1. Introduction
Silicon Carbide (SiC) is a kind of wide bandgap semiconductor material. Thanks to its excellent physical properties, SiC has advantages over Silicon (Si) in power electronics devices [
1]. SiC metal-oxide semiconductor field-effect transistors (MOSFETs) are capable to perform well in high-frequency and high-temperature conditions with negligible tail currents and low switching oscillations [
2,
3]. SiC double-implanted MOSFET (DMOSFETs), also called planar-gate SiC MOSFETs, are the most mature type of SiC MOSFET and they are becoming more popular in the design of high-power-density and high-efficiency power electronics systems, such as motor drivers and charging piles of electric vehicles [
4,
5]. But in terms of reliability, especially the short circuit and avalanche ruggedness, they are far weaker than traditional Si IGBTs, which hinders them from large-scale applications [
6,
7].
Thermal runaway accounts for 38% of the causes of SiC MOSFET’s failures in short-circuit faults [
7], and it is also the main failure mechanism in avalanche breakdown faults [
6,
8,
9]. Thus, for integrality and accuracy, thermal-runaway failures are necessary to be included in the fault simulations of SiC MOSFET. However, finite-element simulation, which is often used to evaluate faults and failure of power electronics devices, is inconvenient for circuit designers. Behavior models are widely used by researchers and engineers in the design, optimization, and faults diagnosis of power electronics systems and design of protection circuits of power devices due to their fast simulation speed and considerable accuracy in circuit system simulations. However, the majority of behavior models of SiC MOSFETs focus on the accuracy of static and dynamic characteristics simulation at an operating temperature that is not extremely high [
10,
11,
12,
13,
14,
15,
16,
17]. They cannot simulate the devices under extreme operating conditions, like short-circuit and avalanche breakdown fault. For users of the devices and the circuit designer, if there exists a behavior model that can describe the faults and thermal-runaway failures of power devices, it is more convenient to carry out some simulations in extreme conditions, validate the design of protection circuit, and evaluate the reliability of the power devices in a power electronics system. For designers of power devices, if a behavior model is capable of reflecting the real cell-level parameters and the physical processes in faults and failures, it is easier to find out the key point in cell-level design to improve the device’s reliability in faults.
In recent years, several papers have paid attention to modeling SiC MOSFET’s faults [
18,
19,
20,
21,
22,
23]. The majority of them focus on fitting short-circuit characteristics of SiC MOSFETs without considering the physical mechanisms, therefore being less generic. Some models simulate short-circuit currents with equations entirely obtained using data fittings [
18,
19]. They cannot fully describe the physical processes of SiC MOSFET’s short-circuit fault and cannot ensure accuracy in different working conditions due to a lack of physical basis. In [
20,
21], electrothermal models are established to simulate temperature-dependent characteristics during short-circuit faults, aiming to clarify the short-circuit failure mechanisms and achieve failure prediction. But they can only be used to calculate the short-circuit withstand time, not for circuit simulation. A physically based short-circuit model of SiC MOSFET is presented in [
22]. It introduces several physical models into parameter calculations for more accurate simulation and can emulate the short-circuit failure, whereas other faults, such as avalanche breakdown, are not included in it. Michele Riccio et al. proposed a temperature-dependent model accounting for both short-circuit and avalanche breakdown faults [
23]. This model is built based on the physical processes of both faults, but the key parameters are all calculated using data fittings, which reduces its physical meaning and may impact its universality.
In a word, the existing models still suffer from one or more of the following weaknesses: (1) the model includes only short-circuit fault and cannot describe avalanche breakdown fault at the same time; (2) it lacks a physical basis, which may decline universality and accuracy of the model; (3) it cannot characterize failure phenomena that occur during faults.
To overcome the aforementioned problems, this paper presents a behavior model of SiC DMOSFET considering thermal-runaway failure in short-circuit and avalanche breakdown faults. The proposed model is more complete and universal than the existing ones because it is built based on cell-level processes of short-circuit and avalanche breakdown faults and can describe thermal-runaway failures in both faults. It can help device users design the protection circuit and evaluate the ruggedness of SiC MOSFET in some to improve the reliability in application. Meanwhile, for designers of SiC DMOSFET, the proposed model can help them intuitively and quickly analyze the influence of the design of the device on the characteristics of it and improve the design.
This paper is organized as follows.
Section 2 clarifies the physical basis of the proposed model: it analyzes the cell-level physical processes of both faults by finite-element simulation and further discusses the thermal-runaway failure mechanisms with reference to existing studies. On this basis,
Section 3 then illustrates the working principles of the model.
Section 4 presents the calculation of the parameters used in the model: junction temperature, carrier mobility, leakage current, and so on. In
Section 5, the proposed model is validated using a short-circuit test and unclamped inductive switching (UIS) test experiments, and a commercialized SiC DMOSFET, C2M0080120D produced by Wolfspeed, is selected as the modeling object. Finally,
Section 6 concludes this paper.
2. Physical Basis
The cell-level physical processes of short-circuit and avalanche breakdown faults and the failure mechanisms of both faults are the physical basis of the proposed model. However, the cell-level phenomena are microscopic and hard to observe. To clarify the physical basis of the proposed model deeply, in this section, a finite-element model of SiC DMOSFET is built and short-circuit and avalanche breakdown test simulations are carried out. Based on the simulation results, the cell-level physical processes of short-circuit and avalanche breakdown faults are analyzed. Furthermore, with reference to the existing studies, the failure mechanisms of SiC DMOSFET during both faults are discussed in depth.
2.1. Finite-Element Cell Model and Fault Simulations
Thus, in order to clarify the physical basis of the proposed model, as shown in
Figure 1, a 2D finite-element cell model of SiC DMOSFET is established in Synopsys Sentaurus TCAD according to the practical structure of the modeling object. The doping distributions of the P-well, N+ and P+ regions are completed using ion implantation simulations, in which Nitrogen is set for the N-type doping and Phosphorus is set for P-type doping. The shape and boundary of the regions in the model are similar with those regions formed using ion implantation in real devices [
24]. The key structure parameters of the cell are given in
Figure 1.
Appropriate physical models are selected in the finite-element simulation to enable the simulated results to reveal the realistic cell-level physical processes. The default drift-diffusion model is used for carrier transportation calculation and Fermi statistics are chosen as the carrier distribution function here. In terms of carrier mobility calculation, the Masetti model describing doping-dependent mobility degradation is used and the high field saturation effect is taken into account. Furthermore, to calculate the electron mobility near SiC/SiO2 interface, the University of Bologna mobility model and the interface charge model is included. Moreover, the combination of the exponential and uniform model, N0exp(−(E − E0)/ES) + N1, is used, to fit the energetic distribution of both donor and acceptor interface traps, where DIT,T = 1.3 × 1013/eV represents the energetic distribution of the interface state density near the band edge and DIT,M = 4.0 × 1012/eV represents that near the middle of band. In the finite-element simulation, ES = 0.069 eV describes the decay rate of the interface state density from band edge to the middle.
For the recombination model, the SRH, Auger and avalanche recombination are all considered and the Okuto–Crowell model is selected to calculate the impact ionization. Also, incomplete ionization and the anisotropy of 4H-SiC are considered. To emulate the lattice temperature variation during faults, the analytic thermoelectric powers model and thermodynamic model are introduced, and the model’s working temperature in static characteristics simulations and the initial temperature in fault simulations are set to 300 K.
The model’s simulated on-resistance is 90.9 mΩ and the simulated breakdown voltage is 1686 V, whose relative error to the measured results of C2M0080120D is approximately 1.6% and 2.3%, respectively. The simulated static characteristics of the model cannot be identical to a realistic device, because they can be significantly influenced by the incomplete physical model and some unreasonable parameters sets in the simulation software. However, it does not prevent the model and simulation from showing the correct semiconductor-level physical processes. The convinced cell structure, doping distribution formed using process simulation, and the proper physical models selected according to realistic physical mechanisms all indicate that the model is able to describe the modeling object’s cell-level physical processes during faults.
Then, short-circuit and UIS test circuits are built in Sentaurus TCAD and the cell model is put into it for fault simulations. In practical short-circuit and UIS test platforms, there may be some branches for protection between drain-gate or some branches between drain-source as buffer circuits. The aforementioned branches may affect the dynamic responses of the tests, but those factors do not actually influence the cell-level physical processes of SiC DMOSFET during faults. Hence, the simulation test circuit is simplified as shown in
Figure 2, in which the inductance
L here represents a small stray inductance of the power loop in the short-circuit test and a large load inductance in the UIS test.
By means of the fault simulations of the finite-element cell model, not only are the current and voltage waveforms obtained, but also the distributions of cell-level physical quantities, such as total current density, electric field, lattice temperature and so on, can be observed. These results may help to straightforwardly analyze the cell-level physical processes of SiC DMOSFET during faults as follows.
2.2. Cell-Level Physical Processes of Short-Circuit Fault and the Thermal-Runaway Failure in It
2.2.1. Cell-Level Physical Processes of Short-Circuit Fault
In
Figure 3a, the solid blue lines and dotted orange line, respectively, represent the drain current
ID and the drain-source voltage
UDS waveforms in short-circuit fault test simulation when the failure does not occur. The process of SiC DMOSFET’s short-circuit fault can be divided into three stages according to its working states, and the vertical black dashed lines show the boundaries between stages.
Figure 3b gives the corresponding maximum lattice temperature
Tmax curve in short-circuit fault test simulation.
With reference to the waveforms and the typical cell-level total current density distributions of these stages obtained from the simulation, as given in
Figure 4, the physical processes of SiC DMOSFET’s short-circuit fault can be demonstrated:
In stage I, gate drive voltage
UGS = −5 V and the SiC MOSFET is blocking. As shown in
Figure 4a, the value of cell-level total current density is very low, which means almost no current flows through the device.
In stage II,
UGS increases to 20 V and the device turns on. As shown in
Figure 4b, the current begins to flow through the channel at the cell level.
ID increases and then falls because the channel electron mobility falls with increasing temperature.
In stage III,
UGS turns to −5 V and the device turns off.
ID does not drop directly to zero, but first drops sharply to a small value and then slowly to zero, which is called tail current. At the cell level, as provided in
Figure 4c, there is still leakage current flowing at the channel, but the value is far lower than that in stage II. The tail current is composed of the leakage current and it increases with the length of short-circuit pulse and maximum lattice temperature, indicating that high temperature is a main cause of non-negligible tail current [
9,
25,
26].
2.2.2. Discussion on the Short-Circuit Failure Mechanism
In the past few years, plenty of researchers have focused on SiC DMOSFET’s failures in short-circuit faults. Most of them reach an agreement that there are two failure modes in SiC DMOSFET’s short-circuit fault: thermal runaway and gate oxide breakdown [
7,
9,
25,
26,
27,
28,
29,
30,
31,
32,
33,
34,
35,
36,
37]. As for the origin of thermal runaway failure, there are several different explanations.
On the one hand, some researchers find that the source metal will melt under high-temperature conditions, causing the incapacity of drain and source during fault, which may lead to the failure [
32,
33,
34]. However, lots of experiment results can prove that delayed thermal runaway exists in SiC DMOSFET’s short-circuit test [
25,
29]. As the
ID waveforms shows in the experiment results provided in those references, when
UGS returns to −5 V,
ID maintains a low value at first and then increase sharply, which indicates that the gate driver can no longer control the device and the thermal-runaway failure happens. If the melting of the source metal is the origin of thermal-runaway failure in a short-circuit fault, the current should increase sharply, which therefore reflects that the melting of the source metal is not the failure mechanism but the result of a failure, and the failure occurs at the semiconductor level.
On the other hand, more studies agree that the main cause of thermal-runaway failure is a non-negligible leakage current generated by a high temperature during the fault, and the leakage current may consist of a channel leakage current, a thermal generation current, and a current generated by parasitic BJT [
7,
9,
25,
26,
28,
29,
31]. When the leakage current is high enough to trigger a positive temperature feedback, the device is out of control and the thermal-runaway failure happens. This mechanism has been demonstrated by sufficient experimental results and elaborate simulation analysis [
9,
25,
26,
29], so leakage current seems to be the more convincing short-circuit failure mechanism.
To clarify the physical processes of thermal-runaway failure in SiC DMOSFET’s short-circuit failure, we lengthen the short-circuit pulses in finite-element fault simulations and the
UDS,
ID and
Tmax curves in the case that the failure occurs can be obtained, as
Figure 5 shows. The processes can also be divided into three stages. In stage I and II here, the physical processes obtained from the simulation results are essentially the same as the case that the failure does not occur. When the
UGS returns to zero at the beginning of stage III,
ID does not decline but increases rapidly, which implies that the gate driver can no longer control the device and the thermal-runaway failure happens. The total current density distribution of stage III is given in
Figure 4d. It can be seen that the leakage current flows through not only the channel, but also the bottom of P-well region. Furthermore, compared with
Figure 4c, the value is much higher.
Figure 6 compares the lattice temperature distribution of short-circuit faults before and after
UGS returns to −5 V for both cases that the failure does not occur and occurs. It is obvious that lattice temperature is much higher after
UGS returns to −5 V in the failure case, especially at the regions beside channel. As provided in
Figure 5b, after the
VGS returns to −5 V,
Tmax does not decline like that in
Figure 3b but still maintains the increasing trend. In
Figure 6b, it can also be seen that the lattice temperature at channel does not decrease when short-circuit failure occurs. This means that there is a positive temperature feedback to hold the temperature. The results are the same with the existing studies [
7,
28,
31].
According to the aforesaid illustrations and discussions, the short-circuit fault and failure part of the proposed model will be established based on the cell-level physical processes shown in simulation results and the failure mechanism discussed here; that a non-negligible leakage current generated by a high temperature leads to the thermal-runaway failure in SiC DMOSFET’s short-circuit fault.
2.3. Cell-Level Physical Processes of Avalanche Breakdown Fault and the Thermal-Runaway Failure within It
2.3.1. Cell-Level Physical Processes of Short-Circuit Fault
Figure 7 gives the
UDS,
ID, and
Tmax waveforms in UIS test simulation when the failure does not occur in avalanche breakdown fault. According to the working state of the device, the processes of avalanche breakdown in UIS test can be divided into four stages and the vertical dashed lines represent the boundaries between stages.
Similarly, with reference to the waveforms and the cell-level total current density distributions of these stages obtained from the simulation, as given in
Figure 8, the physical processes of SiC DMOSFET’s avalanche breakdown fault can be demonstrated:
In stage I, the device is in off-state. As shown in
Figure 8a, at the cell level, there is only leakage current flow through the channel and the value is so small that it can be ignored.
In stage II, the device turns on. ID increases linearly because a large inductance is in series. At the cell level, the current flows through a thin channel closed to the SiC/SiO2 interface, just like the stage II of short-circuit fault.
In stage III,
UGS returns to −5 V and the device withstands very high voltage stress, causing an avalanche breakdown fault to occur at that moment. During a fault,
UDS is clamped to breakdown voltage, and it varies with increasing
Tmax. At the cell level, as shown in
Figure 8c, it is obvious that the PN junction at the corner of the P-well region is broken down and the current path changes from the channel to the P-well region.
In stage IV, ID decreases to zero and UDS drops to VDC, which means the avalanche breakdown fault is over and the device returns to the blocking state. The cell-level total current density distribution is just the same as that in stage I.
2.3.2. Discussion on the Avalanche Breakdown Failure Mechanism
For the failure modes in SiC DMOSFET’s avalanche breakdown faults, a lot of studies have been presented. All of them agree that high temperature causes the thermal-runaway failure. In terms of failure mechanisms, there are four mainstream views: (1) the device’s junction temperature is so high that it will exceed the intrinsic limit of 4H-SiC; (2) the melting of the source metal; (3) the activation of the channel because the threshold voltage decreases with increasing junction temperature; (4) parasitic BJT latch-up at a high junction temperature [
6,
8,
9,
15,
38,
39,
40,
41,
42,
43,
44,
45,
46,
47,
48,
49,
50,
51,
52].
The intrinsic carrier concentration of 4H-SiC is low because of its wide bandgap and it reaches 1.0 × 10
16 cm
−3 at 1270 °C [
38,
39,
53]. Some researchers estimate SiC DMOSFET’s junction temperature at the moment that avalanche failure occurs [
38,
40,
41,
42]. Though they give different estimated varying from 510 °C to 948 °C, which is a wide scale, it is obvious that the junction temperature during an avalanche breakdown fault cannot reach the intrinsic temperature limit of 4H-SiC. Hence, it seems that intrinsic limit is not likely to be SiC DMOSFET’s avalanche failure mechanism.
In the past several years, the majority of researchers agree that the avalanche failure mechanism is that the melting of source metal, like aluminum, leads to a short circuit between source and drain [
38,
41,
43]. The reason given by them is that the estimated junction temperature when an avalanche failure occurs will be close to or exceed aluminum’s melting point and the temperature is not high enough to cause a latch-up of parasitic BJT and reach intrinsic limit. However, according to optical microscope diagram of failure devices shown in [
8,
43,
44], the failure site steadily locates in the source pad near the bonding wire in different experiments.
Figure 9 gives the schematic diagram of the cell-level structure in this area. At the location of source contact, there is a nickel (Ni) layer between the Al layer and 4H-SiC. The melting point of Ni is 1453 °C, which is much higher than that of Al and even 4H-SiC’s intrinsic limit. And in other locations, Al and 4H-SiC are also isolated. Furthermore, as aforementioned, the junction temperature cannot reach the intrinsic limit and the device still have blocking characteristics. The above illustrations suggest that if only the Al melts, the Al cannot be in contact with 4H-SiC and the melting of the Al cannot lead to a short circuit of source and drain. Combined with the junction temperature estimations provided in [
38,
41], the melting of source metal is more likely to be an inducement but not the immediate cause of SiC DMOSFET’s avalanche failure. Meanwhile, the immediate cause is more likely to occur at another site, maybe at the semiconductor level.
As for activation of the channel and parasitic BJT latch-up, failure mechanisms at the semiconductor level are also supported by several researchers [
6,
9,
45,
46]. Lots of researchers disagree with these opinions because they think the junction temperature during SiC DMOSFET’s avalanche breakdown cannot reach the value required by these failure mechanisms at a semiconductor level. However, there is some compelling evidence supporting the fact that mechanisms at the semiconductor level exist in SiC DMOSFET’s avalanche failure: ref. [
46] gives some experiment results to show that SiC DMOSFET’s avalanche failure is related to the turn-off voltage of
UGS; and by means of analytically modeling the parasitic BJT, ref. [
6] indicates that the BJT can be triggered at a temperature that is possible to be reached during avalanche breakdown fault. Moreover, during avalanche breakdown fault, the temperature distribution is not uniform. There exists hot spots in the active area of a die [
43] and the temperature at the hot spot can be much higher than the estimated average junction temperature. Thus, the failure caused by mechanisms at the semiconductor level seems possible to occur in SiC DMOSFET’s avalanche breakdown fault. It may be the immediate cause of avalanche failure after the melting of the source metal. But confirming which of the correct avalanche failure mechanisms of SiC DMOSFET is correct still needs further verification.
To sum up the above discussions, the melting of the source metal seems more likely to be an inducement or a precursor of SiC DMOSFET’s avalanche failure; the immediate cause of the failure is still undecided.
Because SiC DMOSFET’s thermal-runaway failure may be related to the melting of source metal, which occurs out of the semiconductor, the finite-element simulation is not able to emulate this phenomenon. So measured results obtained from experiments, combined with the failure mechanisms discussed above, are used here to explain the physical processes of thermal-runaway failure in SiC DMOSFET’s avalanche breakdown fault.
Figure 10 gives measured
UDS and
ID waveforms in the UIS test experiment and it can also be divided into four stages. It is obvious that the waveforms in stage I, II, and III are almost identical with them in the stages shown in
Figure 6, indicating that the physical processes of them are almost the same, too. In stage IV, the device reaches a critical threshold. Then, the voltage returns to a very low value and the current begins to rise again in accordance with the slope in stage I. From the perspective of external characteristics, the phenomenon is similar to a short circuit between source and drain.
On the basis of the aforesaid illustrations and discussions, the short-circuit fault and failure part of the proposed model will be established, based on the cell-level physical processes shown in simulation results and the failure mechanism discussed here that a non-negligible leakage current generated by a high temperature leads to the thermal-runaway failure in SiC DMOSFET’s short-circuit fault. Also, the avalanche breakdown faults and thermal-runaway failure parts of the proposed can be built based on the above analysis. Critical threshold of the failure can be decided according to the thermal-runaway failure mechanisms discussed above, and it will be illustrated elaborately in Section III.
3. Working Principles of the Proposed Behavior Model
Based on the cell-level physical processes of both faults and thermal-runaway failures discussed in Section II, five working states can be defined to completely describe SiC DMOSFET’s behaviors in short-circuit and avalanche breakdown faults: off-state (
Figure 4a and
Figure 8a), on-state (
Figure 4b and
Figure 8b), leakage current state (
Figure 4c,d), avalanche breakdown state (
Figure 8c), and avalanche failure state. The five states are explained as follows:
Off-state and on-state are two basic working states that describe SiC MOSFET’s behaviors in conduction and blocking conditions, respectively. Leakage current is critical in tail current and thermal-runaway failure in short-circuit fault. Therefore, leakage current state is introduced. It is worth noting that according to semiconductor physics and the simulation results in Section II, leakage current exists in all working states of SiC MOSFET, but it can be ignored when the junction temperature is low. Hence, leakage current should be included in all five states, whereas it plays a leading role in the leakage current state (at a high junction temperature) and can be negligible in other working states. Because avalanche breakdown fault is caused by impact ionization and the thermal-runaway failure in it is induced by different mechanisms, two different working states, avalanche breakdown state and avalanche failure state, should be separately defined to characterize them. If a behavior model of SiC DMOSFET is able to take all five working states into consideration, it can completely describe all behaviors including short-circuit and avalanche breakdown faults and the thermal-runaway failures.
Figure 11 gives the circuit diagram of the proposed behavior model in this paper. It consists of two parts: a traditional behavior model of SiC MOSFET that describes on-state and off-state (the light blue part) and several extended parallel branches for describing leakage current, avalanche breakdown, and avalanche failure states (the pale golden part).
In the traditional behavior model, voltage-controlled current source IMOS characterizes basic voltage-current relationships calculated in the metal-oxide-semiconductor (MOS) structure. CGS, CGD, and CDS are parasitic capacitances between three terminals, which play critical roles in the dynamic responses of the model. D represents the body diode and RG is the gate resistor integrated into the package. RD represents the total resistance of the N− drift layer and JFET region. Furthermore, because channel electron mobility varies with increasing junction temperature in short-circuit faults, it requires special calculations for carrier mobility used in IMOS and RD calculations according to semiconductor and device physics.
A controlled current source ILEAK is included to describe the leakage current state. The physical mechanisms generating the current in leakage current state are totally different from that in on-state, so ILEAK should be placed in a separate branch.
The physical mechanism of the PN junction’s avalanche breakdown is the current amplification effect caused by a high electric field. The current increases very sharply when
UDS reaches a critical value, also called the breakdown voltage. According to the above physical basis, a separate branch made up of a temperature-controlled voltage source
UBR and a controlled switch
SBR is included to characterize SiC DMOSFET’s avalanche breakdown state. Their working principles have been demonstrated in previous work [
54].
As said in
Section 2, from the perspective of external characteristics, the physical process of avalanche failure is similar to a short circuit happening between source and drain. Using the same idea of avalanche breakdown state modeling, as shown in
Figure 11, the processes of an avalanche failure state are modeled as a separate branch consisting of a temperature-controlled switch
SF in the proposed behavior model. The avalanche failure mechanisms discussed in
Section 2 suggests that the melting of Al seems to be an inducement or a precursor of thermal-runaway failure in SiC DMOSFET’s avalanche breakdown fault, and the undecided mechanism occurs after that is more likely to be the immediate cause. Because the direct failure mechanism is still undecided, it can be simplified as a process spending a short period of time. Thus, the control logic of the switch can be defined as follows: after the junction temperature exceeds a critical threshold temperature
TCRIT, the melting of aluminum by a short time
tFD, enables the
SF to turn on and it cannot be turned off again. The
TCRIT characterizes the melting of the source metal, and the
tFD describes the undecided failure mechanism that occurs after the melting of the source metal. According to subsequent experimental verifications, this modeling approach for thermal-runaway failure in SiC DMOSFET’s avalanche breakdown fault is accurate enough and it is easy to modify and improve if the failure mechanisms are clarified further.
The junction temperature T significantly influences SiC DMOSFET’s characteristics in short-circuit and avalanche breakdown faults. So, the calculation of T should be included.
The models of all components and calculations of the key parameters in the proposed behavior model will be elaborated on in
Section 4.
4. Models and Calculations
In this section, the modeling of all components and calculation methods of the key parameters in the proposed behavior model will be elaborated on, including the channel current, parasitic junction capacitances, the resistance of N-type drift layer and JFET layer, leakage current, junction temperature, and key physical parameters used in the above calculations such as intrinsic carrier concentration, threshold voltage and carrier mobility. A physically based and accurate methodology for avalanche breakdown voltage calculation has been explored in previous work [
54] and it will not be repeated in this section.
Figure 12 displays the summary of parameters calculations in the proposed model. The parameters used in this section are listed in
Table A1 in the
Appendix A. The majority of the parameters related to device structure and fabrication can be obtained from the finite-element model of
Section 2 and the open literature. Some fitting parameters can be obtained through curve fitting.
4.1. Channel Current IMOS, Parasitic Junction Capacitance CGS, CDS, CGD, and Body Diode D
In the proposed model, the channel current
IMOS is calculated using (1). The selected model is derived in an ideal N-channel enhanced MOS structure and takes the channel length modulation effect into account [
19]. It is widely used in characterizing MOSFET’s fundamental voltage–current relationship.
In (1), channel electron mobility μCH and threshold voltage VT are the key parameters influencing the variation in current and the calculations of them will be presented later. Cox is the capacitance of oxide per unit area; LCH is the length of channel; LCELL represents the equivalent total width of channel considering the number of cells; λ is the channel length modulation coefficient.
Models of parasitic junction capacitances are necessary for switching process simulations. In the proposed model,
CGS is modeled as a constant capacitance,
CGD and
CDS are, respectively, modeled as
Crss and (
Coss−
Crss), in which
Crss and
Coss is calculated using (2) and (3) [
55].
Cr0,
Vr,
Mr,
δr,
γr,
CrssFD,
Vo,
Mo,
δo,
γo, and
CossFD are all fitting parameters used for parasitic capacitance calculation. In (2) and (3), except
UDS which represents the drain-source voltage, all other parameters are obtained from data fitting according to the datasheet.
The parasitic body diode is an indispensable part of SiC DMOSFET, but it has almost no effect on the fault and failure characteristics of the device. Hence, the body diode is modeled as an anti-parallel ideal diode in the proposed model. The results shown in Section V can verify that the simplified model of the body diode does not influence the accuracy of the proposed model.
4.2. Drain Resistance RD
The resistance of the N-type drift layer and the JFET region of SiC DMOSFET changes with increasing junction temperature, which can significantly influence the characteristics of the device during fault because the junction temperature increases sharply. Thus, the resistances are carefully modeled as follows and they are combined into a drain resistance, RD, in the proposed model.
The resistance of the JFET region can be calculated using (4).
In (4), the
μN represents the electron mobility of the JFET region and the N-type layer, and its calculation will be presented in part E;
q represents elementary charge;
NJFET is the equivalent doping concentration of the JFET region;
WSCR is the width of the space charge region between the JFET region and the P-well region and it can be calculated using (5).
In (5),
ni, the intrinsic carrier concentration, will be modeled physically in
Section 4.5;
εSiC is the permittivity of 4H-SiC;
NDR is the doping of the N-type drift layer;
kB represents the Boltzmann constant;
NPWELL is the equivalent doping of the P-well region;
UCH is the voltage on the channel.
According to the cell-level total current distribution obtained from the finite-element simulation, when SiC DMOSFET is in on-state, there is a trapezoid-like “current diffusion layer” located at the top of the N-type drift layer because the depletion region is blocked. Under the layer, the current distribution is uniform. So, the resistance of the N-type drift layer can be divided into two parts. The upper one, named as
RDRU, represents the resistance of the current diffusion layer and is modeled as a trapezoidal resistance. It can be calculated using (6).
In (6),
HDIFF is the thickness of the current diffusion layer,
WCELL is the width of a cell,
WJFET is the width of the JFET region. The lower one, named as
RDRL, represents the resistance of the layer in which current flows uniformly and it can be calculated using (7).
In (7), HEPI is the thickness of the epitaxial layer.
On this basis, the drain resistance in the proposed model can be defined as (8).
4.3. Leakage Current ILEAK
The leakage current can be negligible in SiC DMOSFET under low-temperature conditions. But in high-temperature cases, it has a strong influence, which is a distinguishing feature of the SiC MOSFET’s short-circuit fault and the main cause of short-circuit failure. Hence, the accurate calculation of the leakage current is necessary. For SiC DMOSFET, the leakage current usually consists of three components: thermal generation current
ITH, diffusion current
IDIFF, and avalanche leakage current
IAV(LEAK) [
12,
14]. According to previous calculations, in SiC DMOSFET’s thermal-runaway failure in a short-circuit fault, the thermal generation current
ITH plays the leading role in total leakage current, and the other two components are miniscule, so they can be ignored in the total leakage current.
In the proposed model, a controlled current source
ILEAK is used to characterize the leakage current, and it is expressed by (9).
In (9), S is the equivalent junction area generating leakage current; τg is the carrier lifetime.
4.4. Junction Temperature
During short-circuit and avalanche breakdown faults, SiC MOSFET generates high power dissipation, which causes the junction temperature to increase sharply and can considerably influence its behavior.
An
RC thermal network is widely used for junction temperature evaluation in behavior models of power semiconductor devices. In the proposed model, a 14-order Cauer thermal network, shown in the red part of
Figure 12, is included to calculate the junction temperature. As provided in
Table 1, the values of the thermal resistance
Rthi and thermal capacitance
Cthi used in this paper are obtained from the SPICE models given via Wolfspeed, Durham, NC, USA.
4.5. Physical Parameters
As mentioned above, some key physical parameters are introduced in the proposed model, such as intrinsic carrier concentration, threshold voltage, and carrier mobility. They have important effects on SiC DMOSFET’s behaviors under fault and failure conditions, because they significantly change with the sharp increase in junction temperature. Thus, it is necessary to calculate them carefully. In this part, the models of intrinsic carrier concentration, threshold voltage, and carrier mobility are presented as follows.
4.5.1. Intrinsic Carrier Concentration
Intrinsic carrier concentration
ni is the key factor deciding the high-temperature characteristics of the power semiconductor devices. In this paper,
ni is calculated using its definition given in (10) [
56].
In (10),
Eg is the band gap of 4H-SiC, which can be calculated using (11) [
57].
Nc and
Nv, provided in (12) and (13) [
56], represent the effective density of state in the conduction band and the valence band, respectively.
In (12) and (13), me is the electron mass, h is the Planck constant.
4.5.2. Threshold Voltage
Threshold voltage
VT is a temperature-sensitive parameter of SiC DMOSFET and it significantly influences the high-temperature behavior of the device. In this paper, modeling of the
VT is given in (14), which considers the fixed charges in gate oxide, the traps at the SiC/SiO
2 interface, and the correction of the surface potential in the flat-band voltage calculation [
39].
In (14),
nF is the density of the fixed charge at the SiC/SiO
2 interface;
ΦMS is the work function difference between the high-doped poly silicon gate and the P-type 4H-SiC at the channel. It can be defined as (15).
In (15),
χ is the electronic affinity of 4H-SiC;
ΨF represents the Fermi potential at the channel and it can be calculated using (16).
V0 is a constant related to the material properties and the design of device and it is defined as (17).
In (17), NCH is the equivalent doping concentration of the channel.
As for
QIT, the charge density of interface traps, is the key parameter influencing the value and high-temperature characteristics of
VT. It also has a significant influence on the channel electron mobility
μCH. The modeling and calculation of it will be presented in the
Appendix B.
4.5.3. Carrier Mobility
In SiC DMOSFET, the carrier mobility is one of the most important parameters affecting the electric conductivity and high-temperature characteristics. It is decided by multiple different physical mechanisms, and generally, it can be calculated using (18).
In (18):
μTOTAL is the carrier mobility that considers several physical mechanisms;
μ1,
μ2,
μ3 are the ones affected by only one physical mechanism. In general, there are four types of physical mechanisms and the corresponding carrier mobilities that should be taken into consideration are bulk mobility (
μB), acoustic-phonon scattering (
μAC), surface roughness scattering (
μSR), and Coulomb scattering at interface traps (
μC). The above four kinds of carrier mobility can be calculated using (19)–(22) [
22,
39,
58,
59,
60,
61,
62,
63,
64].
In the above equations,
μMIN,
μL,
μREF,
K1,
K2,
ΓSR,
ΓCit,
nSCR, and
ζC are the physics-based parameters used in carrier mobility calculation; the effective perpendicular electric field
ENOR and the interface inversion charge
nINV are calculated utilizing the charge sheet model [
60,
65]; the interface trapped charge
nIT is calculated with references to [
22,
66,
67]. The models and calculations of them are complex works and, therefore, will be elaborated in the
Appendix B.
In the channel of SiC DMOSFET, limited by the technology of manufacturing, the quality of the SiC/SiO
2 interface is poor, which has a great impact on the channel electron mobility
μCH [
39,
66]. Thus, as given in (23), the calculation of
μCH should consider all four aforesaid physical mechanisms.
Because the JFET region and N-type drift layer is inside of the 4H-SiC, only bulk mobility should be considered in the calculation of electron mobility
μN in
RD calculation. Hence, as given in (24),
μN is equal to
μB.