If the successful development of GaN-based material could be considered as a vital catalyst for the invention of GaN-based high electron mobility transistor (HEMT), the field-plate (FP) technology could also be considered as one of the most prominent catalysts for significantly improving the power performance of HEMTs so as to satisfy more and more actual requirements in recent power applications. With the excellent ability of these FP techniques to effectively extend the depletion region and replace a single-peak electric field with several peaks,^[1] the electric field distribution could be uniformed, the device breakdown performance and other relevant reliability performance, such as trapping effect,^[2,3] inverse piezoelectric effect^[4,5] could be improved. So far, various studies on the GaN-based HEMTs with different FP structures have been reported,^[6–18] of which it may be one of the important focuses to improve the device breakdown voltage without increasing device dimension. Therefore, it could be concluded, to a great extent, that in order to fabricate a device with high breakdown voltage and little degradation of its other performances, it may be an effective and valuable method to raise the ability of the FP per unit length to enhance the breakdown voltage as much as possible, which deserves to be explored further.

In this paper, a novel AlGaN/GaN HEMT with a source-connected T-shaped field-plate (ST-FP HEMT) is presented for the purpose of a better breakdown enhancement and a better ability of the FP per unit length to increase the breakdown voltage. Systematic simulations of physical mechanism in ST-FP HEMT, relating to the improvement of the breakdown capability together with the ST-FP modulation effects on the channel electric field and potential distributions, are conducted in detail with Silvaco-ATLAS. In addition, the optimization methods, the relationship between the structure of the ST-FP, the breakdown voltage as well as the FP efficiency in ST-FP HEMT are achieved on the basis of the simulation researches. These research results could provide a valuable guideline for reducing the device design complexity and simplifying the actual manufacture processings of the ST-FP HEMT.

Figure 1 shows the schematics of the S-FP HEMT and the proposed ST-FP HEMT. The HEMT is not shown because of its same structure as that of S-FP HEMT except for none FP. As shown in the figure, the ST-FP is characterized by an S-FP and a trench metal. With the exception of the FP structure, all the devices in this paper have the same device structure, which is similar to that used in Ref. [9]. A 3-μm unintentionally doped (UID) GaN buffer layer and a 30-nm UID Al_0.25Ga_0.75N barrier layer are used. And the gate–source spacing, the gate length, and the gate–drain spacing are 1.5 μm, 1.5 μm, and 10 μm, respectively. The same effective FP length L_FP in S-FP HEMT and ST-FP HEMT is used. Silicon nitride passivation is used and the thickness under the FP is defined as T_SiN. L_GT is the distance from the gate to the trench. L_TF is the distance between the trench and the drain side of the ST-FP. The trench depth and width are denoted with T_D and L_T, respectively.

Fig. 1.

	Figure Option ViewDownloadNew Window
	Fig. 1. Schematics of (a) S-FP HEMT and (b) ST-FP HEMT.

The simulations are performed with Silvaco-ATLAS and the main device model is similar to that in our early researches.^[18,19] The background doping concentrations in the AlGaN layer and the GaN layer are both assumed to be 1×10¹⁵ cm⁻³. The positive charges with a density of 1×10¹³ cm⁻² are located along the AlGaN/GaN heterojunction to model the polarization effect, which is based on the experimental result of the approximate charge neutrality at the AlGaN/passivation layer interface.^[20] Two kinds of deep traps are considered, of which the energy level and the density for the deep acceptor trap are E_C−2.85 eV and 1×10¹⁷ cm⁻³ respectively, and those for the deep donor trap are E_C − 0.5 eV and 2×10¹⁷ cm⁻³. The simulations about the breakdown performance are carried out with the gate biased at −6 V. And the breakdown voltage V_BR is defined as the drain voltage where the peak electric field in the channel reaches a GaN breakdown field of 3 MV/cm, whose breakdown criterion has been successfully implemented in our early researches.^[18,19,21]

In order to evaluate the ability of FP to improve the breakdown voltage reasonably, the FP efficiency η,^[22] defined as a ratio of the increment of the breakdown voltage V_BR to the total effective FP length L_FP, is utilized in this paper. Figure 2 shows the relationship between the L_FP of S-FP and the breakdown voltage V_BR as well as the FP efficiency η in S-FP HEMT. In Fig. 2, each V_BR is approximately the maximum breakdown voltage for a given L_FP, which is obtained by optimizing the passivation thickness T_SiN in Fig. 1(a). It can be observed that the breakdown voltage increases while the FP efficiency decreases with the increase of the L_FP. And when the L_FP is beyond 5 μm, the increment of the V_BR with the L_FP is very small, corresponding to a small change of η. This phenomenon is in accordance with that in Ref. [6]. And the simulated V_BR of 475 V for an S-FP HEMT with L_FP = 5 μm in this paper is very close to the experimental data of 470 V for an S-FP HEMT with the same L_FP in Ref. [9]. These above indicate a good validity of our device model.

Fig. 2.

	Figure Option ViewDownloadNew Window
	Fig. 2. Relationship between the LFP of S-FP and the breakdown voltage VBR as well as the FP efficiency η in S-FP HEMT.

Figure 3 shows the nearly optimized channel electric field distributions for the HEMT and four S-FP HEMTs with different values of L_FP. And the area beneath the contour of each electric field distribution in Fig. 3, could be used to estimate the V_BR of each device approximately. The channel electric field distributions away from the left edge of the gate or the FP are very small, thus they are not shown in Fig. 3. As can be seen, the V_BR of HEMT is only 60 V, with only one electric field peak near the gate edge. And it is evident that two nearly equal electric field peaks are formed at the right edge of the gate and the right edge of the S-FP in each S-FP HEMT, indicating an effective modulation of electric field with the S-FP at higher breakdown voltage. In addition, with the increase of L_FP, the V_BR of S-FP HEMT increases and the electric field peak near the right edge of the S-FP shifts toward the drain. And when the L_FP > 5 μm, the increment of the area beneath the contour of the electric field distribution is very small, leading to a small increment of the V_BR and a small decrement of the η. These results are consistent with the analyses in Fig. 2.

Fig. 3.

	Figure Option ViewDownloadNew Window
	Fig. 3. Nearly optimized channel electric field distributions in HEMT and four S-FP HEMTs.

Careful observations of Fig. 3 could reveal that the value of the electric field between the two electric field peaks in each S-FP HEMT drops and the range of this low electric field is enlarged with the increase of L_FP. This indicates that a considerable part of the S-FP between the gate and the drain edge of the FP, especially for a long S-FP, does not really play an effective role in modulating the channel electric field. Thus, in order to enhance the modulation ability of the S-FP, it is full of significance to take some measures to improve the electric field distribution beneath the S-FP, of which the ST-FP HEMT is a kind of structure with great potential, as illustrated in Fig. 1(b) and Fig. 4.

Fig. 4.

	Figure Option ViewDownloadNew Window
	Fig. 4. Nearly optimized channel electric field distributions for the HEMT, the S-FP HEMT, and the ST-FP HEMT.

Figure 4 shows the nearly optimized channel electric field distributions beneath the field-plate or around the gate for the HEMT, the S-FP HEMT with an S-FP of L_FP = 5 μm and an ST-FP HEMT with an ST-FP of L_FP = 5 μm. The optimal T_SiN in the S-FP HEMT is 0.28 μm. And the optimal structure parameters of the ST-FP HEMT is achieved by optimizing the T_D and T_SiN for given L_GT = 0.5 μm and L_T = 0.5 μm. The optimal T_D and T_SiN are 0.46 μm and 0.5 μm, respectively. It is clear that a new electric field peak beneath the FP is formed in the ST-FP HEMT. And a broad electric field distribution near the drain edge of the FP in the ST-FP HEMT channel is obtained compared with that in the S-FP HEMT. This leads to a high V_BR of 810 V and a high η of 150 V/μm in the ST-FP HEMT, which exhibits a remarkable improvement compared with a V_BR of 475 V and an η of 83 V/μm in the S-FP HEMT.

Fig. 5.

	Figure Option ViewDownloadNew Window
	Fig. 5. Potential distributions in (a) HEMT, (b) S-FP HEMT, (c) ST-FP HEMT.

In order to get a physical insight into the essential mechanism of the modulation of the ST-FP, the potential distributions at off-state corresponding to the three devices in Fig. 4 are investigated as shown in Fig. 5. It can be observed that most of potential lines crowds around the drain side of the gate in HEMT, indicating a high electric field peak could be formed, which is in accord with the result in Fig. 4. And for the S-FP HEMT with an L_FP of 5 μm, the high potential line density appears at two positions: one is located at the drain side of the gate and the other at the drain side of the S-FP. This is attributed to the modulation effect of the S-FP, resulting in two electric field peaks in S-FP HEMTs. For the ST-FP HEMT, a more uniform distribution of potential lines is observed and three high potential line density areas are formed at the gate edge, the trench edge and ST-FP edge, which shows an excellent and effective modulation of potential lines. Thus, a more uniform distribution of electric field as well as an outstanding blocking capability could be obtained in the ST-FP HEMT.

In the ST-FP HEMT, there are many structure parameters of the ST-FP which are needed to be considered during the device optimization. And each parameter may influence the breakdown voltage and the FP efficiency. Here, we take the L_T and the L_GT for example. For the purpose of illustration, the effects of the L_T and the L_GT in the ST-FP HEMT on the electric field distribution, V_BR as well as η are shown in Fig. 6 and Fig. 7, respectively. For devices with a fixed L_GT of 0.5 μm and each case of L_T in Fig. 6 and for devices with a fixed L_T of 0.5 μm and each case of L_GT in Fig. 7, they are all optimized to obtain the maximum V_BR by optimizing the T_D and T_SiN. As shown in Fig. 6, with increasing L_T, the electric field peaks near the drain side of the trench shift toward the drain, and the V_BR only shows a slight decrease, resulting in a slight drop of η. This indicates that the V_BR and η are both weak functions of L_T. When the trench width L_T is set to be 0.5 μm, the V_BR and η show strong functions of L_GT in Fig. 7. The smaller the L_GT, the greater the V_BR and η are. Besides, it could be found that the T_D is always a little smaller than the T_SiN in both figures. Based on these analyses, the optimization procedure of the ST-FP HEMT could be extracted. During the optimization, first, we could choose a proper trench width L_T according to the actual fabrication processing condition. And then for different given gate–trench spacings L_GT, the T_D and T_SiN are optimized to achieve the target breakdown voltage.

Fig. 6.

	Figure Option ViewDownloadNew Window
	Fig. 6. (a) Electric field distributions along the channel and (b) relationship between the LT, VBR as well as η for ST-FP HEMTs with different values of LT.

Fig. 7.

	Figure Option ViewDownloadNew Window
	Fig. 7. ST-FP HEMT with different LGT, (a) Electric field distributions along the channel and (b) relationship between the LGT, VBR as well as η for ST-FP HEMTs with different values of LGT.

The off-state breakdown performance of the ST-FP HEMT with an ST-FP is investigated in this paper. With the help of two-dimensional simulations, ST-FP HEMT exhibits a good blocking capability compared with HEMT and S-FP HEMT. This benefits from the advantage that the ST-FP could effectively modulate the potential distributions, which could extend the whole electric field distribution between gate and drain and force all the electric field peaks around the gate edge, trench edge and ST-FP edge to be approximately equal but less than breakdown field. Simulations also show that the V_BR and η are both weak functions of the trench width L_T and are strong functions of the gate-trench spacing L_GT. These results above are of great value for the design and actual manufacture processes of ST-FP HEMTs.