A simulation study of field plate termination in Ga2O3 Schottky barrier diodes
Wang Hui1, 2, 3, 4, Jiang Ling-Li1, 3, 4, Lin Xin-Peng1, 3, 4, Lei Si-Qi1, 3, 4, Yu Hong-Yu1, 3, 4, †
Department of Electrical and Electronic Engineering, Southern University of Science and Technology, Shenzhen 518055, China
School of Electronic Science and Engineering, Southeast University, Nanjing 210096, China
Shenzhen Key Laboratory of The Third Generation Semiconductor, Shenzhen 518055, China
Guangdong GaN Devices Engineering and Technical Research Center, Shenzhen 518055, China

 

† Corresponding author. E-mail: yuhy@sustc.edu.cn

Project supported by the Research Fund of Low Cost Fabrication of GaN Power Devices and System Integration, China (Grant No. JCYJ20160226192639004), the Research Fund of AlGaN HEMT MEMS Sensor for Work in Extreme Environment, China (Grant No. JCYJ20170412153356899), and the Research Fund of Reliability Mechanism and Circuit Simulation of GaN HEMT, China (Grant No. 2017A050506002).

Abstract

In this work, the field plate termination is studied for Ga2O3 Schottky barrier diodes (SBDs) by simulation. The influence of field plate overlap, dielectric material and thickness on the termination electric field distribution are demonstrated. It is found that the optimal thickness increases with reverse bias increasing for all the three dielectrics of SiO2, Al2O3, and HfO2. As the thickness increases, the maximum electric field intensity decreases in SiO2 and Al2O3, but increases in HfO2. Furthermore, it is found that SiO2 and HfO2 are suitable for the 600 V rate Ga2O3 SBD, and Al2O3 is suitable for both 600 V and 1200 V rate Ga2O3 SBD. In addition, the comparison of Ga2O3 SBDs between the SiC and GaN counterpart reveals that for Ga2O3, the breakdown voltage bottleneck is the dielectric. While, for SiC and GaN, the bottleneck is mainly the semiconductor itself.

1. Introduction

In the past decades, wide band gap semiconductor devices, such as SiC and GaN, have been studied extensively for their superior properties. Nowadays, Ga2O3, with a wider band gap (4.8 eV), receives more and more attention.[1,2] The Baliga merit of Ga2O3 is about 10 times that of 4H–SiC and four times that of GaN,[3] suggesting that it is more suitable for high-power and high-voltage applications.

As a new semiconductor device, the Ga2O3 device is just beginning to be studied, and mainly in the Schottky barrier diodes (SBDs)[4,5] and MOSFETs.[68] For Ga2O3 SBD, edge termination is necessary because the electric field will concentrate on the edge of Schottky junction. In past studies, several ways of terminating edge have been applied to SiC SBDs, including field plates, guarding rings, mesa structures, high-resistivity layers or the combination of such techniques.[911] Among these methods, the field plate is very simple and widely adopted when the voltage requirement is not that high, such as the case of 1 kV.[12] For Si or SiC, SiO2 is often used as the field plate dielectric, while for Ga2O3 whose dielectric is proper still needs to be investigated.

Recently, Cha et al.[13] studied the field plate structure of Ga2O3 SBD with a p-type drift layer by simulation. They found that the dielectric should possess both high permittivity and a high breakdown field. This work focuses on a Ga2O3 SBD with an n-type drift layer. The field plate is studied to provide improved reverse electric field (E-field) distribution of Ga2O3 SBDs, which is conducive to breakdown performance. The work is based on simulations and three kinds of dielectrics. A series of field plate overlaps and dielectric thicknesses are studied systematically.

2. Simulation details

The simulations are carried out by using Silvaco TCAD software. The parameters of Ga2O3 are cited mainly from Ref. [14]. The Ga2O3 band gap and affinity are 4.8 eV and 4 eV, respectively. The work function of the Schottky electrode is 5 eV. The thickness and electron concentration of the Ga2O3 drift layer are 7 μm and 1×1016 cm–3, respectively. The drift layer thickness is the same as that in Refs. [12] and [15]. Figure 1(a) indicates that the breakdown voltage for such a Ga2O3 drift layer is larger than 4000 V theoretically. The electron concentration of Ga2O3 substrate is 1×1019 cm–3. SiO2, Al2O3, and HfO2 are employed as the field plate dielectric, respectively. The dielectric are the same as that in Ref. [13], their permittivity values are 3.9, 9.3, and 22, respectively. The maximum electric field intensity in Ga2O3 drift layer, at the edge of the Schottky junction (Ej) and the edge of the plate (Ep), as a function of the overlap are studied at different values of dielectric thickness and reverse voltage. In addition, the maximum electric field intensity in the dielectric (Ed), located at the edge of plate, as a function of dielectric thickness is also studied. The reverse bias voltages are 600 V and 1200 V, respectively. The range of field plate overlap is 1 μm–20 μm. The thickness ranges of SiO2, Al2O3, and HfO2 are 200 nm–500 nm, 300 nm–700 nm, and 300 nm–700 nm, respectively. The device structure is shown in Fig. 1(b).

Fig. 1. (color online) (a) Breakdown voltages versus doping concentration for different Ga2O3 thicknesses; (b) schematic diagram of the field plate termination of Ga2O3 SBD.

In addition, 4H–SiC and GaN counterparts with similar device structures are also presented for comparison. The thickness and electron concentration of the drift layer are the same as those of Ga2O3. The field plate overlap is 20 μm. The thickness range of SiO2, Al2O3, and HfO2 are 200 nm–300 nm, 300 nm–600 nm and 300 nm–900 nm, respectively. The SiC band gap and affinity are 3.2 eV and 3.5 eV, and the GaN band gap and affinity are 3.4 eV and 4 eV, respectively. The work function of the Schottky electrode is 4.5 eV for SiC and 5 eV for GaN, thus keeping the same Schottky barrier height as that in the case of Ga2O3.

3. Results and discussion

Figure 2(a) shows the variations of Ej and Ep with overlap and SiO2 thickness at 600 V. It is seen that, as the overlap increases from 1 nm to 5 μm, the Ej decreases quickly and the Ep increases slightly; as it exceeds 5 μm, their variations both become very slow. To save the termination area, 10 μm length overlap is sufficient. When the SiO2 thickness is 200 nm, the Ej and Ep eventually tend to 2.66 MV/cm and 2.9 MV/cm, respectively. In this case, Ej is smaller than Ep. As SiO2 thickness is 250 nm, both of the values tend to 2.81 MV/cm and 2.66 MV/cm, respectively. In this case Ej becomes greater than Ep. As SiO2 thickness further increases, their gap becomes larger. Since the function of the field plate is to share the high termination electric field, it is ideal when Ej and Ep are both small and equal. Therefore, Fig. 1(a) reveals that the optimal SiO2 thickness is between 200 nm and 250 nm at 600 V.

Fig. 2. (color online) Plots of Ej and Ep versus field plate overlap for different SiO2 thicknesses and reverse bias voltage of (a) 600 V and (b) 1200 V.

Figure 2(b) shows the variations of Ej and Ep with the overlap and SiO2 thickness at 1200 V. It is seen that the variation tendency is similar to that in Fig. 2(a), but differences still exist obviously. In Fig. 2(b), when SiO2 thickness is 200 nm, Ej is still smaller than Ep, while their gap becomes larger than that in Fig. 2(a). When SiO2 thickness is 250 nm, Ep begins to be greater than Ej, which is diametrically opposite to that in Fig. 2(a). When SiO2 thickness further increases, Ej is still greater than Ep, while their gap becomes smaller. Figure 2(b) reveals that the variations of Ej and Ep at 1200 V are different from those in the case of 600 V, and the optimal SiO2 thickness will be a little greater than 250 nm at 1200 V.

Figure 3(a) shows the variations of Ej and Ep with the overlap and Al2O3 thickness at 600 V. It shows that when Al2O3 thickness is 300 nm, Ej is much smaller than Ep; as it increases to 400 nm, both of the values become almost equal; as it further increases to 500 nm, Ej begins to be larger than Ep. Figure 3(a) reveals that the optimal Al2O3 thickness is about 400 nm at 600 V.

Fig. 3. (color online) Plots of Ej and Ep versus field plate overlap for different Al2O3 thickness. The reverse bias is (a) 600 V and (b) 1200 V.

Figure 3(b) shows the variations of Ej and Ep with the overlap and Al2O3 thickness at 1200 V. Comparing with Fig. 3(a), it is seen that in the condition of 400 nm Al2O3 thickness, the gap between Ej and Ep turns larger, signifying that it is no longer the optimal thickness in this case. Moreover, the gap between Ej and Ep becomes very close when Al2O3 thickness is 500 nm. So it can be deduced that the optimal Al2O3 thickness is a little less than 500 nm at 1200 V.

Figure 4(a) shows the variations of Ej and Ep with the overlap and HfO2 thickness at 600 V. It is seen that Ej is much smaller than Ep when HfO2 thickness is 500 nm; as it increases to 600 nm, their gap becomes smaller; as it increases to 650 nm, Ej is just a little smaller than Ep; as it further increases to 700 nm, Ej starts to be greater than Ep. Figure 4(a) reveals that the optimal HfO2 thickness is between 650 nm and 700 nm at 600 V.

Fig. 4. (color online) Plots of Ej and Ep versus field plate overlap for different HfO2 thickness. The reverse bias voltage is (a) 600 V and (b) 1200 V.

Figure 4(b) shows the variations of Ej and Ep with the overlap and HfO2 thickness at 1200 V. Comparing with Fig. 4(a), it is seen that Ej is still smaller than Ep, their gap even becomes larger when HfO2 thickness is 500 nm. The scenario for HfO2 with 600 nm in thickness is similar to that with 650 nm in thickness. When HfO2 thickness is 700 nm, Ej is just a little smaller than Ep. It means that at 1200 V, the optimal HfO2 thickness is greater than 700 nm.

Figures 24 reveal that there is an optimal dielectric thickness for a certain circumstance. This is because, on the one hand, the dielectric thickness should be sufficiently thick so that the peak electric field inside the dielectric will not exceed its breakdown intensity; on the other hand, it should also be sufficiently thin so that the field plate can influence the electric field distribution inside the semiconductor and provide sufficient field relief at the edge.[9] It is also observed that, as the reverse bias increases, the optimal thickness increases accordingly in the condition of same dielectric, which is very significant for designing the different voltage rate SBDs. Moreover, as the relative permittivity increases from 3.9 (SiO2) to 22 (HfO2), the optimal dielectric thickness also increases accordingly. In addition, the Ej and Ep both become smaller at optimal dielectric thickness as the permittivity increases. For example, at 600 V, the lowest values of Ej and Ep are both about 2.8 MV/cm for SiO2, and decrease to about 2.4 MV/cm for Al2O3 and about 2.1 MV/cm for HfO2, indicating that the higher the permittivity, the lower the breakdown risk of Ga2O3 will be.

Figure 5 shows the relationships between Ed and dielectric thickness for different dielectrics. In Fig. 5(a), it is seen that Ed decreases with SiO2 thickness increasing. While it is worth noting that it is more than 8 MV/cm and 14 MV/cm at 600 V and 1200 V, respectively. Figure 2 shows that the maximum electric field in Ga2O3 is about 2.8 MV/cm and 4.5 MV/cm at 600 V and 1200 V, respectively. Since the critical breakdown electric field (Ec) of Ga2O3 and SiO2 are 8 MV/cm[3] and 15 MV/cm, respectively,[16] suggesting that that SiO2 is suitable to be the dielectric for the 600-V Ga2O3 SBDs but not quit suitable for the 1200 V ones.

Fig. 5. (color online) Variations of Ed with dielectric thickness, the dielectric is (a) SiO2, (b) Al2O3, and (c) HfO2.

Figure 5(b) shows that Ed also decreases with Al2O3 thickness increasing, while its decreasing rate is less than the decreasing rate of Ed that decreases as SiO2 thickness increases. The value of Ed is about 5 MV/cm and 8.5 MV/cm at 600 V and 1200 V, respectively. The corresponding maximum electric field in Ga2O3 is about 2.4 MV/cm (Fig. 3(a)) and 3.8 MV/cm (Fig. 3(b)) at 600 V and 1200 V, respectively. Since the Ec of Al2O3 is 11.2 MV/cm–13.8 MV/cm,[16] implying that Al2O3 is suitable for both the 600 V and 1200 V Ga2O3 SBDs.

Figure 5(c) shows that Ed increases slightly as HfO2 thickness increases, which is different from the scenarios of SiO2 and Al2O3. The Ed is about 3 MV/cm and 5.2 MV/cm at 600 V and 1200 V, respectively. The corresponding maximum electric field in Ga2O3 is about 2.1 MV/cm (Fig. 4(a)) and 3.3 MV/cm (Fig. 4(b)) at 600 V and 1200 V, respectively. Since the Ec of HfO2 is 3.9 MV/cm–6.7 MV/cm,[16] suggesting that it is only suitable for the 600-V Ga2O3 SBDs. All the results indicate that the dielectric permittivity of HfO2 should approach that of Ga2O3. Reference [13] indicates that for Ga2O3 SBD with p-type Ga2O3 drift layer the best choice is HfO2, thus our results also show that the scenario of the n-type drift layer is different that of the p-type one.

Figure 6 shows the variations of Ej and Ep with dielectric thickness for SiC SBDs at 600 V and 1200 V. Figure 6(a) shows that at 600 V, the optimal SiO2 thickness is between 200 nm and 250 nm, the optimal Al2O3 thickness is between 400 nm and 500 nm, and the optimal HfO2 thickness is about 800 nm. Figure 6(b) shows that at 1200 V, the optimal SiO2 thickness is a little larger than 250 nm, the optimal Al2O3 thickness is a little larger than 500 nm and the optimal HfO2 thickness is about 900 nm. Comparing with the results of Ga2O3, it is known that at the same reverse bias, the optimal dielectric thickness is larger for the SiC counterpart. Moreover, since the Ec of SiC is 2.5 MV/cm, suggesting that such a SiC drift layer is not able to withstand 1200 V (Fig. 6(b)). In Fig. 6(a), it can also be observed that such a drift layer is not suitable for the 600 V SiC SBD when the dielectric is SiO2 or Al2O3.

Fig. 6. (color online) Plots of Ej and Ep versus dielectric thickness for SiC SBDs. The reverse bias voltage is: (a) 600 V and (b) 1200 V.

Figure 7 shows the variations of Ej and Ep with dielectric thickness for the GaN SBDs at 600 V and 1200 V. Figure 7(a) is for the case with a reverse bias of 600 V, an optimal SiO2 thickness of nearly 250 nm, an optimal Al2O3 thickness is nearly 500 nm, and an optimal HfO2 thickness larger than 800 nm. Figure 7(b) is for the case with a reverse bias of 1200 V, the optimal SiO2 thickness is larger than 250 nm, the optimal Al2O3 thickness is larger than 500 nm, and the optimal HfO2 thickness is a little larger than 900 nm. Comparing with the results of Ga2O3 and SiC, it is known that at the same reverse bias, the optimal dielectric thickness for the GaN SBDs is even larger than the SiC ones. Furthermore, since the Ec of GaN is 3.4 MV/cm,[17] suggesting that such a GaN drift layer is also not able to withstand 1200 V (Fig. 7(b)).

Fig. 7. (color online) Plots of Ej and Ep with dielectric thickness for GaN SBDs. The reverse bias voltage is (a) 600 V and (b) 1200 V.

Figure 8 shows the relationship between Ed and dielectric thickness for the SiC and GaN SBDs. It is seen that at the same dielectric thickness and reverse bias, the Ed of GaN SBDs is less than that of the SiC ones. Combing Figs. 7(a), 8(a), and 8(b), it is clear that such a GaN drift layer is able to withstand 600 V when the dielectric is SiO2 or Al2O3. It is because in such cases the maximum electric field does not exceed the Ec of GaN nor SiO2 nor Al2O3 at 600 V. Figure 8(c) shows that the electric field intensity is about 5 MV/cm in HfO2 at 600 V. Since the Ec of HfO2 is 3.9 MV/cm–6.7 MV/cm, suggesting that it is not suitable for the 600-V rate SiC nor GaN power devices. Moreover, figure 6(a) already indicates that the drift layer is not appropriate for the 600 V SiC SBDs either, when the dielectric is SiO2 or Al2O3, implying that the drift layer is not suitable for the 600-V rate SiC devices either, no matter which dielectric is chosen.

Fig. 8. (color online) Plots of Ed as versus dielectric thickness for SiC and GaN SBDs, the dielectric is (a) SiO2, (b) Al2O3, (c) HfO2.

Comparing Ga2O3 SBDs with the SiC and GaN counterparts, it is known that for Ga2O3 devices, the breakdown voltage bottleneck is the dielectric. While for SiC and GaN devices, the bottleneck is mainly the semiconductor itself.

4. Conclusions

In this work, the field plate termination in a Ga2O3 Schottky barrier diode (SBD) is studied systematically. Three kinds of oxides, i.e., SiO2, Al2O3, and HfO2, are used as the field plate dielectric, respectively. The influence of field plate overlap and dielectric thickness on the maximum electric field in Ga2O3 as well as the dielectric are studied. The results reveal that the optimal dielectric thickness increases with reverse bias increasing. The maximum electric field intensity decreases in SiO2 and Al2O3, while it increases in HfO2 with the increase of the dielectric thickness. Moreover, the results also indicates that SiO2 and HfO2 are suitable for the 600-V rate Ga2O3 SBD, the Al2O3 is suitable for both 600-V and 1200-V level Ga2O3 SBD. Furthermore, the comparison among Ga2O3, SiC and GaN devices reveals that, for the Ga2O3 device, the breakdown voltage bottleneck is the dielectric. While, for SiC and GaN devices, the bottleneck is mainly the semiconductor itself.

Reference
[1] He Q Mu W Fu B Jia Z Long S Yu Z Yao Z Wang W Dong H Qin Y 2018 IEEE Electron Dev. Lett.
[2] Liu X Yue C Xia C Zhang W 2016 Chin. Phys. 25 017201
[3] Higashiwaki M Sasaki K Murakami H Kumagai Y Koukitu A Kuramata A Masui T Yamakoshi S 2016 Semicond. Sci. Technol. 31 034001
[4] Yang J Ahn S Ren F Pearton S Jang S Kim J Kuramata A 2017 Appl. Phys. Lett. 110 192101
[5] He Q Mu W Dong H Long S Jia Z Lv H Liu Q Tang M Tao X Liu M 2017 Appl. Phys. Lett. 110 093503
[6] Chabak K D McCandless J P Moser N A Green A J Mahalingam K Crespo A Hendricks N Howe B M Tetlak S E Leedy K 2018 IEEE Electron Dev. Lett. 39 67
[7] Moser N A McCandless J P Crespo A Leedy K D Green A J Heller E R Chabak K D Peixoto N Jessen G H 2017 Appl. Phys. Lett. 110 143505
[8] Moser N McCandless J Crespo A Leedy K Green A Neal A Mou S Ahmadi E Speck J Chabak K 2017 IEEE Electron Dev. Lett. 38 775
[9] Tarplee M C Madangarli V P Zhang Q Sudarshan T S 2001 IEEE Trans. Electron Dev. 48 2659
[10] Kumta A S Jinghua X 2009 IEEE Trans. Electron Dev. 56 2925
[11] Chen F P Zhang Y M Zhang Y M Tang X Y Wang Y H Chen W H 2011 Chin. Phys. 20 117301
[12] Konishi K Goto K Murakami H Kumagai Y Kuramata A Yamakoshi S Higashiwaki M 2017 Appl. Phys. Lett. 110 103506
[13] Choi J H Cho C H Cha H Y 2018 Results Phys. 9 1170
[14] Silvaco 2013 https://www.silvaco.com/tech_lib_TCAD/simulationstandard/2013/
[15] Sasaki K Wakimoto D Thieu Q T Koishikawa Y Kuramata A Higashiwaki M Yamakoshi S 2017 IEEE Electron Dev. Lett. 38 783
[16] McPherson J Kim J Shanware A Mogul H 2003 Appl. Phys. Lett. 82 2121
[17] Duan X L Zhang J C Xiao M Zhao Y Ning J Hao Y 2016 Chin. Phys. 25 087304