Analysis of Temperature Cycling And Insulation Characteristics of High-Reliability IGBT Modules

IGBT is a new type of power semiconductor device that integrates the advantages of BJT (bipolar junction transistor) and MOSFET (metal-oxide-semiconductor field-effect transistor), featuring high voltage, large current, high input impedance, low drive power, and fast switching speed. It is in extremely high demand in applications such as missile servo motor control systems, laser weapons, and fighter jet flight control systems. Its reliability largely determines the reliability of the entire device. As the working voltage and current of IGBT increase and the chip size continuously decreases, the chip power density increases sharply, making heat dissipation and reliability the key issues that must be addressed.

Ceramic substrates are the most widely used key materials for IGBT modules, featuring excellent thermal conductivity, heat resistance, insulation properties, and low expansion coefficient, and are suitable for aluminum wire bonding. Ceramic copper-clad substrates consist of a metal circuit layer and a ceramic layer. Due to the significant thermal expansion difference between the ceramic and the metal, the thermal stress generated during use can cause the substrate to crack and fail. Cracks typically occur at areas of stress concentration or high strain in the material. After sufficient cycles, cracks initiate at the stress concentration or high strain areas of the material, and further expansion of the cracks occurs under the action of cyclic loads until the material completely fractures. Therefore, the study of the thermal cycling reliability of ceramic substrates is of great significance.

Current status of ceramic substrate materials

The DBC (ceramic copper-clad board) materials for IGBT modules mainly include three types: aluminum oxide ceramic substrate, aluminum nitride ceramic substrate, and silicon nitride ceramic substrate.

● Al2O3 is the most commonly used material, featuring excellent insulation, chemical stability, and mechanical properties. The process is relatively mature, the cost is low, but Al2O3 has a low thermal conductivity and its thermal expansion coefficient does not match well with that of semiconductor chips (typically Si has 2.8×10-6·K-1). It is suitable for medium and low-power IGBT modules.

● AlN has a high thermal conductivity, approximately 6 times that of Al2O3. Its thermal expansion coefficient is relatively compatible with semiconductor chips. However, it is difficult to directly cover copper on its surface, and the cost is about 4 times that of Al2O3. AlN may decompose into hydrated aluminum oxide at higher temperatures and in higher humidity. Its flexural strength and fracture toughness are relatively low, making it prone to cracking during the thermal cycling process after welding, which affects the reliability of the entire power module. It is suitable for high-power IGBT modules.

● The thermal expansion coefficient of Si₃N₄ matches the semiconductor chip best. Its mechanical properties are more than twice those of Al₂O₃ and AlN, its thermal conductivity is more than 2.5 times that of Al₂O₃, it has high temperature lightness, excellent thermal shock resistance, and its cost is about 2.5 times that of Al₂O₃. For high-power IGBT modules, silicon nitride is currently the optimal material.

Temperature Cycling Test

To ensure that the IGBT meets the JM2 level assessment requirements, it is necessary to evaluate the reliability of the IGBT. Currently, the commonly used method is the temperature cycling test. The IGBT is heated and cooled as a whole using a temperature shock test chamber, causing temperature changes throughout the module. According to the requirements of GJB128A "Test Methods for Semiconductor Discrete Devices", the temperature range is -55 to 150℃, the transfer time should not exceed 1 minute, and the holding time should not be less than 10 minutes. In IEC60749-25 "Mechanical and Climate Variation Tests for Semiconductor Devices - Part 25: Temperature Cycling", the holding time should be ≥ 15 minutes. Therefore, the holding time of the temperature cycling test has been extended to 30 minutes to verify the reliability of the ceramic substrate.

① Test Objective

To determine the ability of IGBT to withstand extreme high and low temperatures, and the impact of alternating exposure to such extreme temperatures on the insulation withstand voltage of IGBT. Also, to investigate the corresponding failure phenomena of IGBT over time under harsh usage and storage conditions.

② Test Conditions

Before the test, it is necessary to confirm that the temperature box of the equipment is within the calibration validity period to ensure the validity of the test results. The placement of the module should ensure that it does not obstruct the air flow within the test chamber. The test temperature conditions are -55 to 150℃, with a holding time of 30 minutes, a total of 1000 cycles. The transfer time between the hot zone and the cold zone should not exceed 1 minute. After the test, the insulation withstand voltage test should be conducted within 8 hours to be effective.

③ Failure Mechanism

The ceramic substrate is a dual-material three-layer composite structure composed of copper-ceramic-copper. During the temperature cycling test, when the substrate as a whole uniformly undergoes temperature loads that vary over time, due to the mismatch in thermal expansion coefficients between copper and ceramics and the presence of deformation constraints, stress concentration occurs at the interface, especially at the geometric abrupt changes (commonly referred to as singular points).

When the external temperature load reaches 150℃, the copper layer on the ceramic substrate will undergo plastic deformation. During the temperature cycling process, the plastic deformation of the copper layer accumulates significantly, and stress concentration occurs at the geometric abrupt change point of the copper layer and the ceramic interface. Due to the relatively weak stress singularity at the interface end, when stress concentration occurs at the interface, the failure of the bonding material will start from the position of the stress concentration, thereby generating cracks. At the same time, during the manufacturing process of the ceramic substrate, there is a significant difference from 1066℃ to room temperature, and the substrate has certain residual stress. This will cause the crack to deviate from the original crack direction and extend into the ceramic matrix, resulting in failure. In addition, the ceramic is formed by powder sintering, and there are extremely tiny cracks or voids as inherent defects. These inherent defects will also act as weak points of the ceramic matrix and induce the crack to extend in the direction of the defect. After the crack extends a certain length, it continues to expand in the direction parallel to the interface, ultimately leading to the complete fracture of the substrate.

Twenty IGBT modules with AlN, Si3N4, Al2O3 as the ceramic substrates and 9% zirconium added to Al2O3 were used for 500 cycles (JM2 level) and 1000 cycles (JM3 level) of temperature cycling tests. Before the tests, insulation withstand voltage tests were conducted on the modules. The insulation withstand voltage test was carried out at the 100th cycle, and then every 50 cycles after that until the 1000th cycle.

The AlN substrate had 1 module insulation withstand voltage failure on the 200th occasion, 2 module insulation withstand voltage failures on the 250th occasion, and another 2 module insulation withstand voltage failures on the 300th occasion. As a result, all 5 modules had insulation withstand voltage failures. On the 500th occasion, 3 Al2O3 module insulation withstand voltages were also found to be failure. After 1000 temperature cycles， the insulation withstand voltages of Si3N4 and Al2O3 (doped with 9% zirconium) ceramic substrates were all qualified. This proves the rationality of the theoretical analysis of crack propagation in ceramic substrates. The reliability of AlN is inferior to that of Si3N4 and Al2O3， and the reliability of Al2O3 is inferior to that of Si3N4.

Simulation

Taking the 650V/200A IGBT module as the research object, the steady-state temperature field of different ceramic substrates was simulated using ANSYS finite element method. The thermal resistances of different substrates were compared to provide the best heat conduction solution.

Under the same power and heat exchange conditions, the highest steady-state working temperature of the Al2O3 IGBT module was 125.39℃, corresponding to a bottom temperature of 103.00℃, and the thermal resistance was 0.022℃/W. The highest steady-state working temperature of the FRD chip was 89.95℃, corresponding to a bottom temperature of 65.21℃, and the thermal resistance was 0.049℃/W.

Steady-state operating temperature distribution of alumina IGBT chips and FRD chips

Under the same conditions of power and heat exchange, the maximum stable operating temperature of the IGBT module using Si3N4 is 117.75℃, with the bottom temperature being 104.74℃ and the thermal resistance being 0.013℃/W. The maximum stable operating temperature of the FRD chip is 82.08℃, with the bottom temperature being 64.65℃ and the thermal resistance being 0.036℃/W.

Steady-state operating temperature distribution of silicon nitride IGBT chips and FRD chips

Under the same power and heat exchange conditions, the maximum steady-state operating temperature of the IGBT module using AlN is 116.76℃, corresponding to a bottom temperature of 101.10℃ and a thermal resistance of 0.015℃/W. The maximum steady-state operating temperature of the FRD chip is 80.93℃, corresponding to a bottom temperature of 63.82℃ and a thermal resistance of 0.034℃/W.

The comparison of IGBT structures and thermal resistances of different ceramic materials shows that the thermal resistances of AlN and Si3N4 are comparable, while the thermal conductivity of Al2O3 is relatively poor and its thermal resistance value is higher.

Conclusion

This paper conducted a temperature cycling test using a 650V/200A IGBT module. Based on the test results, the following conclusions were drawn:

● The failure of the ceramic substrate occurred at the edge of the substrate near the soldering interface on the ceramic side.

● Considering that the actual processing of aluminum nitride substrates is twice as thick as that of silicon nitride substrates, a model with different DBC thermal resistances was established using ANSYS finite element method. The calculation results indicated that the thermal resistance of the aluminum nitride substrate was consistent with that of the silicon nitride substrate.

● The performance of the silicon nitride ceramic copper-clad laminate is the best. High-reliability IGBT modules should use silicon nitride as the substrate material.