In this study, a split-gate AlGaN/GaN heterostructure field-effect transistor with an auxiliary gate was fabricated. Through experiment and analysis, it was discovered that by applying a constant potential (usually 0 V or negative potential) to the auxiliary gate, a negative bias can be formed between the auxiliary gate and the channel. This consumes the two-dimensional electron gas in the channel, leading to significant improvements in the saturation characteristics of split-gate devices. By applying different potentials to the main gate and the auxiliary gate, a variety of device working modes can be obtained, and the threshold voltage can be altered across a large range. These advantages make split-gate devices with auxiliary gates more suitable for increasingly complex integrated circuit applications.
I. INTRODUCTION
The AlGaN/GaN heterostructure field-effect transistor (HFET), a type of wide-bandgap semiconductor electronic device, possesses the benefits of high breakdown voltage and high electron mobility. These advantages have led to it being widely used in high frequency and high power applications.1–6 As a unique AlGaN/GaN HFET with a novel working mechanism, split-gate devices have shown broad prospects in low power consumption voltage amplifier applications.7 However, when only the open region is turned on, the saturation mechanism of common split-gate devices is similar to that of ungated devices, which is mainly related to the self-heating effect and a virtual gate formed by surface electron injection.7–11 Current saturation caused by the self-heating effect requires a strong electric field, which can only work at large drain-source voltages (VDS). The virtual gate depends on the surface trap state, so its stability and reliability are very poor. Furthermore, there is still some debate about the effect and mechanism of the current saturation affected by the virtual gate.10,11
In this paper, a split-gate AlGaN/GaN HFET with an auxiliary gate was designed. For this newly designed device, the auxiliary gate became the decisive factor affecting device saturation rather than the virtual gate or the self-heating effect. Therefore, the saturation process is no longer dependent on the surface trap state or large VDS values. As a result, device saturation can be realized stably at lower VDS, thus greatly improving the saturation characteristics of the device. Through various potential combinations of the main gate and auxiliary gate, split-gate devices with auxiliary gates can work in several modes with different threshold voltages (Vth), which makes them suitable for complex circuit applications.
II. EXPERIMENTS
The AlGaN/GaN heterostructure material used in the experiment was grown on a 350 µm SiC substrate. Through metal organic chemical vapor deposition (MOCVD), a 1 µm GaN buffer layer, a 400 nm i-GaN channel layer, a 0.8 nm AlN insertion layer, a 15.5 nm Al0.23Ga0.77N barrier layer, and a 3 nm GaN cap layer were grown in sequence. The Hall measurement at room temperature displayed a sheet carrier density of 8.02 × 1012 cm−2 with a mobility of 2122 cm2 V−1 s−1. The device isolation was defined by inductively coupled plasma (ICP) etching. Using electron beam evaporation, Ti/Al/Ni/Au were deposited successively as ohmic contacts followed by rapid thermal annealing at 850 °C for 30 s in the N2 atmosphere. The metal used for Schottky contact was Ni/Au, which was also deposited by electron beam evaporation. As the passivation film, a 150 nm SixN1−x layer was deposited by plasma-enhanced chemical vapor deposition (PECVD). During device fabrication, all patterns were defined by UV-lithography. As Fig. 1 illustrates, four AlGaN/GaN HFETs with different structures and sizes were prepared. Samples 1, 2, and 4 were all split-gate devices, and sample 3 was a normal device. Since sample 4 had two gates, we named the split main gate as gate1 and the normal auxiliary gate as gate2. For samples 1, 2, and 3, although each device had only one gate, we still used gate1 to call it for convenience of expression. For all these four samples, the gate1 length (LG1) was 40 µm, the total channel width (W) was 100 µm, the gate1-source distance (LG1S) was 6 µm, and the gate1-drain distance (LG1D) was 16 µm. Samples 1, 2, and 4 had an opening in the middle of gate1 with a 3 µm opening width (WO). For sample 4, gate2 was located in the middle of gate1 and drain, and the gate2 length (LG2) was 10 µm. Of all the samples, only sample 2 was not passivated. The current–voltage (I–V) characteristics of the devices were measured at room temperature by a Keysight B1500A semiconductor parameter analyzer. The capacitance–voltage (C–V) characteristics were also measured at room temperature using a Keysight B1520A at 1 MHz.
III. RESULTS AND DISCUSSION
The C–V characteristics of the four samples are shown in Fig. 2. During the C–V measurement, gate2 of sample 4 is floating. It can be seen that when the gate1-source voltage (VG1S) is about −2.8 V, the capacitance of the four samples decreases rapidly to a minimum, which means that the two-dimensional electron gas (2DEG) underneath gate1 is almost depleted around VG1S = −2.8 V. When VG1S < −2.8 V, the normal device has been turned off, while the split-gate device can continue to conduct due to the existence of the opening in gate1. After that, the modulation of channel current by gate1 bias mainly works through the fringe electric field and polarization Coulomb field (PCF) scattering. The negative gate1 bias will directly reduce the open region 2DEG density (n2D) through the fringe electric field and further reduce the open region effective width ().12–17 Meanwhile, the gate1 bias will cause PCF scattering through the inverse piezoelectric effect and affect the open region electron mobility (μ).7,18–22 As VG1S decreases, n2D, , and μ will all decrease under the influence of fringe electric field and PCF scattering, resulting in the decrease in channel current, so as to realize the modulation of channel current by gate1 bias.
The I–V characteristics of the four samples are shown in Figs. 3 and 4. Figure 3(a) displays the I–V characteristics of split-gate sample 1 with VG1S values between −40 and −4 V. At this time, only the ungated open region is conductive. The saturation mechanism of the ungated region is related to the self-heating effect and the virtual gate formed by surface electron injection.7–11 Due to a passivation layer growing on the surface of sample 1, the surface trap states are relatively less, so the virtual gate can be disregarded. Therefore, the current saturation in the open region is mainly affected by the self-heating effect, which requires a sufficiently large electric field. As displayed in Fig. 3(a), the saturation voltage of the device under various VG1S is very large, especially when the absolute value of negative VG1S is small. At this time, the channel current increases along the arc with an increase in values of VDS, and there is no obvious current saturation even if VDS reaches 20 V. Sample 2 is a split-gate device with the same device size as sample 1, but it is not passivated, so the effect of the virtual gate on current saturation needs to be considered. The I–V characteristics of split-gate sample 2 are presented in Fig. 3(b). Compared with Fig. 3(a), it can be observed that for the split-gate device, the existence of a virtual gate reduces the saturation voltage to a certain extent, but the improvement in saturation characteristics is not obvious. There are still some controversies about the effect and mechanism of virtual gate modulating current saturation.10,11 In addition, the formation of a virtual gate depends on the trap states on the device surface, so it is difficult to stably obtain a virtual gate with low potential.
Using sample 1 as a basis, sample 4 introduced an auxiliary gate to improve the saturation characteristics of the split-gate device. Figure 4(a) displays the I–V characteristics of sample 4 when VG2S is constant at 0 V and VG1S = −40 to −4 V. It can be seen that VG1S has a very weak ability to modulate drain–source current (IDS), while the effective modulation range is very large, which is similar to the I–V characteristics shown in Figs. 3(a) and 3(b). However, for sample 4 with a VG2S value of 0 V, the saturation voltage is small, and the linear region and the saturation region of the I–V curve can be easily distinguished. This shows that the saturation characteristics are vastly superior to those of samples 1 and 2. The channel potential underneath gate2 rises as VDS increases, while the potential on gate2 is fixed at 0 V. Therefore, a negative potential difference forms between gate2 and the channel below it to consume the channel 2DEG in this region. As VDS increases, the absolute value of the negative potential difference between gate2 and the channel grows larger, and more 2DEG is consumed. Finally, the channel pinches off and the current reaches saturation. Compared to the virtual gate formed by surface electron injection, which is dependent on the surface trap state, the position and potential of the auxiliary gate are given artificially and can be adjusted freely according to specific needs. It is more stable and controllable than the virtual gate and improves the saturation characteristics of the device.
Comparing Figs. 3(a), 3(b), and 4(a), it can also be found that sample 2 can be turned off when VG1S = −32 V, while sample 1 and sample 4 with 0 V VG2S cannot be turned off even if VG1S = −40 V. This means that after passivation of the split-gate device, the modulation ability of the gate1 bias to the channel current is weakened. As previously mentioned, when the 2DEG underneath gate1 is depleted, in addition to the effect of fringe electric field, PCF scattering also modulates the current in the open region by changing electron mobility. PCF scattering originates from the nonuniform barrier strain distribution, which is mainly caused by gate bias through the inverse piezoelectric effect.18,19 The passivation process will introduce additional compressive strain into the ungated region of the AlGaN barrier, which is consistent with the effect of negative bias on the gate region, thus weakening the nonuniform distribution of barrier strain.23 This result will further lead to a more uniform distribution of polarization charge. In other words, the additional polarization charge density reduces, thus weakening the PCF scattering. The ability of PCF scattering to modulate the current is weakened by the passivation process, which makes it more difficult to turn off the split-gate device with the passivation layer.
When VG2S is constant at −1 V and VG1S = −40 to −4 V, the I–V characteristics of sample 4 are shown in Fig. 4(b). This time, the I–V curve of sample 4 still shows the characteristics of weak modulation ability, large effective modulation range, and excellent saturation characteristics, similar to those shown in Fig. 4(a). However, smaller VG2S values lead to an overall reduction in channel current under the same gate1 bias, which makes sample 4 more suitable for low power consumption applications. Figure 4(c) shows that by further reducing the gate2 potential to −2 V, the channel current under the same VG1S of sample 4 also falls. Using the constant current method, when the saturation current reaches the order of 10−7 A, the corresponding gate bias is defined as Vth. It can be found that when VG2S is set to −2 V, the Vth of sample 4 switches from less than −40 to −22 V. The variation in the electrical characteristics of sample 4 with the gate2 bias can be explained from the following aspects. When VG2S changes from 0 to −2 V, for the gate2 region of sample 4, the channel n2D will decrease and the additional polarization charge will increase. Both of these will increase the PCF scattering of channel electrons and further reduce the electron mobility.24 A decrease in both gate2 region n2D and electron mobility will reduce IDS at a constant VG1S, which means that the channel current can reach the order of 10−7 A with a lower |VG1S|, thus resulting in the change in Vth. When gate1 is floating and only a varying voltage is applied to gate2, the I–V characteristics of sample 4 are shown in Fig. 4(d). It can be seen that the VG2S of the device has a strong ability to modulate the IDS, and only a negative VG2S with a small absolute value is needed to turn off the channel current. These characteristics are very similar to those shown in Fig. 3(c), indicating that when the channel current is only modulated by unopened gate2, the working mode of sample 4 is the same as that of the normal device. At this time, the Vth of sample 4 can be obtained as about −3 V.
IV. CONCLUSION
A split-gate AlGaN/GaN heterostructure field-effect transistor with an auxiliary gate was prepared, and a series of experimental tests and theoretical calculations were carried out. By introducing an auxiliary gate with 0 V potential, the saturation characteristics of the split-gate device were greatly improved. With the decrease in VG2S, the channel current will decrease, which makes the split-gate device with an auxiliary gate more suitable for low power consumption applications. In addition, for this unique device, the threshold voltage can be modulated across a wide range, which is difficult to be realized by devices with other structures. Excellent saturation characteristics and various working modes with different threshold voltages make the split-gate device with an auxiliary gate have a very wide prospect in circuitry applications.
ACKNOWLEDGMENTS
This work was supported by the National Natural Science Foundation of China (Grant Nos. 11974210 and 11574182).
AUTHOR DECLARATIONS
Conflict of Interest
The authors have no conflicts to disclose.
DATA AVAILABILITY
The data that support the findings of this study are available from the corresponding author upon reasonable request.