Simulation of AlGaN/GaN HEMTs’ Breakdown Voltage Enhancement Using Gate Field-Plate, Source Field-Plate and Drain Field Plate

Abstract

A 2-D simulation of off-state breakdown voltage (V_BD) for AlGaN/GaN high electron mobility transistors (HEMTs) with multi field-plates (FPs) is presented in this paper. The effect of geometrical variables of FP and insulator layer on electric field distribution and V_BD are investigated systematically. The FPs can modulate the potential lines and distribution of an electric field, and the insulator layer would influence the modulation effect of FPs. In addition, we designed a structure of HEMT which simultaneously contains gate FP, source FP and drain FP. It is found that the V_BD of AlGaN/GaN HEMTs can be improved greatly with the corporation of gate FP, source FP and drain FP. We achieved the highest V_BD in the HEMT contained with three FPs by optimizing the structural parameters including length of FPs, thickness of FPs, and insulator layer. For HEMT with three FPs, FP-S alleviates the concentration of the electric field more effectively. When the length of the source FP is 24 μm and the insulator thickness between the FP-S and the AlGaN surface is 1950 nm, corresponding to the average electric field of about 3 MV/cm at the channel, V_BD reaches 2200 V. More importantly, the 2D simulation model is based on a real HMET device and will provide guidance for the design of a practical device.

1. Introduction

Gallium nitride (GaN)-based high electron mobility transistor (HEMT) has become an attractive candidate for high power applications, due to integrating lots of outstanding physical properties like high breakdown voltage, high frequency application and low on-resistance [1,2,3]. For power devices, the property of high breakdown voltage is particularly significant. On account of the limit of technique process, the off-state breakdown voltage (V_BD) of AlGaN/GaN HEMTs is still far from the limitation of GaN material [4].

Field-plate (FP) technology is expected to be a feasible and effective way to increase the breakdown voltage by reducing the peak value of an electric field along the channel [5]. From the perspective of field-plate position, FPs can be divided into gate field-plate (FP-G), source field plate (FP-S), and drain field plate (FP-D). Zhang, N.Q. et al. first proposed the FP-G and achieved a high V_BD of 570 V, but resulting in severe degradation of frequency characteristics [6]. Since then, multiple grating FP-G [7,8] and novel FP-G structures [9,10] have been invented to reduce the frequency degradation. Relevant analytical models [11,12] and reliability improvement [13] were also reported, providing a deeper insight into the relationship between FPs and V_BD. In terms of FP-S, a T-shaped FP-S was designed by Mao Wei et al. and it attained high V_BD and high efficiency remarkably [14]. In order to enhance V_BD more effectively, the combination of both FP-G and FP-S is inevitable [15,16,17]. An extremely high V_BD of 8300 V was achieved by using the thick poly-AlN passivation on HEMTs with both FP-G and FP-S [18]. Toshiki Kabemura et al. made a 2-D analysis of breakdown characteristics of FP HEMTs with a high-k passivation layer, confirming that the V_BD would increase with relative permittivity increasing [19]. For investigating the impact of FP-G and FP-S on the capacitances, Aamir et al. modeled the bias dependence of terminal capacitances, and the proposed model is in excellent agreement with measured data [20]. Meanwhile, the incorporation of FP-S and FP-D was considered to be an effective way to improve V_BD and reduce the on-resistance to about 0.6 mΩ·cm² [21]. However, in comparison with FP-G and FP-S, FP-D is rarely induced to enhance V_BD but mostly induced to improve reverse breakdown voltage [22]. In fact, FP-D can reduce the undesired electric field peak from the metal peak introduced by annealing [23]. It is necessary to combine the three FPs. Up to now, the properties of HEMT with FP-G, FP-S and FP-D simultaneously are seldom discussed. At the same time, considering the complexity of structure, more structural parameters such as length of FPs, thickness of FPs and insulator layer are also needed to be optimized in detail. Hence, systematic simulation is necessary and will provide the guidance for experiment.

In this paper, the effect of geometrical variables of FP and insulator layer on field distribution and V_BD were investigated systematically. The electric field distribution and V_BD under different FP were compared by simulation using Silvaco TCAD [24]. Moreover, the FP-G, FP-S and FP-D are employed simultaneously in an AlGaN/GaN HEMT to improve V_BD effectively. The physical models are illustrated in Section 2. The parameters including length of FPs, thickness of FPs, and insulator layer of FPs were optimized for improving V_BD. The breakdown characteristics of HEMT with FP-G, FP-S and FP-D were simultaneously discussed.

2. Physic Models

The two-dimensional numerical simulations were carried out by Silvaco TCAD. The simulated model was based on the experimental structure. Figure 1 shows a not-to-scale cross-section view of the device structure analyzed in this paper. The gate length L_G was 3 μm, the source-to-gate distance L_SG was 9 μm, and the gate-to-drain distance L_GD was 22 μm. The field-plate length L_FP-G’ was 2 μm. The insulator was assumed to be SiO₂ in this simulation, which was usually used in our lab. The permittivity of SiO₂ was set to 5, corresponding to the value of the experimental measurement. Likewise, Si₃N₄ [25] and Al₂O₃ [26] could also be selected as passivation, and the kind of insulator would not change the trend of V_BD versus FPs [19]. The thickness of insulator layer t₁ and t₂; length of FPs L_FP-G, L_FP-S and L_FP-D; and the thickness of FPs t_FP-G and t_FP-S are variable in the following Section. The leakage current [27], especially from the buffer layer, may result in electron injection from the channel into the buffer layer and a large drain current [28]. Setting traps in the GaN buffer is a usual way to decrease the leakage current at pinch-off state [19,29,30], but it may cause a no-convergence problem. Moreover, we aimed to present the relevance between electric field distributions and FPs clearly and exclude the punch-through from the buffer layer. Therefore, with the purpose of simplifying calculation, we set the GaN buffer layer to be 200 nm, reducing the influence of buffer leakage current flow from source to drain [31]. Besides, the thicknesses of the Al_0.23Ga_0.77N barrier and GaN channel are 25 nm and 10nm respectively.

Detailed simulation conditions are as follows. The 1×10¹³ cm⁻² positive charges were placed along the Al_0.23Ga_0.77N/GaN heterojunction to create the two-dimension electron gas, and the channel mobility was set to 1500 cm²/V·s [32]. Also, the field-dependent mobility model and Shockley-Read-Hall (SRH) Recombination model were used. Thermal impact is common in the experiments [33,34], but we did not discuss it during the simulation. The breakdown voltage simulation in this paper was based on the Selberherr’s impact ionization model. The ionization coefficients were AN = AP = 2.98 × 10⁸ and BN = BP = 3.44 × 10⁷ [35]. Moreover, simulations about breakdown performance were all carried out with the gate biased at −6 V, keeping devices on the off-state. Finally, V_BD was defined as the drain voltage when the peak electric field in the channel reached 3 MV/cm [10,14,22].

3. Influences of Structural Parameters of Different Field-Plates

3.1. Gate Field-Plate

We first discuss the influence of gate field-plate on V_BD. Figure 2 shows the device structure only with FP-G. Three parameters are studied, including the insulator layer thickness t₁, the FP-G length L_FP-G and the FP-G thickness t_FP-G. Figure 3a shows the distributions of electric field along the heterojunction interface for t₁ from 100 nm to 300 nm when L_FP-G is 4μm and t_FP-G is 100 nm. For the sake of presenting the electric field peaks intuitively, we only display the electric field distributions partly around the region from gate to drain. As shown in Figure 3a, for thin insulator layer t₁ (t₁ < 250 nm), the electric field at the FP-G edge is higher than the electric field at the gate edge. Due to the modulation of FP-G, the electric field at the FP-G edge reaches the breakdown standard first. Consequently, V_BD increases with the increase of t₁. However, when t₁ further increases, the modulated effect of FP-G becomes weaker. Thus, the electric field peak shifts to the gate edge, which leads to the decrease of V_BD. For the structure in Figure 2, the modulate effect induced by FP-G is optimal when t₁ = 250 nm, as shown in Figure 3b.

Then, we change the L_FP-G from 0 μm to 20 μm and keep t₁ fixed at 250 nm. Without a field plate at the drain side, the electric field at the gate edge increases rapidly and reaches the critical breakdown field 3 MV/cm at around V_D = 55 V. For short L_FP-G (L_FP-G<4 μm), the modulate effect of field-plate is too small to reduce the electric field peak at the gate edge, so V_BD increases slowly with the increase of L_FP-G. As shown in Figure 3c,d, with the optimized t₁, L_FP-G from 4 μm to 8 μm, all can reduce the electric field peak value along the channel effectively, and the electric field peak value of the gate edge reaches 3 MV/cm until V_D = 448 V. When the L_FP-G is more than 20 μm, V_BD will decrease. It is attribute to the narrower distance between the field-plate edge and the drain, which leads to the electric field at the FP-G edge reaches the critical breakdown electric field at around V_D = 435 V.

The following parameter is FP-G thickness t_FP-G. It is set to 100 nm, 300 nm, 400 nm, 800 nm and 8 μm, respectively. In previous researches, there were few studies on the relationship between breakdown characteristics and the thickness of the field-plate. As shown in Figure 3e, V_BD increases with the increase of t_FP-G. This phenomenon shows that thicker FP-G can reduce the electric field at the FP-G edge. However, the improvement of V_BD is not obvious even when the t_FP-G reaches 8 μm. In view of the highest V_BD achievement with less changes in technology progress, we choose t_FP-G = 100 nm as a suitable value for following researches.

3.2. Source Field-Plate

The FP-S is usually used in the GaN-based HEMTs in combination with the gate field-plate, with the aim of achieving a great improvement of V_BD. Figure 4 shows the structure with FP-S. Firstly, the relationship between FP-S and V_BD is studied in the HEMT without the FP-G. The thickness of the gate was fixed at 100 nm. From Figure 5, the effect of FP-S on V_BD is similar to that of FP-G. With regard to t₂, the optimum value of t₂ is less than that of t₁. Because in this device, when the FP-G is fixed to −6 V, the FP-S is fixed to 0 V, resulting in a higher electric field at the FP-S edge than that at the FP-G edge. Hence, to achieve the same effect of modulating the electric field by FP, the FP-S should be closer to the AlGaN surface. The variation trend of V_BD with the length and thickness of FP-S is the same as that of FP-G. As we know, the electric field is in direct proportion to the electric potential difference, and the latter can be performed by the density of the potential line distribution. Therefore, the higher electric field corresponds to the narrower potential lines. From the potential line distributions of Figure 5g,h, we find that potential lines in (g) of FP-G edge are less crowded than that in (h) of FP-S edge even at high drain voltage, which is in accordance with electric field distributions [11]. In a word, in order to enhance V_BD, FP-G is more efficient than FP-S when the structure only has one FP.

3.3. Drain Field-Plate

For FP-D, it is usually used to improve the reverse breakdown voltage. In fact, the electric field peak from the metal peak introduced by annealing could also be reduced by FP-D. However, the electric field peak does not appear on the drain edge in the Silvaco simulation, because of the flat and smooth edge of ohmic electros. Thus, single FP-D usually has little impact on electric field distribution along the channel. However, the electric field peak at the FP-S edge will increase if the distance between FP-S and FP-D is too close. Therefore, we only explored the relationship between FP-D length and V_BD when the structure has both FP-S and FP-D, as shown in Figure 6. Figure 7 shows that when L_FP-D>13 μm, V_BD decreases due to the narrow distance between the two FPs. In consideration of passivation quality in the practical experiment, breakdown may occur between the two FPs. So, it may be more appropriate when L_FP-D is set to 1 μm.

4. Devices Contained with FP-G, FP-S, and FP-D

For much higher V_BD, there is no doubt that the FP-G should be combined with the FP-S [20]. Figure 8 shows the structure of the device contained with FP-G, FP-S and FP-D simultaneously. The V_BD is affected by seven variables, t₁, t₂, L_FP-G, L_FP-S, L_FP-D, t_FP-G, and t_FP-S. Since many parameters should be optimized, the procedure is significant. Because the electric field peak at the gate edge is mainly affected by FP-G, it can be suppressed by a moderate value of t₁, L_FP-G and t_FP-G. Therefore, we can fix these three parameters to adjust the electric field peak at the gate edge, close to but not higher than 3 M/cm. At the same time, less changes in the previous structure and little influence of FP thickness in enhancing V_BD, t_FP-S and L_FP-D can be confirmed too. Thus, we can confirm five parameter values firstly, as shown in Table 1 Then we can focus on the other two parameters, t₂ and L_FP-S.

As shown in Figure 9, for different L_FP-S which vary from 18 μm to 28 μm, the moderated value (up to the highest V_BD) of t₂ is different. We can find the reason in Figure 10a. Only when three peaks are uniform and close to 3 MV/cm, could we achieve the highest V_BD. Meanwhile, comparing Figure 11a with Figure 11b, the potential lines distribution of the former are more crowded than that of the latter, resulting in a lower V_BD [14].

However, no matter what the value of t₂ is, there is the highest V_BD of all lengths for a FP-S of 24 μm. When L_FP-S is between 24 μm to 26 μm, combined with the optimal t₂ = 1.7 μm, we can acquire the highest V_BD of 2200 V. Then, when L_FP-S is 28 μm, due to the narrow distance between FP-S and FP-D, the electric field peak of the FP-S edge will become higher, which makes V_BD decrease. Figure 10b shows the electric field distributions for different t₂ along the channel in HEMT when L_FP-S is fixed to 24 μm. When t₂ = 1950 nm, the two electric field peaks of FP-G edge and FP-S edge are approximate but not high enough to cause breakdown. So, when V_BD is around 2200 V, the highest electric field peak appears at the gate edge. Hence, the crucial way to enhance breakdown voltage is to adjust relative FP parameters. By means of t₂ and L_FP-S, the two electric field peaks at each FPs edge could be uniformed to be closely equal but less than 3 M/cm; and if the highest electric field peak appears at the gate edge, we would acquire the highest V_BD.

5. Conclusions

In this paper, the effect of geometrical variables of FP and insulator layer on field distribution and V_BD were investigated systematically. An AlGaN/GaN HEMT with FP-G, FP-S and FP-D simultaneously was also studied. On account of the incorporation of three FPs, the structure becomes more complex, and more parameters are needed for it to be optimized. Therefore, systematic simulation of these parameters could provide a correct direction for the design of a practical device. By Silvaco TCAD, we optimize the length of FPs, thickness of FPs, and the insulator layer, and acquire the highest V_BD = 2200 V. The mechanism of V_BD enhancement by inducing FPs is the effective modulation of potential lines distribution in the channel. By studying the potential lines and electric field distribution along the AlGaN/GaN heterojunction interface, we can draw some conclusions. In terms of kinds of FP, FP-G can modulate potential lines distribution more uniformly and suppress the electric field peak at FP edge more effectively, compared with FP-S and FP-D. In terms of multiple field-plate combinations, the crucial way to enhance breakdown voltage is to adjust the relative FP parameters. Consequently, the electric field peaks at gate edge, FP-G edge and FP-S edge are uniform to be closely equal but lower than 3 M/cm. Particularly, L_FP-S and t₂ play the most important roles in modulating electric field peaks at FP-G and FP-S edge. Furthermore, the field-plate thickness should be taken into account for the inducing field-plate, in which the breakdown voltage increases with the increase of the field-plate thickness.

Author Contributions

Conceptualization, B.L. and Q.Z.; methodology, B.L. and J.Q.; validation, Q.Z.; formal analysis, B.L. and Q.Z.; writing—original draft preparation, B.L. and Q.Z.; writing—review and editing, H.W.; supervision, H.W.

Funding

This work was supported by Science and Technologies plan Projects of Guangdong Province (Nos. 2017B010112003, 2017A050506013), and by Applied Technologies Research and Development Projects of Guangdong Province (Nos. 2015B010127013, 2016B010123004), and by Science and Technologies plan Projects of Guangzhou City (Nos. 201504291502518, 201604046021, 201704030139, 201905010001), and by Science and Technology Development Special Fund Projects of Zhongshan City (Nos. 2017F2FC0002, 2017A1009).

Conflicts of Interest

The authors declare no conflict of interest. The funders had no role in the design of the study.

Figures and Table

Figure 1. Schematic of HEMT with FP-G, FP-S and FP-D.

Figure 2. Device structure with FP-G analyzed in this study.

Figure 3. The electric field distributions along the channel for different values of (a) t1, (c) LFP-G, and (e) tFP-G. The VBD for different values of (b) t1, (d) LFP-G, and (f) tFP-G.

Figure 3. The electric field distributions along the channel for different values of (a) t1, (c) LFP-G, and (e) tFP-G. The VBD for different values of (b) t1, (d) LFP-G, and (f) tFP-G.

Figure 4. Device structure with FP-S analyzed in this study.

Figure 5. The electric field distributions along the channel for different values of (a) t2, (c) LFP-S, and (e) tFP-S. Off-state breakdown voltage VBD for different values of (b) t2, (d) LFP-S, and (f) tFP-S. The Potential line distributions along the channel when (g) t1 = 250 nm, LFP-G = 4 μm, tFP-G = 1.4 μm, VBD = 521 V, and (h) t2 = 120 nm, LFP-S = 18 μm, tFP-S = 1.4 μm, VBD = 473 V.

Figure 5. The electric field distributions along the channel for different values of (a) t2, (c) LFP-S, and (e) tFP-S. Off-state breakdown voltage VBD for different values of (b) t2, (d) LFP-S, and (f) tFP-S. The Potential line distributions along the channel when (g) t1 = 250 nm, LFP-G = 4 μm, tFP-G = 1.4 μm, VBD = 521 V, and (h) t2 = 120 nm, LFP-S = 18 μm, tFP-S = 1.4 μm, VBD = 473 V.

Figure 6. Device structure with FP-S and FP-D analyzed in this study.

Figure 7. The electric field distributions along the channel for different values of (a) LFP-D. Off-state breakdown voltage VBD for different values of (b) LFP-D.

Figure 8. Device structure contained with FP-G, FP-S, and FP-D.

Figure 9. VBD as a function of t2 and LFP-S.

Figure 10. (a) The electric field distributions along the channel in HEMT with different LFP-S under highest VBD; (b) the electric field distributions along the channel in HEMT when LFP-S = 24.

Figure 11. The Potential distributions in HEMT. (a) LFP-S = 18 μm with t2 = 1250 nm at VBD = 1420 V; (b) LFP-S = 24 μm with t2 = 1950 nm at VBD = 2200 V.

Table 1

Optimized parameter values.

Parameter	Optimized Value
t1	250 nm
LFP-G	4 μm
tFP-G	100 nm
tFP-S	1.4 μm
LFP-D	1 μm