A New Split Gate Resurf Stepped Oxide UMOSFET Structure with High Doped Epitaxial Layer for Improving Figure of Merit (FOM)

Featured Application: The split gate resurf stepped oxide with highly doped epitaxial layer (HDSGRSO) UMOSFET has extremely low on-state resistance, therefore, it can be used in synchronous rectification circuits to replace rectifier diodes to reduce rectification losses. It can greatly improve the efficiency of the DC / DC converter and there is no deadtime voltage caused by the Schottky barrier voltage. On the other hand, it can be applied to the high frequency switching due to its excellent dynamic characteristics. Abstract: The split gate resurf stepped oxide with highly doped epitaxial layer (HDSGRSO) UMOSFET has been proposed. The epitaxial layer of HDSGRSO u-shape metal oxide semiconductor ﬁeld e ﬀ ect transistor (UMOSFET) has been divided into three parts: the upper epitaxial layer, the lower epitaxial layer and the middle epitaxial layer with higher doping concentration. The research shows that the reduced SURface ﬁeld (RESURF) active has been enhanced due to the high doped epitaxial layer, which can modulate the electric ﬁeld distribution and reduce the internal high electric ﬁeld. Therefore, the HDGRSO UMOSFET has a higher breakdown voltage (BV), a lower on-state speciﬁc resistance ( R SP ) and a better ﬁgure of merit (FOM). According to the results of Technology Computer Aided Design (TCAD) simulations, the FOM (BV 2 / R SP ) of HDSGRSO UMOSFET has been improved by 464%, and FOM ( R SP × Q gd ) of HDSGRSO UMOSFET has been reduced by 27.9% compared to the conventional structure, respectively, when the BV is 240 V. Furthermore, there is no extra special process required in this advanced fabrication procedure, which is relatively cost-e ﬀ ective and achievable. writing—review L.W.; funding acquisition, L.W. All authors


Introduction
The power MOSFET (metal oxide semiconductor field effect transistor) has been widely used in the analog and digital circuits. It is one of the most vital research fields to reduce the on-state specific resistance (R SP ) for a certain breakdown voltage (BV). However, the R SP in conventional MOSFET (MOSFET) is limited by the 1D Silicon limit where the R SP and the BV are oppositely affected by the doping concentration. In order to improve the tradeoff between R SP and BV, the two-dimensional charge coupling method has been proposed [1]. There are two kinds of power MOSFET structures, one is the power Superjunction (SJ) MOSFET [2][3][4]. The SJ MOSFET can decrease the R SP , however,

Device Structure and Mechanism
In order to achieve the desired charge-coupling, the split-gate electrode within the oxide coated trenches is designed to connect the source during operation in the blocking mode, as shown in Figure 1. The cross-sectional view of the HDSGRSO UMOSFET is presented in Figure 1b. Conventionally, the doping concentration of the n-drift is uniform. In this study, the epitaxial layer of HDGRSO UMOSFET has been divided into three parts with different doping concentration and thickness: the upper, the middle and lower epitaxial layers. The middle epitaxial layer has a higher doping concentration compared with the upper and lower epitaxial. In order to emphasize the modulation effect of the high doped epitaxial layer on the electric field, the doping concentrations of the upper and lower epitaxial layers are set to same, and the thickness of the high doped epitaxial layer is L 2 .
In this work, we used the TCAD to compare the performance of two structures. The potential contours and the electric field distribution along the trench surface (cut2) and in the middle of Mesa active area (cut1) for two structures at V DS = 240 V are represented as shown in the Figure 2. The doping concentration of the n-drift (N D and N 2 ) was 1 × 10 15 /cm −3 and 5 × 10 15 /cm −3 for SGRSO and HDSGRSO UMOSFET, respectively, and the doping concentration of the high doped epitaxial layer was 1.5 × 10 16 /cm −3 . It can be seen from the picture that the potential lines originally concentrated on the lower part moves to the upper part of the n-drift region, and the electric field of HDSGRSO UMOSFET distribution was more uniform than that of SGRSO UMOSFET. Furthermore, a small electric field peak appeared at the interface between the upper and the middle epitaxial layers, as shown in Figure 2, at −13.5 µm in the y-axis. Therefore, the high doped epitaxial layer can modulate the electric field distribution and reduce the internal high electric field to improve the BV. the doping concentration of the n-drift is uniform. In this study, the epitaxial layer of HDGRSO UMOSFET has been divided into three parts with different doping concentration and thickness: the upper, the middle and lower epitaxial layers. The middle epitaxial layer has a higher doping concentration compared with the upper and lower epitaxial. In order to emphasize the modulation effect of the high doped epitaxial layer on the electric field, the doping concentrations of the upper and lower epitaxial layers are set to same, and the thickness of the high doped epitaxial layer is L2.
In this work, we used the TCAD to compare the performance of two structures. The potential contours and the electric field distribution along the trench surface (cut2) and in the middle of Mesa active area (cut1) for two structures at VDS = 240 V are represented as shown in the Figure 2. The doping concentration of the n-drift (ND and N2) was 1  10 15 /cm −3 and 5  10 15 /cm −3 for SGRSO and HDSGRSO UMOSFET, respectively, and the doping concentration of the high doped epitaxial layer was 1.5  10 16 /cm −3 . It can be seen from the picture that the potential lines originally concentrated on the lower part moves to the upper part of the n-drift region, and the electric field of HDSGRSO UMOSFET distribution was more uniform than that of SGRSO UMOSFET. Furthermore, a small electric field peak appeared at the interface between the upper and the middle epitaxial layers, as shown in Figure 2, at -13.5 μm in the y-axis. Therefore, the high doped epitaxial layer can modulate the electric field distribution and reduce the internal high electric field to improve the BV.  Figure 3 shows the three-dimensional view of the electric field distributions for two structures with the same n-drift doping concentration (ND and N2) 5  10 15 /cm −3 . The small electric field peak (at point A) brought by the high doped epitaxial layer is shown in the Figure 3b. It is obvious that the electric field in the drift region of HDSGRSO UMOSFET is more uniform than that of the SGRSO UMOSFET at the same n-drift doping concentration, and the volume under the electric field curved surface of HDSGRSO UMOSFET is bigger contrast to the SGRSO UMOSFET. As a result, the BV of HDSGRSO UMOSFET is 240 V, while that of SGRSO UMOSFET is 190 V.  Figure 3 shows the three-dimensional view of the electric field distributions for two structures with the same n-drift doping concentration (N D and N 2 ) 5 × 10 15 /cm −3 . The small electric field peak (at point A) brought by the high doped epitaxial layer is shown in the Figure 3b. It is obvious that the electric field in the drift region of HDSGRSO UMOSFET is more uniform than that of the SGRSO UMOSFET at the same n-drift doping concentration, and the volume under the electric field curved surface of HDSGRSO UMOSFET is bigger contrast to the SGRSO UMOSFET. As a result, the BV of HDSGRSO UMOSFET is 240 V, while that of SGRSO UMOSFET is 190 V.   Figure 4 exhibits the fabrication process and structure of the HDSGRSO UMOSFET. The fabrication process of epitaxial layer is the only difference compared to traditional processes. First, a 3.2 μm n-type epitaxial layer is formed on the N + substrate. Then, the middle and upper n-type epitaxial layer are formed on the first epitaxial in the same way. The doping concentration of the bottom and the upper epitaxial layers is the same, but the middle epitaxial layer has a higher doping concentration (a). Secondly, dry etching can be used to form the deep trench. An oxide layer is formed by thermally oxidized on the trench surface and then etched off to smooth the trench corner (b). Next, a 1.2 μm oxide is deposited to form the field oxide (c). Polysilicon is then deposited in the trench to form the source electrode which is surrounded by the field oxide (d). Afterwards, a thick oxide is deposited and etched back to separate the gate and the source electrode (e). Following, the gate oxide is formed by thermally oxidizing for 25 min under 1050 degrees dry oxygen conditions, and the polysilicon is deposited again to obtain the gate electrode (f). Subsequently, the p-body and n + source region are formed by implanted ions, respectively (g). Finally, a thick oxide is deposited on the gate surface to separate the gate and source metallic electrodes, and then the fabrication of the device is finished (h).  Figure 4 exhibits the fabrication process and structure of the HDSGRSO UMOSFET. The fabrication process of epitaxial layer is the only difference compared to traditional processes. First, a 3.2 µm n-type epitaxial layer is formed on the N + substrate. Then, the middle and upper n-type epitaxial layer are formed on the first epitaxial in the same way. The doping concentration of the bottom and the upper epitaxial layers is the same, but the middle epitaxial layer has a higher doping concentration (a). Secondly, dry etching can be used to form the deep trench. An oxide layer is formed by thermally oxidized on the trench surface and then etched off to smooth the trench corner (b). Next, a 1.2 µm oxide is deposited to form the field oxide (c). Polysilicon is then deposited in the trench to form the source electrode which is surrounded by the field oxide (d). Afterwards, a thick oxide is deposited and etched back to separate the gate and the source electrode (e). Following, the gate oxide is formed by thermally oxidizing for 25 min under 1050 degrees dry oxygen conditions, and the polysilicon is deposited again to obtain the gate electrode (f). Subsequently, the p-body and n + source region are formed by implanted ions, respectively (g). Finally, a thick oxide is deposited on the gate surface to separate the gate and source metallic electrodes, and then the fabrication of the device is finished (h).

Fabrication Procedure
Appl. Sci. 2020, 10, x FOR PEER REVIEW 5 of 11  Figure 5 shows the characteristics of the SGRSO and HDSGRSO UMOSFETs for same bias condition; the structural parameters adopted in the simulations are shown in the Table 1 Figure 5 shows the characteristics of the SGRSO and HDSGRSO UMOSFETs for same bias condition; the structural parameters adopted in the simulations are shown in the Table 1. The breakdown voltage for the two structures is shown in the Figure 5a, and the limit of drain current density is 1 µA/µm 2 . It is obvious that the breakdown voltage of both is about 240 V. The relationship between the R SP and gate voltage is shown in the Figure 5b for the two structures at V DS = 0.3 V. It can be found that the R SP of the HDSGRSO UMOSFET is much smaller than that of the SGRSO UMOSFET, and the gap between them increases with the increase of the gate voltage. The R SP of the HDSGRSO UMOSFET can be reduced by 76%~79%, in contrast to the conventional structure at the same gate voltage. Table 1. Structural parameters adopted in the simulations.

Parameter
Value The drain I-V characteristics of the SGRSO and HDSGRSO UMOSFETs with same bias condition are shown in Figure 5c when the breakdown voltage is 240 V. It is obvious that the HDSGRSO UMOSFET has a relatively large current than that of SGRSO UMOSFET when the V GS is 2 V, 5 V, 10 V and 15 V, respectively. Moreover, the current of the HDSGRSO UMOSFET when the V GS is 5 V even exceeds the current of the SGRSO UMOSFET at 15 V especially. The transconductance of HDSGRSO and SGRSO UMOSFETs is shown in the Figure 5d. It is similar to the output characteristics of the two structures that the transconductance of HDSGRSO UMOS at the V GS of 0.5 V is bigger than that of SGRSO UMOSFET when the V GS is 1.5 V. The transconductance of HDSGRSO UMOSFET increases by approximately 178% contrast to the SGRSO UMOSFET, when the V GS is 1.5 V.
An extensive analysis of the HDSGRSO UMOSFET has been given in contrast with SGRSO UMOSFET with the same structure parameters. The model of bandgap narrowing, concentration dependent mobility model, parallel electric field-dependent mobility model, Shockley-Read-Hall (SRH) model and impact-ionization model have been adopted in the simulation work [16]. Figure 6a shows the BV and FOM (BV 2 /R SP ) at different doping concentrations of the n-drift region. Table 1 shows the parameters we adopted in the simulations. The BV of the two structures increases with the decreasing of the doping concentration of n-drift within the range of the 1 × 10 15 /cm −3 −6 × 10 15 /cm −3 . When the doping concentration of the drift region is 5 × 10 15 /cm −3 , the FOM of HDSGRSO UMOSFET achieves the optimal value, which improved by 106% contrasted with the SGRSO UMOSFET, and the BV of HDSGRSO and SGRSO UMOSFETs were 239 V and 191 V respectively, which improved by 25.1%. It is worth noting that when the BV of the two structures was about 240 V, the FOM of HDSGRSO UMOSFET improved by 402% in contrast to the SGRSO UMOSFET as the doping concentration of the drift region is 5 × 10 15 /cm −3 and 1 × 10 15 /cm −3 for HDSGRSO and SGRSO UMOSFETs. The BV and FOM (BV 2 /R SP ) at different widths of the doped layer L 2 are shown in Figure 6b. It gets the optimal FOM and BV when the L 2 is 4 µm. The BV and FOM (BV 2 /R SP ) at different thickness of the upper n-drift (L 3 ) is shown in Figure 6c when the L 2 is 4 µm. It achieves the highest FOM and BV when L 3 is 3.2 µm. Figure 6d shows the BV and FOM (BV 2 /R SP ) at different doping concentrations of the high doped epitaxial layer when the n-drift region is 5 × 10 15 /cm −3 . It is obvious the BV of the HDSGRSO UMOSFET is gradually increasing with the increase of the doping concentration of the high doped epitaxial layer. It gets optimal FOM and BV when the doping concentration of the high doped epitaxial layer is 1.5 × 10 16 /cm −3 , because the modulation effect of the electric field in the n-drift gradually increases and reaches the optimal level with the increasing of the doping concentration of high doped epitaxial layer. However, when the doping concentration of the middle epitaxial layer is higher than 1.5 × 10 16 /cm −3 , the charge balance in the n-drift will be broken and the BV will drop steeply. Figure 7 shows the BV and FOM (BV 2 /R SP ) at different thicknesses of the drift region. It indicates that there exists a saturated BV at a certain thickness. When the L is 14 µm, the FOM value and BV reaches the optimal value, respectively, and the BV is about 240 V. With the increase of epitaxial thickness, the R SP gradually increases, resulting in the decrease of FOM. The L is 14 µm for HDSGRSO UMOSFET, whereas the L' is 16 µm for SGRSO UMOSFET when the BV is 240 V. Because of the enhancing of the electric field by high doped epitaxial layer, the breakdown occurs before the n-drift is depleted completely. Therefore, it can reduce the thickness of the epitaxial layer and R SP . As a result, the FOM has been increased by 464% as compared with the SGRSO UMOSFET when the BV is about 240 V, according to the simulation results. The drain I-V characteristics of the SGRSO and HDSGRSO UMOSFETs with same bias condition are shown in Figure 5c when the breakdown voltage is 240 V. It is obvious that the HDSGRSO UMOSFET has a relatively large current than that of SGRSO UMOSFET when the VGS is 2 V, 5 V, 10 V and 15 V, respectively. Moreover, the current of the HDSGRSO UMOSFET when the VGS is 5 V even exceeds the current of the SGRSO UMOSFET at 15 V especially. The transconductance of HDSGRSO and SGRSO UMOSFETs is shown in the Figure 5d. It is similar to the output characteristics of the two structures that the transconductance of HDSGRSO UMOS at the VGS of 0.5 V is bigger than that of SGRSO UMOSFET when the VGS is 1.5 V. The transconductance of HDSGRSO UMOSFET increases by approximately 178% contrast to the SGRSO UMOSFET, when the VGS is 1.5 V. An extensive analysis of the HDSGRSO UMOSFET has been given in contrast with SGRSO UMOSFET with the same structure parameters. The model of bandgap narrowing, concentration dependent mobility model, parallel electric field-dependent mobility model, Shockley-Read-Hall (SRH) model and impact-ionization model have been adopted in the simulation work [16]. Figure 6a shows the BV and FOM (BV 2 /RSP) at different doping concentrations of the n-drift doping concentration of the high doped epitaxial layer is 1.5  10 16 /cm −3 , because the modulation effect of the electric field in the n-drift gradually increases and reaches the optimal level with the increasing of the doping concentration of high doped epitaxial layer. However, when the doping concentration of the middle epitaxial layer is higher than 1.5  10 16 /cm −3 , the charge balance in the ndrift will be broken and the BV will drop steeply.  Figure 7 shows the BV and FOM (BV 2 /RSP) at different thicknesses of the drift region. It indicates that there exists a saturated BV at a certain thickness. When the L is 14 μm, the FOM value and BV reaches the optimal value, respectively, and the BV is about 240 V. With the increase of epitaxial thickness, the RSP gradually increases, resulting in the decrease of FOM. The L is 14 μm for HDSGRSO UMOSFET, whereas the L' is 16 μm for SGRSO UMOSFET when the BV is 240 V. Because of the enhancing of the electric field by high doped epitaxial layer, the breakdown occurs before the n-drift is depleted completely. Therefore, it can reduce the thickness of the epitaxial layer and RSP. As a result, the FOM has been increased by 464% as compared with the SGRSO UMOSFET when the BV is about 240 V, according to the simulation results. For the power MOSFET used in the high frequency field, the switching losses is the most important part of the total power loss. The time period associated with the increase of the drain current and decrease of the drain voltage is determined by the charging of the device capacitances. The Figure 8a shows the input capacitances (Ciss) and reverse transfer capacitances (Crss) of both structures when VGS = 0 V and f = 1 MHz. The Crss of the HDSGRSO UMOSFET has a slight increase in comparison with the ordinary structure because of the highly doped epitaxial layer. However, the capacitances difference between the two structures is gradually reduced as the drain voltage increases. When the drain voltage exceeds 120 V, the Crss of the two structures are almost same. The Crss of HDSGRSO and SGRSO UMOSFETs is 2 pF and 0.8 pF, respectively, at VDS = 140 V. On the other hand, the doping concentration of the p-body is slightly reduced under the same boron dose due to the highly doped epitaxial layer, which causes a slight reduction of the Ciss for HDSGRSO UNOSFET. It can be seen that the HDSGRSO UMOSFET has lower Ciss contrast to the SGRSO UNOSFET when the drain voltage increases from 0 to 140 V. The Ciss of HDSGRSO and SGRSO UMOSFETs is 1400 pF and 1490 pF, respectively, at VDS = 140 V, which is of benefit to improve the switching speed of the For the power MOSFET used in the high frequency field, the switching losses is the most important part of the total power loss. The time period associated with the increase of the drain current and decrease of the drain voltage is determined by the charging of the device capacitances. The Figure 8a shows the input capacitances (C iss ) and reverse transfer capacitances (C rss ) of both structures when V GS = 0 V and f = 1 MHz. The C rss of the HDSGRSO UMOSFET has a slight increase in comparison with the ordinary structure because of the highly doped epitaxial layer. However, the capacitances difference between the two structures is gradually reduced as the drain voltage increases. When the drain voltage exceeds 120 V, the C rss of the two structures are almost same. The C rss of HDSGRSO and SGRSO UMOSFETs is 2 pF and 0.8 pF, respectively, at V DS = 140 V. On the other hand, the doping concentration of the p-body is slightly reduced under the same boron dose due to the highly doped Appl. Sci. 2020, 10, 7895 8 of 10 epitaxial layer, which causes a slight reduction of the C iss for HDSGRSO UNOSFET. It can be seen that the HDSGRSO UMOSFET has lower C iss contrast to the SGRSO UNOSFET when the drain voltage increases from 0 to 140 V. The C iss of HDSGRSO and SGRSO UMOSFETs is 1400 pF and 1490 pF, respectively, at V DS = 140 V, which is of benefit to improve the switching speed of the HDSGRSO UMOSFET. The FOMs of R SP × Q gd have been adopted to compare the dynamic characteristics of two structures. The FOM (R SP × Q gd ) of HDSGRSO and SGRSO UMOSFETs is 180 and 250 mΩ·nC, as shown in the Figure 8b. The FOM of devices with high doped epitaxial layer structure was reduced by 27.9% in contrast to SGRSO UMOSFET because the introduction of the high doped epitaxial layer increases the doping concentration in the drift region, which makes the R SP decrease greatly.
in comparison with the ordinary structure because of the highly doped epitaxial layer. However, the capacitances difference between the two structures is gradually reduced as the drain voltage increases. When the drain voltage exceeds 120 V, the Crss of the two structures are almost same. The Crss of HDSGRSO and SGRSO UMOSFETs is 2 pF and 0.8 pF, respectively, at VDS = 140 V. On the other hand, the doping concentration of the p-body is slightly reduced under the same boron dose due to the highly doped epitaxial layer, which causes a slight reduction of the Ciss for HDSGRSO UNOSFET. It can be seen that the HDSGRSO UMOSFET has lower Ciss contrast to the SGRSO UNOSFET when the drain voltage increases from 0 to 140 V. The Ciss of HDSGRSO and SGRSO UMOSFETs is 1400 pF and 1490 pF, respectively, at VDS = 140 V, which is of benefit to improve the switching speed of the HDSGRSO UMOSFET. The FOMs of RSP × Qgd have been adopted to compare the dynamic characteristics of two structures. The FOM (RSP × Qgd) of HDSGRSO and SGRSO UMOSFETs is 180 and 250 mΩ•nC, as shown in the Figure 8b. The FOM of devices with high doped epitaxial layer structure was reduced by 27.9% in contrast to SGRSO UMOSFET because the introduction of the high doped epitaxial layer increases the doping concentration in the drift region, which makes the RSP decrease greatly.   Table 2 shows the switching time and switching energy loss of two structures at 120 V, 4 A. The HDSGRSO UMSOFET has a lower turn-on delay time (t d(on) ) because of the smaller input capacitance, however, the rise time(t r ) of the HDSGRS UMOSFET is slightly higher due to the larger C rss , which causes results in the turn-on loss (E on ) to be very close. On the other hand, the lager C rss causes a certain increase of the turn-off delay (t d(off) ) time and fall time (t f ), which leads to an increase of turn-off loss (E off ) in contrast to the SGRSO UMOSFET, and this needs more research in the future. The performance of the HDSGRSO UMOSFET has been compared with the one-dimensional silicon limit and other works in Figure 9. The HDSGRSO UMOSFET has the lowest R SP and the highest BV compared with the previous power UMOSFETs.
Appl. Sci. 2020, 10, x FOR PEER REVIEW 9 of 11 Table 2 shows the switching time and switching energy loss of two structures at 120 V, 4 A. The HDSGRSO UMSOFET has a lower turn-on delay time (td(on)) because of the smaller input capacitance, however, the rise time(tr) of the HDSGRS UMOSFET is slightly higher due to the larger Crss, which causes results in the turn-on loss (Eon) to be very close. On the other hand, the lager Crss causes a certain increase of the turn-off delay (td(off)) time and fall time (tf), which leads to an increase of turnoff loss (Eoff) in contrast to the SGRSO UMOSFET, and this needs more research in the future. The performance of the HDSGRSO UMOSFET has been compared with the one-dimensional silicon limit and other works in Figure 9. The HDSGRSO UMOSFET has the lowest RSP and the highest BV compared with the previous power UMOSFETs.  Figure 9. Performance comparing of one-dimensional silicon limit, two-dimensional charge-coupling limits for the pitches of 2 and 8 μm. Some references are [10,24,25] and this work.

Conclusion
In this article, a split-gate resurf stepped oxide with high doped epitaxial layer (HDSGRSO) UMOSFET has been proposed. The epitaxial layer of HDGRSO UMOSFET has been divided into three parts: the upper epitaxial layer, the lower epitaxial layer and the middle epitaxial layer with Figure 9. Performance comparing of one-dimensional silicon limit, two-dimensional charge-coupling limits for the pitches of 2 and 8 µm. Some references are [10,24,25] and this work.

Conclusions
In this article, a split-gate resurf stepped oxide with high doped epitaxial layer (HDSGRSO) UMOSFET has been proposed. The epitaxial layer of HDGRSO UMOSFET has been divided into three parts: the upper epitaxial layer, the lower epitaxial layer and the middle epitaxial layer with higher doping concentration. The TCAD simulation results show that the RESURF active has been enhanced due to the high doped epitaxial layer, which can modulate the distribution of the electric field and reduce the internal high electric field. As a result, the BV of HDGRSO UMOSFET has been improved greatly. Meanwhile, the n-drift region has a higher doping concentration due to the high doped epitaxial layer, which can reduce the R SP for a given BV. Therefore, the FOM of the device with high doped epitaxial can be improved significantly. According to the simulation results, the FOM (BV 2 /R SP ) of HDSGRSO UMOSFET has been improved by 464%, and FOM (R SP × Q gd ) of HDSGRSO UMOSFET has been reduced by 27.9% compared to the conventional structure, respectively, when the BV is 240 V. Furthermore, the only difference between this procedure and the traditional procedure is the fabrication of the epitaxial layer, and no extra special process is required in this advanced fabrication procedure, which is relatively cost-effective and achievable.