Improved MRD 4H-SiC MESFET with High Power Added Efficiency

An improved multi-recessed double-recessed p-buffer layer 4H–SiC metal semiconductor field effect transistor (IMRD 4H-SiC MESFET) with high power added efficiency is proposed and studied by co-simulation of advanced design system (ADS) and technology computer aided design (TCAD) Sentaurus software in this paper. Based on multi-recessed double-recessed p-buffer layer 4H–SiC metal semiconductor field effect transistor (MRD 4H-SiC MESFET), the recessed area of MRD MESFET on both sides of the gate is optimized, the direct current (DC), radio frequency (RF) parameters and efficiency of the device is balanced, and the IMRD MESFET with a best power-added efficiency (PAE) is finally obtained. The results show that the PAE of the IMRD MESFET is 68.33%, which is 28.66% higher than the MRD MESFET, and DC and RF performance have not dropped significantly. Compared with the MRD MESFET, the IMRD MESFET has a broader prospect in the field of microwave radio frequency.

In this paper, we reported a 4H-SiC MESFET with improved multi-recessed double-recessed p-buffer layer (IMRD) structure. Traditional 4H-SiC MESFETs have been experimentally verified. Many experimental results have been reported so far [18,19], but they are all fixed structures, and the effect of structural parameters on the results has not been studied. Based on multi-recessed 4H-SiC MESFETs with double-recessed p-buffer layer (MRD 4H-SiC MESFET) [20], we adopt a new design method optimized by technology computer aided design (TCAD) simulation and verified in advanced design system (ADS) software. The IMRD MESFET has both the excellent performance of MRD MESFET and better power-added efficiency (PAE) and provides a new idea for high-power operational amplifier design at the device level. Figure 1a,b are the cross-sectional views of the MRD MESFET and the IMRD MESFET, respectively. In Figure 1a, the MRD MESFET contains a high-purity semi-insulating substrate (SI-Substrate), a p-type buffer layer (P-Buffer), an n-type channel layer (N-Channel), and two doped n-type cap layers (Source and Drain), the Nickel Schottky gate has a work function of 5.1 eV. By high energy ion implantation and high temperature annealing processes, two recessed areas are formed on both sides of the gate. The main difference between the two devices is the recessed regions on both sides of the gate. The length and width on both sides are L 1 = 0.15 µm, L 2 = 0.2 µm, H 1 = 0.1 µm and H 2 = 0.1 µm, other parameters are shown in Table 1. In the optimized IMRD MESFET, L 1 = 0.5 µm, L 2 = 0.2 µm, H 1 = 0.15 µm and H 2 = 0.2 µm, and the other parameters are consistent with the MRD MESFET.

Device Structure and Description
Micromachines 2019, 10, x FOR PEER REVIEW 2 of 7 MRD MESFET and better power-added efficiency (PAE) and provides a new idea for high-power operational amplifier design at the device level. Figure 1a,b are the cross-sectional views of the MRD MESFET and the IMRD MESFET, respectively. In Figure 1a, the MRD MESFET contains a high-purity semi-insulating substrate (SI-Substrate), a p-type buffer layer (P-Buffer), an n-type channel layer (N-Channel), and two doped ntype cap layers (Source and Drain), the Nickel Schottky gate has a work function of 5.1 eV. By high energy ion implantation and high temperature annealing processes, two recessed areas are formed on both sides of the gate. The main difference between the two devices is the recessed regions on both sides of the gate. The length and width on both sides are L1 = 0. 15    The 2D TCAD simulator, Sentaurus is used in this paper. The simulation temperature is set to 300 K. A 5.1 eV Nickel Schottky gate work function is applied. The main model used in the simulation are Mobility (Enormal Doping Dep HighFieldsaturation (GradQuasiFermi)), Recombination (Auger SRH (DopingDep)), Incomplete Ionization, Effective Intrinsic and Density (BandGapNarrowing (OldSlotboom)) [21]. The interface state has a great impact on the device, in this paper, the gate of the two devices are formed by a metal-to-SiC contact to form a Schottky contact, so the gate does not have to consider the interface state. When the MESFET is passivated with a material such as Si3N4 or SiO2, the performance is slightly degraded. When verifying the conventional 4H-SiC MESFET, Si3N4  The 2D TCAD simulator, Sentaurus is used in this paper. The simulation temperature is set to 300 K. A 5.1 eV Nickel Schottky gate work function is applied. The main model used in the simulation are Mobility (Enormal Doping Dep HighFieldsaturation (GradQuasiFermi)), Recombination (Auger SRH (DopingDep)), Incomplete Ionization, Effective Intrinsic and Density (BandGapNarrowing (OldSlotboom)) [21]. The interface state has a great impact on the device, in this paper, the gate of the two devices are formed by a metal-to-SiC contact to form a Schottky contact, so the gate does not have to consider the interface state. When the MESFET is passivated with a material such as Si 3 N 4 or SiO 2 , the performance is slightly degraded. When verifying the conventional 4H-SiC MESFET, Si 3 N 4 was used as the passivation layer. When Si 3 N 4 is used as the passivation layer, the device performance is reduced by less than SiO 2 [22,23]. As a theoretical analysis, the influence of passivation on the device is not considered here. After obtaining the simulation results, the obtained model parameters are used in the ADS software to measure the PAE of the two devices. In the ADS simulation, the bias conditions of the device are shown in Figure 2. Direct current (DC) voltage Vg1 is −3.5 V, Vdd is 28 V, input power Pavs is 24 dBm, and frequency f is 1.2 GHz.

Device Structure and Description
Micromachines 2019, 10, x FOR PEER REVIEW 3 of 7 was used as the passivation layer. When Si3N4 is used as the passivation layer, the device performance is reduced by less than SiO2 [22,23]. As a theoretical analysis, the influence of passivation on the device is not considered here. After obtaining the simulation results, the obtained model parameters are used in the ADS software to measure the PAE of the two devices. In the ADS simulation, the bias conditions of the device are shown in Figure 2. Direct current (DC) voltage Vg1 is −3.5 V, Vdd is 28 V, input power Pavs is 24 dBm, and frequency f is 1.

Effect of the Length and Height of the Recessed Regions on the PAE
The effect of the length and height of the recessed regions on PAE is shown in Figure 3. When a parameter changes, the remaining parameters are the default values in Figure 1a. We can see in Figure  3a, when L1 increases, PAE also increases. When L1 reaches 0.5 μm, PAE reaches a maximum value; when L2 is less than 0.1 μm, the PAE increases with L2. When L2 reaches 0.2 um, PAE reaches a maximum value. When L2 is larger than 0.1 μm, PAE decreases with the increase of L2. In Figure 3b, the trend of the effect of H1 and H2 on PAE agrees well. When H1 and H2 are less than 0.2 μm, PAE of H1 and H2 increase with the increase of H1 and H2. When H1 and H2 reach 0.2 μm, PAE of H1 and H2 reach the maximum value. When H1 and H2 are larger than 0.2 μm, PAE of H1 and H2 decrease as H1 and H2 increases.

Effect of the Length and Height of the Recessed Regions on the PAE
The effect of the length and height of the recessed regions on PAE is shown in Figure 3. When a parameter changes, the remaining parameters are the default values in Figure 1a. We can see in Figure 3a, when L 1 increases, PAE also increases. When L 1 reaches 0.5 µm, PAE reaches a maximum value; when L 2 is less than 0.1 µm, the PAE increases with L 2 . When L 2 reaches 0.2 um, PAE reaches a maximum value. When L 2 is larger than 0.1 µm, PAE decreases with the increase of L 2 . In Figure 3b was used as the passivation layer. When Si3N4 is used as the passivation layer, the device performance is reduced by less than SiO2 [22,23]. As a theoretical analysis, the influence of passivation on the device is not considered here. After obtaining the simulation results, the obtained model parameters are used in the ADS software to measure the PAE of the two devices. In the ADS simulation, the bias conditions of the device are shown in Figure 2. Direct current (DC) voltage Vg1 is −3.5 V, Vdd is 28 V, input power Pavs is 24 dBm, and frequency f is 1.

Effect of the Length and Height of the Recessed Regions on the PAE
The effect of the length and height of the recessed regions on PAE is shown in Figure 3. When a parameter changes, the remaining parameters are the default values in Figure 1a. We can see in Figure  3a, when L1 increases, PAE also increases. When L1 reaches 0.5 μm, PAE reaches a maximum value; when L2 is less than 0.1 μm, the PAE increases with L2. When L2 reaches 0.2 um, PAE reaches a maximum value. When L2 is larger than 0.1 μm, PAE decreases with the increase of L2. In Figure 3b, the trend of the effect of H1 and H2 on PAE agrees well. When H1 and H2 are less than 0.2 μm, PAE of H1 and H2 increase with the increase of H1 and H2. When H1 and H2 reach 0.2 μm, PAE of H1 and H2 reach the maximum value. When H1 and H2 are larger than 0.2 μm, PAE of H1 and H2 decrease as H1 and H2 increases.

Optimized Results and Mechanism Discussion
Through further analysis of the results obtained in 3.1, we found that between L 1 , L 2 , H 1 and H 2 there is a relatively independent relationship. When L 2 , H 1 and H 2 take different values, the trend of PAE with L 1 is almost the same as Figure 3a. The maximum value of PAE is found in L 1 = 0.5 µm.
In addition, for L 2 , H 1 and H 2 , the maximum value of PAE are found in L 2 = 0.2 µm, H 1 = 0.2 µm and H 2 = 0.2 µm. Based on the above results, L 1 = 0.5 µm, L 2 = 0.2 µm, H 1 = 0.2 µm and H 2 = 0.2 µm were selected as the optimal structural parameters, and the optimized device was obtained. The main parameters of the device are shown in Table 2. It can be seen from the table that although its saturated drain current I dsat is too small, which is 35.2% lower than that of the MRD MESFET, and the DC characteristics are greatly weakened, its PAE reaches 70.85%, which is 33.40% higher than that of the MRD MESFET.  Figure 4 shows the effect of the length and height on I dsat . It can be seen from Figure 4a that when L 1 = 0.5 µm and L 2 = 0.2 µm, the decline of I dsat is small; In Figure 4b, it can be clearly seen that when H 1 is less than 0.15 µm, I dsat decreases, but the value of decrease is not very large. When H 1 is greater than 0.15 µm, I dsat drops sharply. Similarly, for H 2 , when H 2 is less than 0.2 µm, the decrease of I dsat is not large. When H 2 is larger than 0.2 µm, the decline of I dsat is more obvious. The main cause of the weakening of the DC characteristics is that the thickness of the channel region becomes extremely thin, resulting in narrowing the channel of most electrons, and thus the DC characteristics are deteriorated. Based on the results above, L 1 = 0.5 µm, L 2 = 0.2 µm, H 1 = 0.15 µm and H 2 = 0.2 µm were selected as structural parameters, and the PAE was measured to be 68.33%. After optimization, the PAE of the IMRD MESFET is 28.66% higher than that of the MRD MESFET. Through further analysis of the results obtained in 3.1, we found that between L1, L2, H1 and H2 there is a relatively independent relationship. When L2, H1 and H2 take different values, the trend of PAE with L1 is almost the same as Figure 3a. The maximum value of PAE is found in L1 = 0.5 μm. In addition, for L2, H1 and H2, the maximum value of PAE are found in L2 = 0.2 μm, H1 = 0.2 μm and H2 = 0.2 μm. Based on the above results, L1 = 0.5 μm, L2 = 0.2 μm, H1 = 0.2 μm and H2 = 0.2 μm were selected as the optimal structural parameters, and the optimized device was obtained. The main parameters of the device are shown in Table 2. It can be seen from the table that although its saturated drain current Idsat is too small, which is 35.2% lower than that of the MRD MESFET, and the DC characteristics are greatly weakened，its PAE reaches 70.85%, which is 33.40% higher than that of the MRD MESFET.   Figure 4b, it can be clearly seen that when H1 is less than 0.15 μm, Idsat decreases, but the value of decrease is not very large. When H1 is greater than 0.15 μm, Idsat drops sharply. Similarly, for H2, when H2 is less than 0.2 μm, the decrease of Idsat is not large. When H2 is larger than 0.2 μm, the decline of Idsat is more obvious. The main cause of the weakening of the DC characteristics is that the thickness of the channel region becomes extremely thin, resulting in narrowing the channel of most electrons, and thus the DC characteristics are deteriorated. Based on the results above, L1 = 0.5 μm, L2 = 0.2 μm, H1 = 0.15 μm and H2 = 0.2 μm were selected as structural parameters, and the PAE was measured to be 68.33%. After optimization, the PAE of the IMRD MESFET is 28.66% higher than that of the MRD MESFET.    H) on Vt, Gmmax, Cgs and Cgd, the transconductance Gmmax, its physical meaning is the first-order partial conductance of the output drain current and the input voltage, Cgs is the gate-source capacitance, and Cgd is the gate-drain capacitance. In combination with Figure 3, it can be seen from Figure 5a,c,e that as L1 increases, the  Figure 5 illustrates the effect of recessed region parameters (L & H) on V t , Gm max , C gs and C gd , the transconductance Gm max , its physical meaning is the first-order partial conductance of the output drain current and the input voltage, C gs is the gate-source capacitance, and C gd is the gate-drain capacitance. In combination with Figure 3, it can be seen from Figure 5a,c,e that as L 1 increases, the threshold voltage V t , the transconductance Gm max , the gate-source capacitance C gs and the gate-drain capacitance C gd decrease, but the PAE gradually increases; when L 2 increases, V t increases first, then tends to be stationary, Gm max gradually decreases, the change trend of C gd and C gs are about the same. When L 2 is less than 0.1 µm, the changes of C gs and C gd are relatively stable. When L 2 is larger than 0.1 µm, C gs and C gd are gradually increased. Combined with Figures 3a and 5a,c,e, it can be found that the effects of C gs and C gd on PAE are obvious. It can be seen from the bias conditions of Figure 2 that C gs and C gd affect the input impedance and output impedance respectively, and the power consumed by the capacitor is proportional to the size of the capacitor. According to the definition of PAE, these two capacitors affect the input power and output power, so these two capacitors have a greater impact on the PAE. Similarly, in conjunction with Figure 3, it can be seen from Figure 5b,d,f that, when H 1 is less than 0.2 µm, as H 1 increases, V t , Gm max , C gs and C gd decrease, while PAE increases. When H 1 is larger than 0.2 µm, PAE decreases with H 1 . For H 2 , when H 2 increases gradually, V t and Gm max also decrease gradually, but when H 2 is less than 0.05 µm, C gs and C gd decrease with increasing H 2 . When H 2 is greater than 0.05 µm, the change of C gs and C gd are relatively stable. When H 2 is less than 0.2 µm, PAE increases with the increase of H 2 . When H 2 is greater than 0.2 µm, PAE decreases as H 2 increases. Combined with Figures 3b, 4b and 5b,d,f, although the capacitances C gs and C gd have a greater influence on PAE, as the H 2 increases, the conductive channel of the device still decreases. When the thickness of the conductive channel is less than 0.05 µm, the conductive channel is extremely thin, and its conductive properties are greatly weakened. At this time, both the AC signal and the DC signal will be greatly attenuated in the channel, resulting in a decrease in output power. Under the combined action of several factors, the PAE decreases as H 2 increases.
Micromachines 2019, 10, x FOR PEER REVIEW 5 of 7 threshold voltage Vt, the transconductance Gmmax, the gate-source capacitance Cgs and the gate-drain capacitance Cgd decrease, but the PAE gradually increases; when L2 increases, Vt increases first, then tends to be stationary, Gmmax gradually decreases, the change trend of Cgd and Cgs are about the same. When L2 is less than 0.1 μm, the changes of Cgs and Cgd are relatively stable. When L2 is larger than 0.1 μm, Cgs and Cgd are gradually increased. Combined with Figure 3a and Figure 5a,c,e, it can be found that the effects of Cgs and Cgd on PAE are obvious. It can be seen from the bias conditions of Figure 2 that Cgs and Cgd affect the input impedance and output impedance respectively, and the power consumed by the capacitor is proportional to the size of the capacitor. According to the definition of PAE, these two capacitors affect the input power and output power, so these two capacitors have a greater impact on the PAE. Similarly, in conjunction with Figure 3, it can be seen from Figure 5b,d,f that, when H1 is less than 0.2 μm, as H1 increases, Vt, Gmmax, Cgs and Cgd decrease, while PAE increases. When H1 is larger than 0.2 μm, PAE decreases with H1. For H2, when H2 increases gradually, Vt and Gmmax also decrease gradually, but when H2 is less than 0.05 μm, Cgs and Cgd decrease with increasing H2. When H2 is greater than 0.05 μm, the change of Cgs and Cgd are relatively stable. When H2 is less than 0.2 μm, PAE increases with the increase of H2. When H2 is greater than 0.2 μm, PAE decreases as H2 increases. Combined with Figure 3b, Figure 4b and Figure 5b,d,f, although the capacitances Cgs and Cgd have a greater influence on PAE, as the H2 increases, the conductive channel of the device still decreases. When the thickness of the conductive channel is less than 0.05 μm, the conductive channel is extremely thin, and its conductive properties are greatly weakened. At this time, both the AC signal and the DC signal will be greatly attenuated in the channel, resulting in a decrease in output power. Under the combined action of several factors, the PAE decreases as H2 increases.  The power-added efficiency (PAE) is the difference in output and input power in place of radio frequency (RF) output power in the drain efficiency equation, which takes power gain G p into account.
η c is the drain efficiency and represents the ratio of output power to DC input power. In order to obtain a higher PAE, a larger power gain G p is required. It can be seen from the equivalent circuit diagram of the device, the size of the PAE is determined by these parameters together. For the threshold voltage V t , only when the absolute value of V t is reduced, can small voltage make the channel turn on. When the input voltage is constant, the smaller the absolute value of the threshold voltage is, the larger the channel current is, and the larger the output power P out is, so that the PAE becomes larger. When Gm max is reduced, its DC characteristic will also decrease. Since the input voltage is constant, the decrease of Gm max causes the bias current I d to decrease, so that the P dc is reduced, according to the definition of PAE, the decrease of P dc will cause the increase of PAE. The gate-source capacitance C gs and the gate-drain capacitance C gd have a greater influence on the PAE, in the equivalent circuit, C gs and C gd are in the input loop and output loop respectively. Because of the larger C gs in the bias, the more AC input energy it consumes, the larger the C gd , the smaller the AC output power of the load is. In addition, as the channel becomes thinner and thinner, it affects both the capacitance and the carrier's passage through the channel region. Although the PAE increases as the capacitance decreases, the extremely thin channel also makes the DC and AC characteristics affected seriously. From the analysis of 3.1-3.3, when the structural parameters of the device are changed, in order to obtain the optimal PAE, it is necessary to weigh the various parameters and find the optimal structural parameters on the premise that the other performance is guaranteed. Perhaps the above design process will sacrifice a small part of the performance, but the efficiency is greatly improved, achieving energy saving and emission reduction, which is very beneficial to the construction of green earth.

Conclusions
An improved MRD 4H-SiC MESFET with high power added efficiency is analyzed and studied by co-simulation of ADS and TCAD Sentaurus software in this paper. Based on MRD 4H-SiC MESFET, we optimize the MRD MESFET on both sides of the gate. In the recessed area, the DC, RF parameters and efficiency of the device are weighed, and the IMRD MESFET with the best PAE is finally obtained. The results show that the saturation drain current Idsat of the IMRD MESFET is 311 mA, the threshold voltage V t is −6.99 V, the maximum transconductance Gm max is 46.37 mS, the gate-source capacitance C gs is 0.218 pF, and the gate-drain capacitance C gd is 17.9 fF. The PAE is 68.33%, which is 28.66% higher than the MRD MESFET, and DC and RF performance have not dropped significantly. This paper proposes to lay a device-level theoretical basis and design method for further energy-efficient RF power amplifier.