Measurement of Heat Dissipation and Thermal-Stability of Power Modules on DBC Substrates with Various Ceramics by SiC Micro-Heater Chip System and Ag Sinter Joining

This study introduced the SiC micro-heater chip as a novel thermal evaluation device for next-generation power modules and to evaluate the heat resistant performance of direct bonded copper (DBC) substrate with aluminum nitride (AlN-DBC), aluminum oxide (DBC-Al2O3) and silicon nitride (Si3N4-DBC) ceramics middle layer. The SiC micro-heater chips were structurally sound bonded on the two types of DBC substrates by Ag sinter paste and Au wire was used to interconnect the SiC and DBC substrate. The SiC micro-heater chip power modules were fixed on a water-cooling plate by a thermal interface material (TIM), a steady-state thermal resistance measurement and a power cycling test were successfully conducted. As a result, the thermal resistance of the SiC micro-heater chip power modules on the DBC-Al2O3 substrate at power over 200 W was about twice higher than DBC-Si3N4 and also higher than DBC-AlN. In addition, during the power cycle test, DBC-Al2O3 was stopped after 1000 cycles due to Pt heater pattern line was partially broken induced by the excessive rise in thermal resistance, but DBC-Si3N4 and DBC-AlN specimens were subjected to more than 20,000 cycles and not noticeable physical failure was found in both of the SiC chip and DBC substrates by a x-ray observation. The results indicated that AlN-DBC can be as an optimization substrate for the best heat dissipation/durability in wide band-gap (WBG) power devices. Our results provide an important index for industries demanding higher power and temperature power electronics.


Introduction
Silicon carbide (SiC) and gallium nitride (GaN), are wide band-gap (WBG) semiconductors, are strongly recognized as the best materials for the power electronics applications demanding higher power and higher temperature [1,2]. Since these WBG semiconductors have superior properties, kind In this study, SiC micro-heater chip was designed and used to bond with the various type of ceramic DBC substrates such as AlN and Al2O3, and Si3N4. The Ag sinter joining was used as a die to attach material because which can withstand the high temperature applications and reported in many previous studies. Steady-state thermal resistance from the SiC micro-heater chip to cooling system with different type DBC were measured. In addition, the power cycling test also was performed to investigated the high temperature reliability of the SiC micro-heater chip die attach structure on each type DBC substrate. The failure after power cycling was analyzed by a micro-focus 3D computed tomography (CT) X-Ray system. This method can significantly distinguish with traditional thermal cycling test because thermal properties of material can be considered during repetitive thermal environments. Our method of power cycling test with SiC heater chip may provide in important evaluation index of future packaging systems with WBG semiconductors targeted for hightemperature/power density power modules.

SiC Micro-Heater Chip and Direct Bonded Copper (DBC) Substrate
Geometrically 3D designed SiC micro-heater chip was fabricated with n-doped 4H-SiC by using a lithography technology (see Figure 2a), also it consists of a heater, a probe and an electrode, the material of heater and probe are platinum (Pt). The DBC substrate with five lands of the top side was introduced, and the SiC heater chip was die attached onto the center land and its geometric dimension is as displayed in Figure 2b. At the backside of SiC micro-heater chip and the top side of DBC substrates were metallized in the order of Ti and Ag layer by a sputtering process.  In this study, SiC micro-heater chip was designed and used to bond with the various type of ceramic DBC substrates such as AlN and Al 2 O 3, and Si 3 N 4 . The Ag sinter joining was used as a die to attach material because which can withstand the high temperature applications and reported in many previous studies. Steady-state thermal resistance from the SiC micro-heater chip to cooling system with different type DBC were measured. In addition, the power cycling test also was performed to investigated the high temperature reliability of the SiC micro-heater chip die attach structure on each type DBC substrate. The failure after power cycling was analyzed by a micro-focus 3D computed tomography (CT) X-Ray system. This method can significantly distinguish with traditional thermal cycling test because thermal properties of material can be considered during repetitive thermal environments. Our method of power cycling test with SiC heater chip may provide in important evaluation index of future packaging systems with WBG semiconductors targeted for high-temperature/power density power modules.

SiC Micro-Heater Chip and Direct Bonded Copper (DBC) Substrate
Geometrically 3D designed SiC micro-heater chip was fabricated with n-doped 4H-SiC by using a lithography technology (see Figure 2a), also it consists of a heater, a probe and an electrode, the material of heater and probe are platinum (Pt). The DBC substrate with five lands of the top side was introduced, and the SiC heater chip was die attached onto the center land and its geometric dimension is as displayed in Figure 2b. At the backside of SiC micro-heater chip and the top side of DBC substrates were metallized in the order of Ti and Ag layer by a sputtering process. In this study, SiC micro-heater chip was designed and used to bond with the various type of ceramic DBC substrates such as AlN and Al2O3, and Si3N4. The Ag sinter joining was used as a die to attach material because which can withstand the high temperature applications and reported in many previous studies. Steady-state thermal resistance from the SiC micro-heater chip to cooling system with different type DBC were measured. In addition, the power cycling test also was performed to investigated the high temperature reliability of the SiC micro-heater chip die attach structure on each type DBC substrate. The failure after power cycling was analyzed by a micro-focus 3D computed tomography (CT) X-Ray system. This method can significantly distinguish with traditional thermal cycling test because thermal properties of material can be considered during repetitive thermal environments. Our method of power cycling test with SiC heater chip may provide in important evaluation index of future packaging systems with WBG semiconductors targeted for hightemperature/power density power modules.

SiC Micro-Heater Chip and Direct Bonded Copper (DBC) Substrate
Geometrically 3D designed SiC micro-heater chip was fabricated with n-doped 4H-SiC by using a lithography technology (see Figure 2a), also it consists of a heater, a probe and an electrode, the material of heater and probe are platinum (Pt). The DBC substrate with five lands of the top side was introduced, and the SiC heater chip was die attached onto the center land and its geometric dimension is as displayed in Figure 2b. At the backside of SiC micro-heater chip and the top side of DBC substrates were metallized in the order of Ti and Ag layer by a sputtering process.

Die Attach Material
Two types of Ag particles were mixed and used to make Ag paste. The first type comprised flake-shaped Ag particles with an average lateral diameter of 8 µm and a thickness of 260 nm, as shown in Figure 3a (AgC-239, Fukuda Metal Foil and Powder Co., Ltd., Kyoto, Japan). The second type comprised spherical-shape particles with an average diameter of 400 nm, as shown in Figure 3b (FHD, MitsuiMining and Smelting Co., Ltd., Tokyo, Japan). The two kinds of Ag particles were mixed as the Ag filler with a weight ratio of 1:1 [24]. These hybrid particles were then stirred magnetically for 10 min and vibrated ultrasonically for 30 min with viscosity of 100-200 Pa·s solvent (alkylene glycols and polyols) using a hybrid mixer (HM-500, Keyence Corporation, Osaka, Japan) to fabricate Ag paste to achieve a uniform mixture. The amount of the solvent was maintained at approximately 10 wt % in the paste to keep a suitable viscosity of 150-250 centi poise (cPs) at room temperature. The Ag paste was then printed onto the center land of the DBC substrates which were purchased from Mitsubishi Materials Corporation. For the bonding process, a step-by-step profile with low pressure (0.4 MPa), including 120 • C and 180 • C for 5 min as pre-heating, and then 250 • C for 60 min, was applied to prevent pores formation caused by evaporation of organic solvent (see Figure 3b). Finally, the sintered cross section is formed as shown in Figure

Die Attach Material
Two types of Ag particles were mixed and used to make Ag paste. The first type comprised flake-shaped Ag particles with an average lateral diameter of 8 μm and a thickness of 260 nm, as shown in Figure. 3a (AgC-239, Fukuda Metal Foil and Powder Co., Ltd., Kyoto, Japan). The second type comprised spherical-shape particles with an average diameter of 400 nm, as shown in Figure 3b (FHD, MitsuiMining and Smelting Co., Ltd., Tokyo, Japan). The two kinds of Ag particles were mixed as the Ag filler with a weight ratio of 1:1 [24]. These hybrid particles were then stirred magnetically for 10 min and vibrated ultrasonically for 30 min with viscosity of 100-200 Pa·s solvent (alkylene glycols and polyols) using a hybrid mixer (HM-500, Keyence Corporation, Osaka, Japan) to fabricate Ag paste to achieve a uniform mixture. The amount of the solvent was maintained at approximately 10 wt % in the paste to keep a suitable viscosity of 150-250 cps at room temperature. The Ag paste was then printed onto the center land of the DBC substrates which were purchased from Mitsubishi Materials Corporation. For the bonding process, a step-by-step profile with low pressure (0.4 MPa), including 120 °C and 180 °C for 5 min as pre-heating, and then 250 °C for 60 min, was applied to prevent pores formation caused by evaporation of organic solvent (see Figure 3b). Finally, the sintered cross section is formed as shown in Figure 3c. For power supply, 10EA Au wires of 45 μm diameter (TANAKA DENSHI KOGYO, Yoshida, Japan) on the four electrode were each interconnected by an ultrasonic wire bonding machine (Kulicke & Soffa Industries Inc, 4500 digital series, Singapore).  Figure 4 shows the machine of power cycling test and thermal resistance measurement and its component. The machine consists of four parts, power supplier, controller, monitor and watercooling system. Over 200 W of power is supplied from the power source to the electrode, and when heat is generated from the top of the heater chip, heat is transferred to the heat sink (25 °C) through the die attach and the DBC substrate. At this time, SiC heater/DBC die-attached specimen is mounted on the water cooling plate by diamond thermal grease (thermal interface material, TIM). To prevent delamination between specimen and cooling plate during power cycling, the DBC substrate was fixed to the water-cooling plate with a force of 20 N. Then, ∆T is calculated by the temperature of the top surface of the SiC heater chip (Tchip) and the temperature of the thermocouple attached to the watercooling plate (Tthermocouple), the equation is as follows:  Figure 4 shows the machine of power cycling test and thermal resistance measurement and its component. The machine consists of four parts, power supplier, controller, monitor and water-cooling system. Over 200 W of power is supplied from the power source to the electrode, and when heat is generated from the top of the heater chip, heat is transferred to the heat sink (25 • C) through the die attach and the DBC substrate. At this time, SiC heater/DBC die-attached specimen is mounted on the water cooling plate by diamond thermal grease (thermal interface material, TIM). To prevent delamination between specimen and cooling plate during power cycling, the DBC substrate was fixed to the water-cooling plate with a force of 20 N. Then, ∆T is calculated by the temperature of the top surface of the SiC heater chip (T chip ) and the temperature of the thermocouple attached to the water-cooling plate (T thermocouple ), the equation is as follows:

Test Machine and Method
Here, the thermal resistance (R th ) can be calculated by dividing this temperature difference (∆T) by the input power (Q) as follows: where, Q (W) is Joule heating by the electrical power of SiC micro-heater chip and T is the temperature (K).
Micromachines 2018, 9, x 5 of 10 Here, the thermal resistance (Rth) can be calculated by dividing this temperature difference (∆T) by the input power (Q) as follows: where, Q (W) is Joule heating by the electrical power of SiC micro-heater chip and T is the temperature (K).

Steady-State Thermal Resistance
For thermal resistances measurement, a specimen was packaged by the method introduced in the previous section (see Figure 3). After the die bonding, the specimen interconnected without noticeable defects such as oxidation, and its actual manufactured and interconnected SiC heater die attached specimen are shown in Figure 5a. Furthermore, the electrode and temperature sensor of SiC heater chip was Au wire bonded with the DBC substrate after die-attach bonding. The Au wire bonded with the DBC substrate was given at upper left corner in Figure 5a. In addition, an enlarged image of the packaged SiC micro-heater chip circuit is shown in Figure 5b. When the power is supplied, the heat transfer path is shown in Figure 6a, where each layer has a thermal resistance. Here, since all materials except ceramic are the same, input power dependent thermal resistance performance can be directly compared according to ceramic type. The results of thermal resistance performance are shown in Figure 6b. The thermal resistance of the SiC micro-heater chip power modules on the DBC-Al2O3 substrate at power over 200 W was about twice higher than DBC-Si3N4 and also higher than DBC-AlN. The thermal resistance of DBC-AlN was lower than that of DBC-Si3N4. In addition, thermal resistance of the SiC micro-heater chip power modules on the DBC-Al2O3 substrate tended to increase thermal resistance significantly with increasing power but there is almost not so much change for the DBC-AlN and DBC-Si3N4 with the increased power. The reason of the lowest thermal resistance of AlN is that the thermal conductivity of AlN (285 W·m −1 ·K −1 , for single crystals) is significantly higher than Al2O3 (25 W·m −1 ·K −1 ) and Si3N4 (60-90 W·m −1 ·K −1 ) [21,25]. Therefore, AlN-DBC can be considered to have the best heat dissipation performance of an excellent power module when applied properly in the area where heat dissipation is intensively required.

Steady-State Thermal Resistance
For thermal resistances measurement, a specimen was packaged by the method introduced in the previous section (see Figure 3). After the die bonding, the specimen interconnected without noticeable defects such as oxidation, and its actual manufactured and interconnected SiC heater die attached specimen are shown in Figure 5a. Furthermore, the electrode and temperature sensor of SiC heater chip was Au wire bonded with the DBC substrate after die-attach bonding. The Au wire bonded with the DBC substrate was given at upper left corner in Figure 5a. In addition, an enlarged image of the packaged SiC micro-heater chip circuit is shown in Figure 5b. When the power is supplied, the heat transfer path is shown in Figure 6a, where each layer has a thermal resistance. Here, since all materials except ceramic are the same, input power dependent thermal resistance performance can be directly compared according to ceramic type. The results of thermal resistance performance are shown in Figure 6b. The thermal resistance of the SiC micro-heater chip power modules on the DBC-Al 2 O 3 substrate at power over 200 W was about twice higher than DBC-Si 3 N 4 and also higher than DBC-AlN. The thermal resistance of DBC-AlN was lower than that of DBC-Si 3 N 4 . In addition, thermal resistance of the SiC micro-heater chip power modules on the DBC-Al 2 O 3 substrate tended to increase thermal resistance significantly with increasing power but there is almost not so much change for the DBC-AlN and DBC-Si 3 N 4 with the increased power. The reason of the lowest thermal resistance of AlN is that the thermal conductivity of AlN (285 W·m −1 ·K −1 , for single crystals) is significantly higher than Al 2 O 3 (25 W·m −1 ·K −1 ) and Si 3 N 4 (60-90 W·m −1 ·K −1 ) [21,25]. Therefore, AlN-DBC can be considered to have the best heat dissipation performance of an excellent power module when applied properly in the area where heat dissipation is intensively required.

Power Cycling Test
For the power cycling test, the temperature of the cooling plate was constantly controlled to be 50 °C. The water flow rate was 5.4 L/min. The constant current for heating were tested, and constant current is necessary to suppress the surge spike when heating was started. The typical temperature profile of the power cycling is plotted in Figure 7. The input electric current in the test was 2.1 A. The condition of power cycling test was applied 2 s ON and 5 s OFF at the input power of 200 W.
The results indicated that Al2O3 was stopped during the power cycle test due to excessive rise in thermal resistance before 1000 cycles, Si3N4 and AlN specimens were each subjected to more than 20,000 cycles of power cycle tests. In the case of Al2O3 stopping, die-attach and ceramic component were not any degradation but the Pt heater pattern line was partially broken as shown in Figure 8a due to the poor heat dissipation performance of Al2O3. The Pt heater pattern line accumulated a lot of heat, leading to the temperature increase significantly, leading to the film of pattern line partially melted as shown in Figure 8b.

Power Cycling Test
For the power cycling test, the temperature of the cooling plate was constantly controlled to be 50 °C. The water flow rate was 5.4 L/min. The constant current for heating were tested, and constant current is necessary to suppress the surge spike when heating was started. The typical temperature profile of the power cycling is plotted in Figure 7. The input electric current in the test was 2.1 A. The condition of power cycling test was applied 2 s ON and 5 s OFF at the input power of 200 W.
The results indicated that Al2O3 was stopped during the power cycle test due to excessive rise in thermal resistance before 1000 cycles, Si3N4 and AlN specimens were each subjected to more than 20,000 cycles of power cycle tests. In the case of Al2O3 stopping, die-attach and ceramic component were not any degradation but the Pt heater pattern line was partially broken as shown in Figure 8a due to the poor heat dissipation performance of Al2O3. The Pt heater pattern line accumulated a lot of heat, leading to the temperature increase significantly, leading to the film of pattern line partially melted as shown in Figure 8b. Figure 9a shows temperature swing behavior of the Si3N4-DBC substrate, the heater temperature at the beginning was about 240 °C, and when it reached 20,000 cycles, the temperature was about

Power Cycling Test
For the power cycling test, the temperature of the cooling plate was constantly controlled to be 50 • C. The water flow rate was 5.4 L/min. The constant current for heating were tested, and constant current is necessary to suppress the surge spike when heating was started. The typical temperature profile of the power cycling is plotted in Figure 7. The input electric current in the test was 2.1 A. The condition of power cycling test was applied 2 s ON and 5 s OFF at the input power of 200 W.
The results indicated that Al 2 O 3 was stopped during the power cycle test due to excessive rise in thermal resistance before 1000 cycles, Si 3 N 4 and AlN specimens were each subjected to more than 20,000 cycles of power cycle tests. In the case of Al 2 O 3 stopping, die-attach and ceramic component were not any degradation but the Pt heater pattern line was partially broken as shown in Figure 8a due to the poor heat dissipation performance of Al 2 O 3 . The Pt heater pattern line accumulated a lot of heat, leading to the temperature increase significantly, leading to the film of pattern line partially melted as shown in Figure 8b. Figure 9a shows temperature swing behavior of the Si 3 N 4 -DBC substrate, the heater temperature at the beginning was about 240 • C, and when it reached 20,000 cycles, the temperature was about 260 • C. In the case of AlN-DBC substrate, the starting temperature was about 220 • C, and the temperature after 20,000 cycles was about 280 • C (see Figure 9b). The temperature ranges of Si 3 N 4 and AlN are not significantly different, but the temperature change tendency was remarkable. Unlike Si 3 N 4 substrate, which has a small change in temperature, AlN was largely divided into two stages. The stage 1 is from the starting point to 4000 cycles, it reached up to saturation temperature. The stage 2 is from 4000 cycles to 20,000 cycles, at a relatively stable temperature, the temperature slowly reached its peak point. Si3N4 substrate, which has a small change in temperature, AlN was largely divided into two stages. The stage 1 is from the starting point to 4000 cycles, it reached up to saturation temperature. The stage 2 is from 4000 cycles to 20,000 cycles, at a relatively stable temperature, the temperature slowly reached its peak point.   Figure 9a shows temperature swing behavior of the Si3N4-DBC substrate, the heater temperature at the beginning was about 240 °C, and when it reached 20,000 cycles, the temperature was about 260 °C. In the case of AlN-DBC substrate, the starting temperature was about 220 °C, and the temperature after 20,000 cycles was about 280 °C (see Figure 9b). The temperature ranges of Si3N4 and AlN are not significantly different, but the temperature change tendency was remarkable. Unlike Si3N4 substrate, which has a small change in temperature, AlN was largely divided into two stages. The stage 1 is from the starting point to 4000 cycles, it reached up to saturation temperature. The stage 2 is from 4000 cycles to 20,000 cycles, at a relatively stable temperature, the temperature slowly reached its peak point.      To investigate a factor of different temperature behavior, a nondestructive test was performed by a micro-focus 3D computed tomography (CT) X-Ray system (XVA-160N, Uni-Hite System Corporation, Yamato City, Japan) on the specimens after power cycling tests. Figure 10 displays nondestructive test results (X-ray) after power cycling tests after power cycling tests (20,000 cycles). Figure 10a,b show the SiC micro-heater chip power modules on the DBC-Si 3 N 4 substrate and its high magnification image, respectively. Figure 10b,c the SiC micro-heater chip power modules on the DBC-Si 3 N 4 substrate of AlN and its high magnification image Figure 10d, respectively. There was no relationship between the different temperature behavior of the two substrates and the substantial damage to the die attach structure. Therefore, no delamination of the ceramic or degradation of the Ag sinter joint occurred during power cycling tests, and it was clearly found that the temperature rise did not cause physical failure. Despite the excellent thermal conductivity of AlN, the cause of temperature rise shown in Figure 9b can be regarded as an issue of the TIM. The mechanism of air entrainment into the TIM are shown in Figure 11. In the initial state, the DBC substrate and the water cooling plate were closely adhered by TIM (see Figure 11a), after experiencing expansion and shrinkage, there should be some gaps generation between the TIM and substrate, the adhesion strength changed to lower with a little gap generation as displayed in Figure 11b,c. Air entrainment into the TIM also can lead to the temperature swing increase. temperature rise shown in Figure 9b can be regarded as an issue of the TIM. The mechanism of air entrainment into the TIM are shown in Figure 11. In the initial state, the DBC substrate and the water cooling plate were closely adhered by TIM (see Figure 11a), after experiencing expansion and shrinkage, there should be some gaps generation between the TIM and substrate, the adhesion strength changed to lower with a little gap generation as displayed in Figure 11b,c. Air entrainment into the TIM also can lead to the temperature swing increase. Thus, to avoid this phenomenon, it can be prevented by using a soft TIM or warpage suppression mechanism that does not harden under repeated substrate warpage behavior. The results of the power cycling test using the SiC micro-heater chip show the heat resistance performance of the DBC substrate, are significantly different from the traditional thermal shock test. Consequently, the AlN-DBC substrates may have the best heat dissipation and durability, perhaps by optimizing TIM performance for WBG power conversion systems. Thus, to avoid this phenomenon, it can be prevented by using a soft TIM or warpage suppression mechanism that does not harden under repeated substrate warpage behavior. The results of the power cycling test using the SiC micro-heater chip show the heat resistance performance of the DBC substrate, are significantly different from the traditional thermal shock test. Consequently, the AlN-DBC substrates may have the best heat dissipation and durability, perhaps by optimizing TIM performance for WBG power conversion systems. Thus, to avoid this phenomenon, it can be prevented by using a soft TIM or warpage suppression mechanism that does not harden under repeated substrate warpage behavior. The results of the power cycling test using the SiC micro-heater chip show the heat resistance performance of the DBC substrate, are significantly different from the traditional thermal shock test. Consequently, the AlN-DBC substrates may have the best heat dissipation and durability, perhaps by optimizing TIM performance for WBG power conversion systems.

Conclusions
In this study, a power cycling test system based on a SiC micro-heater chip was developed to simulate the active SiC devices where the temperature of the heater chip can be recorded through the power cycling test in real-time. The new approach is useful for the evaluation of high-power devices with a similar condition to the real packaging system. The thermal resistance of of the SiC micro-heater chip power modules on DBC-AlN was lower than that of DBC-Al 2 O 3 and DBC-Si 3 N 4 . In addition, thermal resistance of the SiC micro-heater chip power modules on the DBC-Al 2 O 3 substrate tended to increase thermal resistance significantly with increasing power but there is almost not so much change for the DBC-AlN and DBC-Si 3 N 4 with the same increased power. Furthermore, no delamination of the ceramic or degradation of the Ag sinter joint occurred during power cycling tests for both DBC-AlN and DBC-Si 3 N 4 , and it was clearly found that the temperature rise did not cause physical failure. The results exhibit that both DBC-AlN and DBC-Si 3 N 4 are excellent candidates for high power modules with WBG semiconductor devices, suitable for industries demanding high energy efficiency, when the packaging like the die-attach and the cooling system is well optimized.