, Output characteristics Ic–Vc show an NDR with a PVR of around 100.

Full size image

The Gr is separated by the gap as tested by current–voltage (I–V) measurements (Extended Data Fig. 2). The I–V characteristics of the emitter-Gr/p-Ge junction and the base-Gr/p-Ge junction show an on-to-off current ratio of about 103 at ±3 V, indicating the existence of a Schottky barrier between Gr and Ge (Extended Data Fig. 3a). The temperature-dependent I–V characteristics of the junction indicate a thermionic-emission-dominant mechanism where the Schottky barrier height was determined to be about 0.38 eV (Extended Data Fig. 3b and Methods). See Extended Data Fig. 3 for more junction characteristics. For the transistor, the relationship of the collector current Ic and the base voltage Vb in the transfer characteristics (Ic–Vb) shows an abrupt current change beyond the Boltzmann limit where the subthreshold swing (SS) is below 1 mV dec−1 (Fig. 1d), whereas the one of Ic and the collector voltage Vc in the output characteristics (Ic–Vc) shows an NDR with a peak-to-valley current ratio (PVR) around 100 (Fig. 1e).

Ultralow subthreshold swing

The SS is a basic parameter to characterize the switching performance of a transistor. A smaller SS is preferred for low-power operation; however, it is usually larger than 60 mV dec−1 because of the Boltzmann limit23. When the HOET works, the emitter bias Ve is grounded giving the transistor a common-emitter configuration. When the base bias Vb increases, at a critical base bias Vb-critical, a negative collector current Ic is observed where the current change is rather abrupt (Figs. 1d and 2a). At room temperature, the abrupt current change is beyond the Boltzmann limit where the minimum SS is in the range of 0.38–1.52 mV dec−1 as Vc increases, and the range of the current with an SS less than 60 mV dec−1 is about 1 to 3 orders of magnitude, which could be further increased (Fig. 2b and Methods). For a current with an SS less than 60 mV dec−1, the average SS is from 0.82 mV dec−1 to 6.1 mV dec−1 and the maximum on-current is from 73.9 μA μm−1 to 165.2 μA μm−1, which is one of the best reported results17,24,25,26,27,28,29,30,31 (Fig. 2c).

Fig. 2: An ultralow SS and the SEHC mechanism.
figure 2

a, Transfer characteristics Ic–Vb of a transistor with a 2-μm gap showing an abruptly changing collector current Ic with a minimum SS of 0.38 mV dec−1. b, SS–Ic relationship showing minimum values of SS from 0.38 mV dec−1 to 1.52 mV dec−1 and the range of the current with an SS less than 60 mV dec−1 of about 1 to 3 orders of magnitude. c, Benchmark of the average SS and maximum on-current of the transistor for a current with an SS less than 60 mV dec−1, which is one of the best reported results when compared with impact ionization MOSFETs (I-MOS), depletion-IMOS (DIMOS), heterojunction tunnel field-effect transistors (HJ-TFET), black phosphorus tunnel field-effect transistors (BP TFET), heterojunction tunnel triodes (HJ-TTs), carbon nanotube tunnel field-effect transistors (CNT TFET) and negative-capacitance field-effect transistors (NC-FET)17,24,25,26,27,28,29,30,31. d, Transfer characteristics with temperature-dependent current behaviour. e, The critical base bias Vb-critical when Ic starts to increase abruptly increases linearly with Vc. f, For different biases of Vc, devices show increasing Vb-critical with increasing gap length dgap from 5 μm to 75 μm. g, As Ic increases abruptly, Ie also increases abruptly. h,i, An illustration of a transistor with a gap of a length of dgap in the Gr channel (h) and its energy-band diagram near the Gr channel (i).

Full size image

Stimulated emission of heated carrier mechanism

It is noted that holes are the main conduction carriers in the HOET because Gr is p-type (Extended Data Fig. 4), and the abrupt negative Ic indicates a sudden increase in the hole current flowing out of the collector, which is neither a normal reverse leakage current of the Gr/Ge junctions nor a forward current of the base-Gr/p-Ge junction. Four phenomena shed light on the device operation mechanism. First, the transfer characteristics are temperature dependent (Fig. 2d). Different from the tunnelling behaviour, the current changes more abruptly when the temperature increases. The ultralow SS appears when the temperature is above room temperature, which is the working temperature of most realistic systems. Second, the critical base bias Vb-critical when Ic abruptly changes increases linearly with Vc, and Vc − Vb-critical is about 0.7 V, leading to a forward-biased base-Gr/p-Ge junction (Fig. 2e). Third, at each bias of Vc, Vb-critical increases with increasing gap length dgap (5 μm to 75 μm in 5-μm steps; Fig. 2f and Extended Data Fig. 5). Finally, Ic and Ie increase abruptly at the same time (Fig. 2g). These phenomena can be summarized as that initially both the emitter-Gr/p-Ge junction and the base-Gr/p-Ge junction are reverse biased, and when the base bias increases to a critical value, the base-Gr/p-Ge junction is sufficiently forward biased, so that an exceptional number of holes in the emitter-Gr will suddenly be emitted into the Ge collector, while holes will enter from the emitter to ensure a continuous current from the emitter to the collector. The higher the temperature, the more obvious the phenomenon, and the shorter the gap, the smaller the critical base bias.

We propose a stimulated emission of heated carrier (SEHC) mechanism to explain these phenomena using a structure illustration (Fig. 2h) and an energy-band diagram (Fig. 2i) of the device. There are four processes that collectively lead to the ultralow SS. In process A (carrier heating), Vb and the electric field in the emitter-Gr accelerate holes there to become heated holes; however, they are not hot enough to overcome the emitter-Gr/Ge potential barrier. In process B (carrier injection), holes are injected from Ge into the base-Gr to become high-energy holes with the forward bias there. In process C (carrier diffusion), the injected high-energy holes in the base-Gr will overcome the potential barriers induced by the base-Gr/Ge/emitter-Gr structure by diffusion to arrive at the emitter-Gr. In process D (carrier emission), with a higher energy, these arrived holes will pass their energy to the heated ones in the emitter-Gr through carrier–carrier scattering (CCS)32, making them the stimulated carriers that will continue to participate in the CCS process, causing a stimulated-carrier multiplication. These stimulated carriers with high energy will overcome the emitter-Gr/Ge barrier with the reverse bias there, leading to an abrupt hole current (Methods). The Ge collector current will first increase to a peak abruptly and then decrease when the current of the base-Gr/Ge junction begins to dominate. Compact modelling is used to further explain the multiplication and emission process (Methods and Extended Data Fig. 6). The SEHC mechanism indicates that even a transistor with a continuous Gr channel can still generate an ultralow SS, which is validated by our experiments (Extended Data Fig. 7).

Negative differential resistance

The NDR effect refers to the characteristic where the current of a device decreases as the voltage increases, which has been widely used in modern electronics such as amplifiers, microwave generators, high-frequency oscillators and high-speed digital-to-analogue converters, generally evaluated by the ratio of maximum and minimum currents, that is, the PVR33,34,35,36,37,38,39,40,41. However, when using Si and Ge technologies, the NDR effect generated by a hot-carrier device is limited where the PVR is no more than three42,43,44,45,46. In the HOET, the output characteristics Ic–Vc show an obvious NDR (Figs. 1e and 3a). When the collector bias Vc increases, Ic first increases to a peak value, and then decreases to the reverse currents of the Gr/Ge junctions. The output characteristics are temperature dependent where the NDR gradually vanishes when the temperature decreases (Fig. 3b), and at each bias of Vb, the voltage Vc-peak where Ic achieves its maximum, decreases as the gap length dgap increases (Fig. 3c and Extended Data Fig. 8).

Fig. 3: A negative differential resistance.
figure 3

a, Output characteristics Ic–Vc of a transistor with a 3-μm gap showing the PVR. b, Temperature-dependent output characteristics. c, For different biases of Vb, devices with an increasing gap length dgap from 5 μm to 75 μm show a decreasing Vc-peak. d, Peak and valley currents as well as the corresponding PVR from 90.6 to 24.6 in the output characteristics. e, Benchmark of the PVR for our device compared with devices using Gr and RSTTs using Si and Ge technologies33,34,35,36,37,38,39,40,41,42,43,44,45,46. GO, graphene oxide; hBN, hexagonal boron nitride.

Full size image

These phenomena are consistent with the SEHC mechanism: in the output characteristics, for each negative bias of Vb, as the negative bias of Vc increases, the hot holes at the emitter-Gr are collected by the collector, leading to a large negative Ic, gradually reaching the peak current. When Vc further increases, the bias of the base-Gr/Ge junction changes from a forward bias to a reverse one, and the carrier-injection process stops, leading to the valley current. It should be noted here that although an RSTT can also generate an NDR, the characteristics are different. In an RSTT, the NDR shows in the Ib–Vb curve where Ib decreases as Vb increases because the carriers are accelerated by Vb to become hot and are transferred to the collector. In a HOET, Ib never decreases, no matter how large Vb is (before the device is damaged), which is also true even for a transistor with a continuous Gr channel, indicating that carriers cannot be accelerated to become hot enough only by applying Vb.

The peak and valley currents increase with the base bias Vb (Fig. 3d), and the PVR is from 90.6 to 24.6 (Fig. 3d). When Vb is −3 V, the high PVR is due to a small leakage current of the Gr/Ge junction, and the best PVR is 126 (Extended Data Fig. 9). This result is one of the highest values for a device using Gr33,34,35,36,37,38,39,40,41, and is higher than any RSTT using Si and Ge technologies42,43,44,45,46 (Fig. 3e), and is also comparable to the best result of a tunnel device using two-dimensional materials14.

Multi-valued logic technology

The multifunctional HOET is promising in various applications. For example, multi-valued logic (MVL) uses more than two logic states to enable rapid and low-power data processing with high-density integration15, and the HOET can be used to provide a high inverter gain and reconfigurable logic states for MVL, which has rarely been reported before. A circuit was fabricated using three HOETs (T1–T3) in parallel with a common emitter, a common collector (Ge substrate) and separated bases 1–3, illustrated by an equivalent circuit and device symbols (Fig. 4a,b). The input and output signals are voltage and current, respectively43 where the lowest voltage potential corresponds to the input logic ‘0’ and the smallest current (absolute value) corresponds to the output logic ‘0’.

Fig. 4: HOETs for MVL technology.
figure 4

a, In the circuit, three HOETs (T1, T2 and T3) are connected in parallel with a common Ti/Au emitter, a common Ge collector and separated bases. b, An equivalent circuit using device symbols showing a common emitter and collector and separated bases 1–3. c, The input (IN) and output (OUT) signals are voltage and current, respectively. As Vb3 changes, three abrupt changes of Ic occur, showing that the circuit is a quaternary digital logic inverter. d, The inverter gain gm (dIc/dVb3) can approach 1 mA μm−1 V−1 when the logic state changes. e,f, The dependence of Ie on Vb3 is also the behaviour of a quaternary inverter (e) with a high inverter gain gm (dIe/dVb3) (f). g–i, When different combinations of base biases are applied, the input logic signal Vc can be linked to different output logic signal Ic, leading to a ternary digital logic inverter (g), a ternary follower (h) and a component that can be used to construct an adder (i).

Full size image

To demonstrate a high inverter gain, one base voltage is used as the input signal (IN, taking Vb3 as an example), and the collector current Ic is the output signal (OUT; Fig. 4c). Other voltage biases can be applied to control the shape of the Ic curve, such as the position where an abrupt change happens. As Vb3 changed from 0 V to −12 V, the injected carriers from base 3 first arrived at the emitter of device T3 leading to an abrupt change of Ic, followed by another two abrupt changes when the injected carriers arrived at the emitters of devices T2 and T1. The relationship between Ic and Vb3 demonstrates that the circuit can be seen as a quaternary (0, 1, 2, 3) digital logic inverter that has three characteristics. First, the inverter gain gm (transconductance dIc/dVb3) is high when the logic state changes because of the abruptly changing Ic, approaching 1 mA μm−1 V−1, which can be used to fabricate a low-power MVL (Fig. 4d). Second, more HOETs can be connected in parallel to achieve quinary or even higher system using a simple structure. Third, the dependence of Ie on Vb3 is also the behaviour of a quaternary inverter, which provides more flexibility for circuit design (Fig. 4e,f).

To demonstrate reconfigurable logic states, the output characteristics Ic–Vc are investigated. Vc is used as the input signal (IN), and the collector current Ic is the output signal (OUT). Ic is the sum of the collector currents of the three HOETs. For each collector current, the position where it achieves the maximum value is controlled by its base bias. Therefore, when different combinations of base biases are applied, the logic states of the output signal are different, that is, they are reconfigurable, leading to different functions. When the input logic signal is (2, 1, 0), if the output logic signal is (0, 1, 2), the circuit is a ternary digital logic inverter (Fig. 4g). If the output logic signal is (2, 1, 0), it is a ternary follower (Fig. 4h). If the output logic signal is (0, 2, 1), it can be used to construct an adder (Fig. 4i). More possibilities can be realized by using different base biases, and more HOETs can be connected in parallel to achieve higher system.

Conclusion

Using the SEHC mechanism based on mixed-dimensional materials, the HOET provides another member of the hot-carrier transistor family, generating an ultralow SS that is one of the lowest reported values and a PVR in the NDR effect that is one of the highest for Gr devices. By combining the correct materials and device structure, the HOET can provide a multifunctional and high-performance device with potential applications in low-power and NDR technologies for the post-Moore era.

Methods

Preparation of the substrate

A p-type (100) Ge substrate with a resistivity of 1–10 Ω cm was cleaned by hydrofluoric acid (40 wt%) for 60 s to remove native oxide on the surface. A 30-nm-thick HfO2 insulating layer was deposited on top of the Ge substrate by atomic layer deposition at 200 °C (precursors, tetrakis (dimethylamido) hafnium (Hf (NMe2)4) and water). The bottom of the Ge substrate was scratched, and Ti/Au (5/50 nm) metallization was performed by electron beam evaporation to form an ohmic contact. Electrode metallization using Ti/Au (5/50 nm) on the surface was formed by photolithography and electron beam evaporation. For photolithography, photoresist s-1813 (spin-coated at 3,000 rpm for 30 s, baked at 120 °C for 2 min) and LOR3A (spin-coated at 3,000 rpm for 50 s, baked at 190 °C for 5 min) were used in sequence. The HfO2 layer on the substrate was then patterned by photolithography and reactive ion etching (CF4 50 standard cubic centimetres per minute (sccm), 5.0 Pa, radiofrequency power 100 W, 5.5 min), followed by dilute hydrofluoric acid (5 wt%) etching for 30 s to form a window to the Ge.

Preparation of the monolayer Gr film

Monolayer Gr film was synthesized by chemical vapour deposition on a commercial copper foil (99.9%, 25-μm thick). The copper foil was first annealed at 1,000 °C under a 5-sccm hydrogen flow and then exposed to a mixture of hydrogen (5 sccm) and methane (60 sccm) at a total pressure of 100 Pa for 30 min to grow Gr, followed by slow cooling to room temperature.

Gr film transfer and device fabrication

Gr was transferred by using the typical wet polymethyl methacrylate (PMMA) method. The solution of PMMA (950 kDa molecular weight, Sigma, 4 wt% in ethyl lactate) was first spin-coated on the graphene/Cu foil at 2,000 rpm for 60 s and cured at 180 °C for 15 min. After removing the Cu foil by chemical etching, the PMMA/Gr film was carefully collected on the desired target substrate and baked. PMMA was then removed by immersing in acetone at 50 °C to complete the transfer. Finally, the transferred Gr was patterned by photolithography and oxygen plasma etching (200 W, 180 sccm, 2 min).

The emission process in the SEHC mechanism

When the collector current increases abruptly, (A) the electric field in the emitter-Gr heats the holes there while (B) high-energy holes are injected from Ge into the base-Gr, which will arrive at the emitter-Gr by (C) diffusion. Only one injected hole is shown (Extended Data Fig. 6a). The arrived high-energy holes will pass their energy to the heated ones through (D1) CCS32 (Extended Data Fig. 6b). When the stimulated holes in the emitter-Gr gain enough energy from the injected holes, (D2) a multiplication process happens (Extended Data Fig. 6c), before the stimulated holes with enough energy are (D3) emitted into Ge with the reverse bias there, leading to the abrupt current (Extended Data Fig. 6d).

In the CCS process, after one collision between carriers, the number of carriers that can cross the Gr/Ge barrier will double, and the energy these carriers possess after the collision is still higher than the Gr/Ge barrier. In addition, the lateral electric field generated by Vb can increase these carriers’ energy. Therefore, the carriers after the collision can continue to participate in the CCS, causing the number of carriers that can cross the barrier to double again. Meanwhile, In the energy domain, the CCS will result in a high-energy-band tail in the energy distribution function, indicating an increase in high-energy carrier distribution32. Therefore, the continuous CCS process can cause the number of high-energy carriers in the emitter-Gr to increase repeatedly, leading to a surge in the reverse current. It should be noted that the impact ionization47 cannot be responsible for this phenomenon, as that Ic never increases abruptly no matter how large Vb and the corresponding electric field is applied (before the device is damaged) in the emitter-Gr if the base-Gr/Ge junction is not forward biased.

Modelling of the CCS multiplication process

The base bias Vb determines the current Ic not only by providing a lateral electric field in the emitter-Gr but also by controlling the injected carriers at the base-Gr/Ge junction, leading to a complex relationship among the scattering, the multiplication and Vb, and therefore a complex one between Ic and Vb. In fact, multiplication processes are usually modelled using an experience-based approach47 and we provide an empirical model for the multiplication process in the HOET below.

The critical base bias Vb-critical, where Ic increases abruptly, increases linearly with the collector bias Vc (Fig. 2e) and the gap length dgap (Fig. 2f). On the basis of these experimental results, Ic is described as Ic = M × Irev + I0, M = A/(1 − (Vb − Vc)/(Vb-critical − Vc)), Vb-critical = Bdgap + Vc + C, where Irev is the reverse current of the emitter-Gr/Ge junction before the CCS multiplication, M is the CCS multiplication factor and I0, A, B and C are fitting constants. As shown in Extended Data Fig. 6e, the model fits the experimental results well.

Modelling of the Gr/Ge junction

On the basis of the thermionic-emission current model of a Schottky junction, the Gr/p-Ge Schottky potential barrier height qϕB, ideality factor η, interface state density Dit and series resistance Rs are estimated as follows. The relationship of the forward current IF of the Gr/Ge junction and temperature T is ln(IF/T2) = C − q(ϕB – Vc/η)/k × (1/T), where C is a constant, q is the elementary charge, Vc is the forward voltage bias and k is Boltzmann’s constant3. Using an Arrhenius plot, the slopes of the fitted lines –q(ϕB – Vc/η)/1,000k is plotted against Vc (Extended Data Fig. 3c), and the y intercept at 0 V is S0 = −qϕB/1,000k leading to a qϕB of about 0.38 eV, while the slope is Slope* = q/1,000kη leading to an η of about 1.29. Dit is estimated using η and ϕB based on a relationship of η = 1 + (δ/ε0)(εs/Wd + q2Dit), where δ is the thickness of an interfacial layer between Gr and Ge, ε0 is the permittivity in vacuum (8.85 × 10−14 F cm−1), εs is the relative dielectric constant (16.2) of Ge and Wd is the thickness of the depletion layer of Ge (ref. 3). Wd = (2εs/qNa × (ψbi − Vc − kT/q))0.5 where Na is the doping concentration of Ge (1016 cm−3) and ψbi is the built-in potential barrier height in the semiconductor as ψbi = ϕB − ϕn. ϕn = kT/q × ln(Nv/Na) where Nv is the effective states density of the valence band of Ge (5.7 × 1018 cm−3). On the basis of these models, Dit is estimated to be about 2.6 × 1012 cm−2 eV−1. A series resistance Rs of the junction of about 3 kΩ is extracted by a linear fitting of the forward I–V characteristic at a high voltage bias (Extended Data Fig. 3d).

Characterization

Characterization of the graphene film and the devices was performed using a confocal Raman spectrometer (Jobin Yvon Lab RAM HR800), an optical microscope (Nikon LV100ND), a scanning electron microscope (FEI XL30 SFEG using an accelerating voltage of 10 kV) and an atomic force microscope (Bruker Dimension Icon AFM). The transistors were measured using a semiconductor analyser (Agilent B1500A with a capacitance measurement unit B1500A-A20) and a probe station (Cascade Microtech 150-PK-PROMOTION) at room temperature, and a vacuum probe station (Lake Shore TTPX/TSM1D1001) at low temperature.

For device uniformity, taking the device with a 3-μm gap as an example, 20 devices were fabricated on a wafer. The transfer characteristics (Vc = −2 V) show that for SS, sample size = 20, mean = 0.60 mV dec−1 and standard deviation = 0.29 mV dec−1, whereas the output characteristics (Vb = −4 V) show that for PVR, sample size = 20, mean = 15.80 and standard deviation = 2.84. The uniformity can be improved by advanced processes, such as using a Gr-on-Ge wafer where the Gr is grown directly on Ge instead of manual Gr transfer48.

Opportunities and challenges

The SEHC mechanism can be applied to devices composed of different materials. For example, using a carbon nanotube film/n-Ge junction, n-type HOETs can be fabricated (Extended Data Fig. 10). Complex circuits can be realized by using both the n-type and p-type HOETs. However, for the HOET to ultimately become practical, a series of problems must be solved including increasing the range of the current with an SS less than 60 mV dec−1, improving the PVR in the NDR applications and reducing the hysteresis in the characteristics.

At present, although the on-current of the transistor is high, the off-current is also high, resulting in a limited current range with a SS less than 60 mV dec−1, which is not an intrinsic result of the SEHC mechanism. The off-current should be reduced and the on-current increased to improve the on-to-off current ratio. To reduce the off-current, (1) the quality of the Gr/semiconductor interface should be improved by using a Gr-on-Ge wafer where the Gr is grown directly on Ge instead of being transferred48 and (2) other material combinations can be used, such as Gr combined with wider-bandgap semiconductors, where the potential barrier height of the Gr/semiconductor junction is higher and the intrinsic carrier concentration in the semiconductor is lower. To increase the on-current, the energy of the carriers injected from the base-Gr should be further increased, perhaps by using asymmetric potential barriers of the base/substrate and emitter/substrate junctions, which can be achieved by using different substrate semiconductor materials under the base-Gr and the emitter-Gr. These requirements are also necessary for improving the PVR in the NDR applications.

Typical transfer and output characteristics show that the width of the hysteresis window is about 0.25 V and 0.23 V, respectively. However, it should be noted that the hysteresis is not caused by the SEHC mechanism intrinsically, but by the limited quality of the Gr/Ge interface. Because of the contamination and imperfection during the transfer and fabrication process, the Gr/Ge junction itself shows a large hysteresis. This can be reduced by using a Gr-on-Ge wafer where the Gr is grown directly on Ge instead of being transferred or by using an encapsulation.