Steady-State Temperature-Sensitive Electrical Parameters’ Characteristics of GaN HEMT Power Devices

Abstract

Gallium nitride high-electron-mobility transistor (GaN HEMT) power devices are favored in various scenarios due to their high-power density and efficiency. However, with the significant increase in the heat flux density, the junction temperature of GaN HEMT has become a crucial factor in device reliability. Since the junction temperature monitoring technology for GaN HEMT based on temperature-sensitive electrical parameters (TSEPs) is still in the exploratory stage, the TSEPs’ characteristics of GaN HEMT have not been definitively established. In this paper, for the common steady-state TSEPs of GaN HEMT, the variation rules of the saturation voltage with low current injection, threshold voltage, and body-like diode voltage drop with temperature are investigated. The influences on the three TSEPs’ characteristics are considered, and their stability is discussed. Through experimental comparison, it is found that the saturation voltage with low current injection retains favorable temperature-sensitive characteristics, which has potential application value in junction temperature measurement. However, the threshold voltage as a TSEP for certain GaN HEMT is not ideal in terms of linearity and stability.

1. Introduction

As wide bandgap power semiconductor devices, Gallium nitride high-electron-mobility transistor (GaN HEMT) power devices have attractive advantages such as a fast switching speed, high switching frequency, and low conduction resistance, which penetrate fast chargers, data centers, 5G equipment, etc. [1,2,3,4,5]. With the decrease of GaN HEMTs’ physical size and increase in the power density requirements, the heat flux density of the devices surges significantly. For example, the heat flux density of GaN HEMT is as high as 300 W/cm², about three times that of Si-based IGBTs, as shown in Figure 1 [6]. Since a high heat flux density causes the devices’ junction temperature to increase, the junction temperature as a key reliability parameter for Si-based devices is also valuable for the reliability of GaN HEMT [7,8,9].

The junction temperature is important for the operating condition and reliability of the devices. Firstly, a high junction temperature deteriorates device performance by increasing conduction loss. Secondly, it serves as a major factor contributing to device failure, with the failure rate doubling for every 10 °C increase in the junction temperature [10,11]. The junction temperature is the primary parameter in the device lifetime model that is the basis for device health and lifetime assessment, prediction, and management [12,13]. Therefore, monitoring the junction temperature is a prerequisite for the efficient operation and reliable application of devices.

The junction temperature monitoring methods can be divided into four categories: the optical measurement method, physical contact method, thermal network modeling method, and temperature-sensitive electrical parameters (TSEPs) method. The optical method can provide a map of the temperature distribution, but it requires that the device must be unpacked [14,15]. Although the physical contact method is inexpensive and noninvasive, its slow response speed makes it difficult to track dynamic junction temperature accurately [16]. While the thermal network modeling method is noninvasive, it requires extensive calculations based on accurate and continuously updated thermal network models [17,18]. In contrast, the TSEPs method, which treats the device itself as a thermal sensor, is a promising junction temperature measurement method because of its low cost, noninvasiveness, and potential for real-time on-line monitoring [19,20]. The TSEPs method for GaN HEMT still faces challenges due to device packaging, noise, and aging [21].

According to the time-based characteristics of TSEPs, the junction temperature monitoring methods for GaN HEMT based on TSEPs can be divided into steady-state TSEPs and transient TSEPs. Steady-state TSEPs are the electrical parameters related to the junction temperature when the devices to be measured are in full conduction or full shutdown, such as the saturation voltage with low current injection [22]. Transient TSEPs are the electrical parameters related to the junction temperature of the devices to be tested during the transient switching process, such as turn-on delay time [23]. In comparison, the calibration method for steady-state TSEPs is simple and less susceptible to noise, but it faces the challenge of online application. Although transient TSEPs have the potential for real-time online extraction of the junction temperature, they are susceptible to interference during the measurement process.

Since the reliability research on GaN HEMT is still in the early stages, the TSEPs method and its related characteristics are still being explored [24]. Compared to the transient TSEPs of GaN HEMT [23,25], the steady-state TSEPs are the focus of current research. In [26], the threshold voltage exhibits a negative temperature coefficient, and a good linear relationship with junction temperature, whose sensitivity is about 0.84 mV/K. In [22], the saturation voltage with low current injection or conduction resistance can be utilized as TSEPs, exhibiting a positive temperature coefficient and good linearity, and is suitable for GaN HEMT with different gate structures. In [27], the body-like diode voltage has a positive temperature coefficient, but its sensitivity varies widely among different types of GaN HEMT. As mentioned in [28], the Schottky gate diode forward characteristics can be used effectively to measure the channel temperature of GaN HEMT, but this method is only applicable to GaN HEMT with Schottky gate structures. Similar to gate diode voltage, the gate current can only be used for GaN HEMT with Schottky gate structures and has a positive temperature coefficient and nonlinear characteristics [29]. In contrast, saturation voltage, threshold voltage, and body-like diode voltage devices are more universal as TESPs for GaN HEMT. However, the current research results solely address the relationship between TSEP and junction temperature under specific conditions. The characteristics of TSEP have not been explored, which cannot adequately provide comprehensive guidance for the junction temperature monitoring of GaN HEMT.

In this article, the temperature-sensitive characteristics of commonly used steady-state TSEPs in GaN HEMT are discussed, which is beneficial for the study of the junction temperature monitoring of the reliability of GaN HEMT. This paper is organized as follows. Section 1 emphasizes three measurement methods for TSEPs and the underlying physical mechanisms of temperature variation. Section 2 describes the construction of an experimental platform. Section 3 investigates the influence of different factors on the three types of TSEPs, with a focus on the stability of TSEPs. The results demonstrate that the saturation voltage with low current injection, as a steady-state TSEP, offers significant advantages for practical applications.

2. Measurement Method of Steady-State TESPs

2.1. Saturation Voltage with Low Current Injection

Referring to the IGBT parameter measurement standard of IEC [30], the measurement circuit for saturation voltage with low current injection of GaN HEMT is shown in Figure 2. A voltage source (V_gss) and a constant current source (I_m) are used to provide the gate voltage and test current for the device under test (DUT).

When V_gss > V_gss(th) and V_ds > 0, GaN HEMT is in a forward conduction state. At this time, V_ds is defined as the drain-source voltage drop when forward conduction occurs, which can be expressed as [31]:

$$V_{\mathrm{ds}}=I_{\mathrm{ds}}\cdot \frac{L(d+\Delta d)}{W\mu \epsilon (V_{\mathrm{gss}}-V_{\mathrm{gss}\left(\mathrm{th}\right)})}$$

(1)

where L is the channel length, Δd is the channel thickness, W is the gate width, and d and ε are the thickness and dielectric constant of the AlGaN barrier layer. μ is the electron mobility, I_ds is the drain current, V_gss is the gate voltage, and V_gss(th) is the gate threshold voltage.

Since the electron mobility μ decrease is sensitive to temperature T and has negative temperature characteristics [32], V_ds increase with μ decrease when the drain of the device is injected into a constant test current. Thus, as a TSEP, the saturation voltage with low current injection has a positive temperature coefficient.

2.2. Threshold Voltage

The threshold voltage is defined as the voltage applied to the device gate-source when the device drain is flowing the specified current at an ambient temperature. Typically, the specified drain current is in the μA or mA range [33]. Figure 3 shows the measurement circuit for the threshold voltage of the GaN HEMT. DUT is tested with the drain-gate shorted, and I_m is the constant current source.

The threshold voltage of GaN HEMT can be expressed as [34]:

$$V_{\mathrm{th}}={\varphi }_{\mathrm{b}}-\Delta E_{\mathrm{C}}-\frac{q{N}_{\mathrm{d}}{d}^2}{2\epsilon }-\frac{\sigma d}{\epsilon }$$

(2)

where ϕ_b is the height of the potential barrier, ΔE_C is the conduction band discontinuity, q is the charge and N_d is the doping density, σ is the total polarized charge density at the AlGaN/GaN interface, and d and ε are the thickness and dielectric constant of the AlGaN barrier layer.

Since the height of the potential barrier ϕ_b, the conduction band discontinuity ΔE_C, and the total polarized charge density at the AlGaN/GaN interface σ exhibit significant positive temperature characteristics [35], the threshold voltage shows different trends as the temperature changes, depending on the dominant degree of ϕ_b, ΔE_C, and σ.

2.3. Body-like Diode Voltage

Although there are no parasitic diodes in GaN HEMT, their reverse conduction characteristics are similar to MOSFETs [36]. Referring to the MOSFET body diode measurement method [37], the principle of the body-like diode voltage measurement for GaN HEMT is shown in Figure 4. I_m provides the specified test current to the device, while V_gss provides the specified gate voltage to turn off the device. Unlike the mechanism for measuring the saturation voltage with low current injection, the body-like diode voltage is typically measured by shorting or reverse biasing the gate and drain.

When V_gd > V_gd(th), V_ds < 0, and V_gss < V_gss(th), GaN HEMT is in a reverse conduction state. At this time, V_sd is defined as the voltage drop between the source and drain during reverse conduction, which can be expressed as [36]:

$$V_{\mathrm{sd}}=V_{\mathrm{gd}\left(\mathrm{th}\right)}-V_{\mathrm{gss}}+I_{\mathrm{sd}}\cdot R_{\mathrm{sd}\left(\mathrm{on}\right)}$$

(3)

where V_gd(th) is the device reverse gate threshold voltage, V_gss is the gate voltage, I_sd is the drain source current, and R_sd(on) is the reverse on resistance.

Since temperature affects both the reverse on resistance R_sd(on) and the threshold voltage V_th, and the reverse on resistance R_sd(on) has a positive temperature characteristic, the temperature characteristics of the body-like diode voltage V_sd will correlate with those of the threshold voltage V_th when a constant source drain current I_sd is injected.

3. Experimental Investigation

3.1. Introduction of Experimental Platform

Figure 5a depicts the experimental platform utilized to analyze the properties of three TSEPs, comprising saturation voltage with low current injection, threshold voltage, and body-like diode voltage. The DUT used for the test is the GaN HEMT model GS61008P produced by GaN Systems, the GaN HEMT model IGT60R070D1 produced by Infineon, and the SiC MOSFET model CMF20120D produced by Cree. To mitigate the impact of the temperature on the testing circuit, a split structure is employed. The PCB substrate containing only DUT is placed on the temperature control platform for separate heating, while other test circuits are connected to the substrate via the interface. The drive circuit is utilized to regulate the on–off state of the DUT. The constant current circuit provides a steady test current. To capture feeble variations in signals, the conditioning circuit houses an instrument amplifier that boasts high gain and an outstanding common mode rejection ratio.

During the test, the temperature control platform is set to 25 °C, 50 °C, 75 °C, 100 °C, and 125 °C, respectively. Each temperature point is maintained for 10 min to ensure that the DUT reaches the thermal equilibrium, which can be interpreted as the device junction temperature being equal to the temperature point. Considering that the components in the test circuit would be affected by temperature, the maximum temperature is set to 125 °C to evaluate the characteristics of TSEP. Figure 5b shows the measurement timing diagram. The pulsed I_m is injected into the DUT. After 2 s, the data acquisition is triggered to ensure that the platform is in a stable state. The data are collected during 100 µs after the trigger signal, with the experimental error reduced by multiple measurements and average processing.

3.2. Drive Circuit

To evaluate the effect of the drive voltage on the TSEP characteristics, the adjustable drive circuit is designed as shown in Figure 6, including the conventional drive circuit with SI8271GB-IS and the adjustable voltage unit with LM317 to achieve stable and low-noise output voltage. The adjustable voltage V_cc provided to the drive chip is adjusted through resistor R₁ and R₂, and is fed to the device gate.

3.3. Constant Current Circuit

A controllable constant current source circuit capable of providing a pulse type test current in milliampere levels is illustrated in Figure 7. Transistor Q₁ operates in an amplified state where the base current can be disregarded, and the output collector current I_c is about equal to the emitter current I_e, that is:

$$I_{\mathrm{c}}\approx I_{\mathrm{e}}=\frac{U_{\mathrm{ZD}1}-U_{\mathrm{eb}}}{R_1}$$

(4)

where U_ZD1 is the voltage drop across the Zener diode ZD₁, and U_eb is the voltage between the emitter and base, about 0.7 V. The constant current source output can be adjusted by resistance R₁. In addition, R₂ is used to limit the current and provide a steady-state operating point for ZD₁. The optocoupler chip ACPL-P346 with electrical isolation is used to control the voltage V_out fed to a constant current circuit.

3.4. Conditioning Circuit

In order to accurately measure the weak voltage signal, the conditioning circuit constructed with the instrument amplifier AD8221 is shown in Figure 8, where the resistance R₁ is 49.4 kΩ. The external resistance R_G is used to adjust the voltage gain. The resistance R and the capacitor C_C form two groups of low-pass RC networks to filter the high-frequency signals and suppress common mode interference, while the capacitor C_D affects the differential signal. The gain G of the conditioning circuit is:

$$G=1+\frac{R_1}{R_{\mathrm{G}}}=1+\frac{49.4\mathrm{k}\Omega }{R_{\mathrm{G}}}$$

(5)

To fulfill the measurement requirements, the gain is set to 495 in this paper, and the external resistance R_G is 100 Ω. For a nonperiodic rectangular pulse signal with a pulse amplitude of A and a pulse width of τ, the spectrum is [38]:

$$X(\mathrm{j}\omega )=\frac{2A}{\omega }\sin \left(\frac{\omega \tau }{2}\right)$$

(6)

The instrument amplifier used in this paper has a corner frequency of about 5.5 kHz at a given 495 times gain, and an angular frequency ω₁ of about 11,000π rad/s. According to (6), the signal spectrum at ω₁ is close to 0. Therefore, the high frequency signal with angular frequencies outside ω₁ can be ignored, and the input voltage signal can be completely collected by the instrument amplifier.

4. Experimental Results and Analysis

4.1. Characteristics of Saturation Voltage with Low Current Injection

Based on the measurement circuit shown in Figure 2 and information from the DUT datasheet, the gate voltage V_gss is set to 5 V and the test current I_m is set to 300 mA. The V_ds-T_j curves of two GaN HEMT device models GS61008P are shown in Figure 9. It can be seen that a saturation voltage with low current injection V_ds increases with the junction temperature T_j, which is consistent with the theoretical analysis. In addition, the V_ds has good linearity, whose fitting coefficient reaches approximately 0.998. However, V_ds has a temperature sensitivity of about 0.015 mV/°C and is much smaller than that of Si-based (2 mV/°C) and SiC-based devices (5 mV/°C) [10,39], which is a challenge for utilizing V_ds as a TSEP to extract the junction temperature of GaN HEMT.

To explore the influence of I_m on the temperature sensitivity of V_ds, V_gss is kept at 5 V, and I_m is successively increased from 100 mA to 300 mA. V_ds-T_j curves under different I_m are shown in Figure 10a. It can be seen that the saturation voltage with low current injection of GaN is the function of the test current and the junction temperature. Similar to IGBT devices, V_ds demonstrates nonlinear variation with I_m at a given temperature [40]. The sensitivity of this TSEP depends on the test current I_m level. As the I_m decreases, the sensitivity of V_ds declines from 0.016 mV/°C to 0.006 mV/°C, which makes the measurement of V_ds more difficult. Conversely, increasing I_m improves the sensitivity of V_ds, but it can lead to self-heating during junction temperature monitoring. According to the thermal resistance from the junction to case R_ΘJC and the power dissipation P, the temperature difference from the junction to case under steady-state is

$$\Delta T=P\times R_{\mathsf{\Theta }\mathrm{JC}}$$

(7)

Since R_ΘJC is 0.55 °C/W based on the datasheet and the maximum I_m is 300 mA, the temperature difference from the junction to case is far less than 0.001 °C, which can be ignored.

To explore the effect of V_gss on the temperature sensitivity of V_ds, I_m is set to 300 mA, and V_gss ranges from 3.5 V to 6 V. V_ds-T_j curves under different V_gss are shown in Figure 10b. It can be seen that the larger V_gss is, the better the linear fitting is. However, the sensitivity of V_ds reduces from 0.025 mV/°C to 0.013 mV/°C. Specifically, when V_gss is between about 4.5 V and 6 V, the sensitivity changes slightly within the range of 0.013 mV/°C to 0.015 mV/°C, from which it can be inferred that the conductive channel resistance characteristics are affected by V_gss. When V_gss is low, the conductive channel resistance and its sensitivity change significantly with V_gss, and the temperature sensitivity of V_ds is easily affected by V_gss. Conversely, when the sensitivity of conductive channel resistance tends to be stable with the increase of V_gss, the influence of V_gss on the temperature sensitivity of V_ds is weak.

4.2. Characteristics of Threshold Voltage

The test current I_m of 10 mA is used, based on the schematic diagram presented in Figure 3 and information from the DUT datasheet. Figure 11 shows the V_th-T_j curves of two GaN HEMT devices of model GS61008P. Unlike that of Si-based devices and SiC-based MOSFET devices [41,42], the threshold voltage V_th of the GaN HEMT has a positive temperature coefficient, probably because the temperature-dependent behavior of the barrier height plays a dominant role based on (2). Additionally, the sensitivity range of V_th is between 1.6 mV/°C and 1.9 mV/°C, which is smaller than that of the Si-based device [43]. The fitting coefficient of V_th is weaker than that of V_ds.

To explore the influence of I_m on the temperature-sensitive characteristics of V_th, I_m is set to 5 mA, 10 mA, 15 mA, and 20 mA, respectively. The V_th-T_j curves under different I_m are shown in Figure 12. It can be seen that V_th rises with I_m and exhibits a level of discreteness in contrast to V_ds.

4.3. Characteristics of Body-like Diode Voltage

Based on the measurement circuit shown in Figure 4, the constant current I_m is set to 100 mA, and the gate voltage V_gss is set to 0 V. The V_sd-T_j curves of the two GaN HEMT devices of model GS61008P are presented in Figure 13. It can be seen that V_sd of the body-like diode increases with the junction temperature. Based on (3), the threshold voltage and resistance, which have positive temperature characteristics, make V_sd also have positive temperature characteristics. Additionally, V_sd exhibits excellent linearity, whose fitting coefficient reaches approximately 0.98. The DUT’s V_sd has a temperature sensitivity of about 2 mV/°C, which is comparable to the body diode voltage drop sensitivity of 2.4 mV/°C for SiC MOSFET [37].

To explore the effect of I_m on the temperature sensitivity of the V_sd, V_gss is kept at 0 V, and I_m is set to several groups of values. The relationship between V_sd and junction temperature under different I_m is shown in Figure 14. It is evident that the V_sd maintains good linearity, and the initial value of V_sd has a slight increase while the sensitivity remains relatively constant as I_m increases. Within the set range of I_m, V_sd has a temperature sensitivity that remains around 2 mV/°C.

4.4. Stability of TSEPs’ Characteristics under Temperature Cycles

After subjecting the DUT to temperature cycling in the experiment, the TSEPs show parameter drift under identical test conditions, thus indicating inconsistent stability. The temperature cycle specifically refers to the use of natural cooling to cool down DUT after the test temperature reaches the maximum temperature, and then DUT is tested for the next cycle. To improve the precision of results during temperature cycles, multiple measurements are performed at each temperature point.

To explore the stability of TSEPs, Infineon’s product GaN HEMT IGT60R070D1 and Cree’s product SiC MOSFET CMF20120D are used in the experiment and compared with GaN System’s product GaN HEMT GS61008P used in the previous section.

The stability of three parameters is tested: saturation voltage with low current injection V_ds, threshold voltage V_th, and body-like diode voltage V_sd. For the stability of V_ds, each DUT is subjected to three temperature cycles and the test current of 300 mA is injected. Due to different material characteristics and voltage levels, the initial V_ds value of each device is also different.

Table 1. Initial TSEPs’ values measured at 25 °C.

	GaN HEMT GS61008P	GaN HEMT IGT60R070D1	SiC MOSFET CMF20120D
Vds (mV)	2.57	16.72	297
Vth (V)	2.09	1.47	5.22
Vsd (mV)	1.83	1.51	0.56

displays the V_ds, V_th, and V_sd of each device measured at 25 °C. V_ds of GS61008P is significantly smaller than that of IGT60R070D1 and CMF20120D. To better illustrate the TSEPs’ characteristics, the results are normalized as shown in Figure 15. The curve overlap of all devices is good during three temperature cycles, which indicates that V_ds has good stability and the impact of the temperature cycles on V_ds can be negligible. It is beneficial to the application of V_ds in junction temperature measurement. Furthermore, according to the experimental results, the sensitivity of the GaN HEMT device is significantly lower than that of SiC MOSFET. Therefore, utilizing V_ds as a TSEP is more challenging in the junction temperature extraction of GaN HEMT [22,39].

For the stability of the threshold voltage V_th, each device is subjected to three temperature cycles, and the test current of 10 mA is applied. The V_th of the three devices measured at the initial temperature of 25 °C is shown in Table 1. The normalized V_th of each device is shown in Figure 16. It can be seen that the V_th of GS61008P fluctuates significantly, while IGT60R070D1 and CMF20120D have better stability, which indicates that the stability of V_th needs to be considered when it is used as the TSEP of GaN HEMT. Additionally, the linearity of GS61008P is notably weaker compared to that of IGT60R070D1 and CMF20120D. In particular, for GaN HEMTs GS61008P and IGT60R070D1, the V_th of the two devices presents opposite temperature characteristics, which may be caused by the dominant degree of ϕ_b, ΔE_C, and σ of the two devices at different temperatures.

For the stability of the body-like diode V_sd, each device is subjected to three temperature cycles with the gate voltage set to 0 V and a test current of 100 mA. GaN HEMT exhibits a higher V_sd than SiC MOSFET at 25 °C, per Table 1. The normalized V_sd of each device is presented in Figure 17. It is evident that the V_sd of IGT60R070D1 and CMF20120D exhibits stability, whereas the V_sd of GS61008P has slight fluctuations, which are much smaller than those of the V_th in the same device. In addition, compared to the positive temperature characteristics of GS61008P, the V_sd of IGT60R070D1 as a GaN HEMT exhibits negative temperature characteristics, which are caused by the negative temperature characteristics of V_th based on Equation (3).

5. Conclusions

Currently, the junction temperature monitoring of GaN HEMT based on the TSEPs method remains in the exploratory stage. Research on the characteristics of steady-state TSEPs is beneficial to their application in device reliability. This article presents test guidelines for three TSEPs, saturation voltage with low current injection, threshold voltage, and body-like diode voltage, and focuses on investigating their stability and assessing their temperature-sensitive characteristics. The following conclusions can be drawn:

• The saturation voltage with low current injection as a TSEP still has good temperature-sensitive characteristics, showing good linearity and stability in GaN HEMT. Its sensitivity is influenced by both the injection current and gate voltage. Overall, it has potential value in the field of temperature measurement.
• The threshold and body-like diode voltage as TSEPs exhibit significant variations for different devices. In particular, for some GaN HEMT, the stability of the threshold voltage is not ideal, which needs to be considered in applications.
• Compared to Si and SiC devices, the sensitivity of TSEPs in GaN HEMT is generally lower, especially saturation voltage with low current injection, which poses higher challenges to their application in junction temperature monitoring.