1. Introduction
Trench power MOSFETs have become a superior device in the medium-to-low voltage power application field. In conventional trench MOSFETs, the gate is isolated from the drain region only by the gate oxide. This results in that trench MOSFETs exhibit large switching losses due to a high gate-to-drain capacitance (C
gd), which limits its application. In order to reduce the device-switching losses, many studies, such as a thick-bottom oxide layer (TBOX) design, W-gated, and RESURF stepped oxide (RSO) MOSFET, were proposed [
1,
2,
3,
4]. All of these structures feather a thick oxide between gate electrodes and drain area, to reduce device C
gd. The RSO structure uses a thicker oxide at the lower portion of the trench, to reduce C
gd, while it applies a thinner one at the upper portion of the trench, to be the gate oxide. Because the stepped gate electrode plays a role as an extended field plate (FP) to modulate the electric field (EF) around it, this structure not only reduces the feedback capacitance but also the R
on, by using a low-resistivity epitaxial layer. Although RSO design can reduce the C
gd, switching losses are still a big issue when a device is used in a high-frequency application. Split-gate trench (SGT) devices overcame that problem by adding a source electrode located between the gate and drain [
5,
6,
7,
8,
9]. There are two parts in the trenches for a split-gate structure: The upper electrode is the gate, and the lower one is connected by a separate contact to the source, to play as a field plate to balance the charge in the n
- drift epitaxy region. This field plate is surrounded with a thick oxide to be a MOS structure that induces a silicon depletion region once the electrode is biased at a more negative potential than the n
- silicon region [
10,
11,
12]. Furthermore, the extended field plate along the drift epitaxy layer shapes the electric field in the drift region that enables the drift depletion area to support a higher drain voltage by using a lower resistivity epitaxy layer to reduce device specific on-resistance [
5,
13]. In addition, the C
gd of an SGT can be reduced significantly because the gate electrodes are shielded from the drain region by these FPs [
10,
14,
15].
Even RSO and SGT power MOSFETs can provide an effective way to reduce device feedback capacitance and R
on simultaneously. The on-state resistance for a device used in a higher voltage system (100 to 200 V) increases sharply, owing to a high-resistivity epitaxial layer. For 20–30 V low-voltage SGT devices, the channel resistance portion is dominant and amounts to over 60%–85% of the total device resistance. However, this channel resistance is reduced to only 30%–20% for 60–70 V middle-voltage-rating devices [
16,
17]. When the device rating voltage reaches 150–200 V, the drift resistance occupies about 90% of the total device resistance [
16,
18]. To achieve a high breakdown voltage (V
BR) design without increasing the R
on too much, a gradient, two-stepped oxide or multiple stepped oxide designs were applied to the trenches and shown to improve device performance effectively [
18,
19,
20,
21]. Since the potential of the field plate (bottom gate) on the oxide around it is different everywhere, that leads to a different depletion strength and electric field between two trenches along the cell depth, [
18,
19,
20,
21] use oxide engineering to improve device performance. On the other hand, double split-gate resurf stepped oxide UMOS can overcome the non-uniform problem [
15]; however, the oxide and poly process in the trenches is too complicated. The abovementioned methods could make the drift region have a more uniform EF distribution to sustain a higher V
BR. However, these structures required multiple depositions and etching steps that complicate the fabrication process. Superjunction structures and wide bandgap SiC material devices are alternative ways to provide high-voltage and low-R
on,sp solutions [
22,
23,
24,
25]. However, the built-in superjunction depletion layer limits the scalability to lower voltages (<500 V) [
3]. In addition, besides cost issues, low channel mobility owing to a high density of SiC/SiO
2 interface traps and undesirable higher turn on voltage of the body diode of a wide bandgap SiC power MOSFET make SiC devices less attractive than Si ones for lower-voltage applications [
26,
27,
28]. Lower-voltage SiC power MOSFETs have not yet been demonstrated [
10]. For a device structure with a rating voltage below 200 V, Si SGT power MOSFET dominates and plays an important role in reducing the device R
on,sp in power applications.
In this study, we proposed a 150 V SGT power MOSFET with multiple EPIs, to improve the device characteristics, and applied the same way to design a 200 V SGT power device. The single-EPI structures are wildly used in the low-voltage (<50 V) SGT power MOSFETs design. When required device rating voltage is up to 50–100 V, single-EPI device makes this scheme suffer a sharply increased R
on. A double-EPI-layers structure was used to improve device R
on characteristics in some studies [
29,
30]. Compared to the single-EPI one, the double-EPIs device has a higher device output current than the single one. This unique merit allows for the possibility of the double-EPIs design to reduce R
on,sp, as well as its power consumption. In this study, we wanted to design and modify the EPI structures rather than the complicated fabrication ways mentioned in [
15,
18,
19,
20,
21], to reduce device R
on and sustain a high V
BR at the same time. When device ranting voltage is designed to over 100 V, we find that the R
on of double-EPIs structure is no longer satisfying our expectations. Therefore, a triple-EPIs structure was applied, to modify the EF distributions between two trenches, instead of only depending on its magnitude supported by the bottom EPI. This design makes us have more flexibilities in designing the bottom EPI with a lower resistivity specification, to achieve a lower R
on,sp device. In double-EPIs design, the bottom EPI layer is used to support the V
BR, and the top one could be used to modify the EF and reduce the R
on. For a triple-EPIs structure, the top and bottom EPI layers play the same roles as those is the double-EPIs device. The middle one is used to lower the R
on,sp if the top and bottom EPI layers can be properly designed. We applied ISE-TCAD to simulate and investigate by analyzing device potential and EF distributions with different epitaxial layers for all devices [
31]. The R
on,sp of a triple-epitaxial-layer structure is much lower than those applied with a single- or double-epitaxial layer based on the same fabrication process.
3. Results and Discussion
First, we constructed a 150 V device by using a double-EPIs structure. The trench depth we used here was 6 µm, from the top to the bottom EPI. For a double-EPIs structure, to improve its V
BR, a thicker thickness or a higher-resistivity bottom EPI is required. However, it will increase R
on significantly. Then, we apply the same EPI thickness and trench depth as double-EPIs structure to all devices in the simulation. For comparison, we adopted the same bottom EPI specification for the double structure used as for the single-EPI device. For a triple-EPIs device, the EPI specifications are adjusted to achieve a balance to have a maximum V
BR and a minimum R
on,sp. The EPI information for all structures is list in
Table 2. All the devices simulated here use the same trench depth (6 µm). Different top- and middle-EPI-thickness designs are studied for a triple-EPIs device.
Figure 4a shows the EF distributions with different top- and middle-EPI-thickness designs. We can see that the EF distributions between two trenches can be modified by different top and middle EPI thickness. Our approach to improving the EF distributions between two trenches is similar to that proposed in [
15]. We used triple EPIs and [
15] double split-gates with different bias in the trenches, to achieve the same purpose. The R
on is not affected by the top EPI too much; however, different electric field distributions with different EPI combinations here give us more room to design a high V
BR device. One can expect that the highest breakdown voltage can be obtained in the largest area of the EF integration, with respect to the cell depth [
19,
22]. In our study, the best top-and middle-EPI-thickness ratio to sustain a high V
BR device is 1:2.
Figure 4b presents the simulated V
BR and R
on,sp with different EPI-thickness designs.
Figure 5 and
Figure 6 show the simulated potential and EF distributions for all structures under the same total EPI thickness. From the simulation, it is obvious that the triple-EPIs device can sustain a higher V
BR easily than the others. A middle EPI layer is used to increase the EF magnitudes and then enhance the breakdown, as well as lower the R
on,sp simultaneously. In addition, it offers us more flexibility to adjust the resistivity of the bottom EPI, to further reduce its R
on,sp. The EF distribution curves in the cell center for all structures are shown in
Figure 7. From this figure, the triple-EPIs design shows it has more uniform EF distributions between two trenches to sustain a higher V
BR. In
Figure 7, it is obvious that the two-layer structure can increase the device top electric field between two trenches; however, it decreases to a low value at p
--well/n
-EPI as the single one does. A triple-EPIs structure is designed to enhance the device EF between two trenches at the top and p
- well/n
− EPI area between two trenches to enhance device breakdown voltage. From
Figure 7, we can observe a more uniform electric field distribution and the largest area under EF integration along the cell depth can be found in the triple-EPIs design. Therefore, the breakdown voltage of a triple-EPIs device can be improved. All devices’ performances are summarized in
Table 3. By using the same EPI thickness, the triple-EPIs design has the highest breakdown voltage than other EPI structures.
Then we modified single-EPI and double-EPIs specifications to sustain the same V
BR that a triple design can achieve. To increase the V
BR of these two devices, the thickness and resistance of each EPI layer, as well as the trench depth, have to be increased. The EPI information for all structures is list in
Table 4.
Figure 8 shows the potential profiles for all devices. It can be seen that, in order to sustain a higher rating voltage, the thickness and resistivity of the single-EPI and double-EPIs structure must be thickened and increased to achieve a high V
BR. The EF magnitude distributions of all structures are shown in
Figure 9 and
Figure 10. We can find that, the less EPI layers that are used, the lower the electric field valley, which weakens the support of a high V
BR with a small R
on,sp. The triple-EPIs structure uses a middle EPI to enhance its electric field in the middle of the trench, where there is an EF valley observed in other structures. Therefore, a triple-EPIs structure is much easy to sustain a high V
BR than other devices.
Figure 11 shows the output characteristics for all structures with the same V
BR of 164 V. It can be observed that the R
on of a triple-EPIs design is much lower than those of the others. The triple-EPIs structure can sustain a higher V
BR, owing to a more uniform electric field distribution between two trenches that is attributed to top- and middle-EPI design. It makes a triple-EPIs device more flexible on resistivity and thickness design for bottom EPI to achieve a low R
on characteristic.
Table 5 demonstrates the R
on,sp for all devices with the same V
BR. The simulated R
on,sp of a triple-EPIs device with a rating voltage of 150 V is only 62% and 18.3% of the one for the double-EPIs and single-EPI structure, respectively. Although a double-EPIs structure has better R
on,sp than the single one, the long trench depth, accompanied by a long top EPI thickness, makes it is hard to maintain a uniform electric field between two trenches. Therefore, a higher resistivity bottom EPI spec is required to sustain a high rating voltage that results in a higher R
on than the triple-EPIs design. Compared with other methods mentioned in [
15,
18,
19,
20,
21], the multiple-EPIs structure does not complicate the process in manufacturing, and a higher-V
BR and a lower-R
on,sp device can be achieved.
We also use the same method to construct 200 V SGT devices with different EPI designs. Similar electrical field distributions and output characteristics with
Figure 10 and
Figure 11 can be obtained, respectively, if we modify the best epitaxial specification.
Table 6 lists the parameters that we used for all 200 V SGT devices’ simulation and shows their characteristics. Again, the triple-EPIs structure demonstrates more flexibility to achieve a lower R
on,sp than the single-EPI and double-EPIs devices under the same breakdown voltage design.
Figure 12 compares the specific on-resistance performance of our proposed SGT devices with that of the other middle-voltage devices reported in [
4,
15,
21,
32,
33,
34,
35,
36,
37,
38,
39,
40], ideal silicon limit, and super junction (SJ) limit for cell pitch = 5 and 10 µm in the 50–200 V range. Form
Figure 12, we observe that the triple-EPIs structure and those using a double split-gate device [
15] and stepped oxide SGTs [
18,
20,
21] can achieve a very low R
on,sp in the middle-voltage range because they all can maintain more uniform EF distributions between two trenches. Compared with a double split-gate device and stepped oxide ones, our triple-EPIs devices do not require the complicated double split-gate or oxide-engineering process in the trenches and is compatible with the conventional SGT process.