2024_05_02_c3eb1310eb974b219274g

接受的手稿

IGBT功率模块集成均热板的新型热管理系统的传热和热应力研究

陈依依、李波、王鑫、严玉英、王彦刚、齐芳

个人身份信息：

S2451-9049(18)30233-6

DOI：https://doi.org/10.1016/j.tsep.2019.01.007

参考：TSEP 291

出现在：

热科学与工程进展

收到的日期：

2018年3月

修订日期：

2018年11月

接受日期：

2019年1月

引用本文为：Y. Chen，B. Li，X. Wang，Y. Yan，Y. Wang，F. Qi，IGBT功率模块集成均热板的新型热管理系统的传热和热应力研究，热科学与工程进展（2019），doi：https://doi.org/10.1016/j.tsep.2019.01.007

这是一份未经编辑的手稿的 PDF 文件，已被接受出版。作为对客户的一项服务，我们提供此手稿的早期版本。在以最终形式出版之前，手稿将经过文字编辑、排版和校样审查。请注意，在制作过程中可能会发现可能影响内容的错误，并且适用于期刊的所有法律免责声明均适用。

IGBT功率模块集成均热板的新型热管理系统的传热和热应力研究

陈依依, 李波, 王鑫, 严玉英, 王彦刚, 方琪诺丁汉大学工程学院，NG7 2RD 诺丁汉，英国戴尼克斯半导体有限公司，多丁顿路，林肯，LN6 3LF

抽象的

IGBT功率模块中的热应力会导致严重的热可靠性问题，例如模块变形、性能下降甚至永久性损坏。因此，开发创新、高效的IGBT冷却技术非常重要。本文开发了一种用于冷却 IGBT 功率模块的新型热管理系统。该模块与均热板散热器集成，可显着降低热阻并提高温度均匀性。通过 3D FEM 建模来研究均热板对功率循环下的温度分布、热应力、能量应变耗散密度和寿命的影响。仿真结果表明，所提出的热管理系统在冷却能力、热应力、蠕变和塑性应变能量耗散以及热疲劳寿命方面优于传统的冷却解决方案。对功率循环下焊料层失效机制的研究表明，蠕变导致的主要是功率循环中的损坏，并且热载荷引起的裂纹预计会在边缘开始。

关键词: IGBT电力电子, 电动汽车, 相变冷却, 均温板, 热阻, 热应力, 热疲劳

命名法

蠕变速率系数

横截面积

乙

1 /有效应力（1/MPa）

C 四阶弹性张量

热容

运动切线模量

K热导率（W/mK）

L 厚度（米）

希腊字母

下标

疲劳能量指数

加罗法隆参数

疲劳循环

传热率（W）

活化能 (J/mol)

热阻（K/W）

气体常数（J/mol.K）

温度

时间（秒）

疲劳能量系数

应力（帕）

运动硬化位移应力 (Pa)

温差

蠕变应变能密度（

）

非弹性应变能密度

塑性应变能密度

双点张量积

C	横截面
铬	蠕变
埃尔	弹性的
F	疲劳
在	塑料
特金	运动切线

一、简介

为了缓解全球化石能源危机和环境问题，人们为开发可持续和清洁能源做出了很多努力。混合动力汽车和电动汽车的普及将成为未来几十年不可阻挡的趋势。作为一种半导体开关器件，绝缘栅双极晶体管（IGBT）功率模块是功率转换和电机控制系统中最重要的组件之一。最近，IGBT 模块需要安装在引擎盖下或直接安装在发动机上。汽车额定功率的提高、结构的紧凑化以及高度集成化和小型化都需要IGBT功率在有限的空间内散发更多的热量。例如，在下一代混合动力电动汽车中，IGBT模块的热通量将从100增加到150

到

，但整个模块的温度必须保持低于最大结温

或者

。过高的温度和大的温度梯度会导致多种模块故障，包括焊料分层和裂纹、陶瓷和硅芯片裂纹扩展以及键合线剥离。因此，迫切需要开发先进的IGBT功率模块热管理系统。

采用风冷方式很难为大功率、热通量的IGBT模块提供足够的冷却，特别是当功耗超过约

[3]。微型、迷你通道和针翅式液体冷却解决方案通常可以提供更高的冷却能力

与空气冷却相比（通常

）[4]。奥贡塔拉等人。 [5]研究了颗粒沉积对电子元件多孔翅片散热器热性能的影响。古等人。 [6]从理论上研究了用于冷却三维电子电路的微通道的热性能。然而，流动重新分布导致温度分布不均匀。在射流冲击策略中，蒸发的冷却剂被喷射到平坦的表面上进行冷却。约尔格等人。 [7] 设计了一种直接单冲击射流，用于冷却 MOSFET 电力电子模块。传热速率可达到

. However, cooling performance rapidly degrades as it departs the central of the jet region, resulting in temperature non-uniformity across the surface being cooled [8]. Therefore, much attention has been paid to array impingement cooling to improve temperature uniformity as well as heat transfer coefficient. When water is used as the coolant, heat flux level of

and

can be obtained with single-phase and phase change impingement

. Although jet impingement
can achieve a very high cooling capacity, the applicability of this strategy is limited due to high cost, complex cooling flow redistribution, cooling loop leakage and channel blockage.

As super heat conductive devices with phase change, heat pipe has been applied widely to improve the cooling capacity of high power electronic module [10]. Avenas et al. [11]developed a power chip module with vapour chamber heat spread. Ivanova et al. [12] proposed that heat pipe can be directly integrated in DBC for cooling of power electronics packaging. Figure 1 illustrates the structures of three different thermal management for IGBT module. Case A represents conventional indirect cooling. The DBC layer is soldered on a copper baseplate first and then bolt to a heat sink with grease-based thermal interface material (TIM). TIM accounts for a great part of total thermal resistance so it becomes a dominating obstacle in improving cooling capacity. Case B shows a cross section of an assembly using direct attach method. DBC layer is directly soldered on a copper baseplate. Currently, this direct cooling method is most used due to its compact structure and avoiding using any kind of thermal interface material. Figure 1 (c) shows the novel design in this study. The DBC layer is directly soldered on a vapour chamber-based heat sink. Also, no thermal interfacial material is used and the structure becomes more compact.

(a) Case A

(b) Case B

Figure 1 Structure of IGBT (a) with traditional indirect cooling (b) with direct cooling attached with copper baseplate IGBT module (c) with direct cooling attached with vapour chamber baseplate

In this paper, finite element method (FEM) simulation is used to investigate thermalmechanical performance and reliability under power cycling. It allows us to estimate fatigue life which is a major challenge for the electronic area. In electronic power module system, near-eutectic

and

materials are widely used to joint semiconductor on substrate for providing thermal, mechanical and electrical connection. It is well known that solder layer is the weakest part and break of which leads to failure of electronic power module at high operation temperature. This is because the solders are prone to creep and the
accumulation of plastic strain triggers crack and propagation[13]. Moreover, Sn-3.5Ag solder has lower melting temperature and yield stress and higher coefficient of thermal expansion compared with other packaging layers, which will deteriorate mechanical performance of solder subjected to power cycling. In finite element simulation, a coupled thermo-mechanical model was applied in transient condition. Strain range and energy density are used as the failure parameters. An energy-based model is used to correlate crack initiation and propagation through solder from fracture mechanics approach. The simulation results reveal temperature, elastic-plastic and creep behaviour, deformation and lifetime prediction of Sn-3.5Ag solder joints exposed to power cycling condition.

2. IGBT module description

2.1 Module components

A schematic of the IGBT power module is given by figure 2. The module is composed of six pairs of IGBT chip and diode, Direct Bond Copper (DBC), solder, copper or vapour chamber. The DBC layer is made of two copper layers bonded on an aluminium oxidation

layer. The

functions as heat conductor and electrical insulator due to its good thermal conductivity and electrical insulation. The chip solder has the same size with that of IGBT chips and diodes. Table 1 summarizes sizes of main components and table 2 shows the material physical properties of each layer.

Figure 2 Schematic of the analysed power module

Table 1 Component sizes

Component	Dimensions
Silicon chips
Silicon diodes
DCB copper layer
DCB Aluminium oxidation

Table 2 Material property

THERMAL

CAPACITY

COEFFICIENT OF

THERMAL

EXPANSION

730

6.5

320

413

Vapour chamber

385

5000

904

237

23.1

Silicon

678

2.1

240

242

2.2 IGBT parameters

Power applied to the IGBT chips and diodes is

(high state) and

(low state) per cycle. As suggested by

[14], IGBT cycle is six times longer than diode power cycle in hybrid electrical vehicle application. Figure 3 shows power cycle applied to the IGBT module.

Figure 3 IGBT power loss cycle

2.3 Initial design of IGBT integreted with thermal management

Figure 1c and 4 show the novel design of IGBT module integrated with a vapour chamber. Different from traditional thermal management system, baseplate and thermal paste are not used. The Direct Copper Bond (DCB) layer is directly soldered on lid of vapour chamber. A water cooling plate is bolted to vapour chamber together with IGBT power module.

Figure 4 Explosion view of thermal management system

Figure 5 shows traditional IGBT module (case A) thermal resistance stack in IGBT module. Thermal resistance of each layer is descripted by:

where

is the thickness of material layer,

is the thermal conductivity and

is the crosssectional area. As shown in figure 5, thermal paste and cast cold plate contribute the major thermal resistance, which account for

and

of overall module respectively. Therefore, the removal of those should bring a significant decrease in total resistance.

Silicon = Chip solder = Copper = AlN = Copper = DBC Solder . Copper baseplate - Thermal paste = Cast Al cold plate

Figure 5 Approximate thermal resistances stack in traditional IGBT module[15]

3. Methodology

3.1 Finite element method (FEM) model

Firstly, a coupled thermo-mechanical model is applied in transient condition. The governing equations are given as,

Equation 2 is energy equation, where

is temperature and

is heat generation. Equation 3 is the Hooke's law relating the stress-stress of material, where

is stress tensor,

is strain tensor,

is the fourth order elasticity tensor and ":" stands for the double-dot tensor product[16].

In solid mechanics model, semiconductor silicon and aluminium oxide in DCB substrate are assumed to be a linear-elastic model. Copper is assumed to be elastic-plastic because the stress within DCB substrate exceeds its yield stress easily. The Sn-3.5Ag solder layer is assume to viscoplastic with properties of hardening plasticity and implicit creep as it is widely agreed that rate of dependent plasticity creep occurs in the solder over time[17]. To achieve such effect, Garofalo hyperbolic sine law is applied to the solder layer.

The

is the equivalent uniaxial creep strain,

is the equivalent tensile stress,

is absolute temperature,

is the gas constant;

are creep rate coefficient,

effective stress, Garofalo n parameter and activation energy respectively[18].

Also, the creep analysis uses an implicit time integration method, which includes temperature - dependent constants, together with modelling of creep and kinematics hardening plasticity. In this case, the total strain can be expressed as,

Kinematics hardening plastics model is given as,

is kinematic tangent modulus. Both kinematic and mixed hardening result in a so-called back stress or shift stress, which is a new stress level that is equally far from yielding in tension and compression. Mechanical properties of packaging material and chip solder material are listed in table 3.

Table 3 Viscoplastic properties of Sn-Ag solder

A
44100	0.005	4.2	44995	8.314

The thermal fatigue lifetime modules for prediction of lifetime of solder include stress, strain, energy and damaged based models. Energy based model are the most convenient and accurate
method due to the ability to obtain test conditions more precisely[19, 20]. Therefore, energy based model and Morrow model [21]are used to assess the lifetime of

solder layer under power cycle in this study. Fatigue damage parameters such as inelastic strain and inelastic strain energy density and energy-based model are expressed as follow.

and

are inelastic, plastic and creep strain energy density respectively.

and

are inelastic, plastic and creep strain respectively.

' is fatigue energy coefficient,

is fatigue cycles and

is fatigue energy exponent with value of -0.69 .

3.2 Boundary condition

Figure 6 Mesh detail of the module

The simulation is conducted by Comsol Multiphysics software[22]. It is found that the simulation results of temperature distribution keep almost unchanged after the density exceeds 222,646. Therefore, mesh elements of 222,646 is selected in this study. Figure 6 shows the detail of final mesh of module. The size of mesh ranged between

and

. Thin and weak components such as solder layers, chips and DBC are fined meshed whereas baseplate are coarse meshed.

In this thermo-mechanical coupling model, heat sources are generated from IGBTs and diodes, which is estimated by power loss of these two components. Power loss is mainly caused by conduction losses from chip and collector current and switching loss. Total power losses are applied to each IGBT and diode are

and

respectively. Also, convection heat transfer is added to this model. The ambient temperature is initially set to

and heat transfer coefficient of air is

. To analyse thermal performance of each module, the heat transfer coefficient of cooling solutions is set to

according to the experiment. Table 4
summarizes basic thermal parameters used in modelling. Solid mechanics is applied to evaluate displacement caused by thermal expansion and thermal stress. Only the copper baseplate of IGBT module and four bolt holes are fixed in all direction and four bolt holes are fixed while others displace freely.

Table 4 Boundary condition in IGBT modelling

Boundary condition	Value
Heat source from diode	per diode
Heat source from IGBT	per IGBT
Ambient air temperature
Coolant temperature
Heat transfer coefficient of air
Heat transfer coefficient of coolant

Two models are built. Table 5 gives the details of each model.

Table 5 Descriptions of IGBT integrated thermal management system models

CU	Case B: IGBT is integrated with copper substrate
VC	Case C: IGBT power module is integrated with vapour chamber

3.3 Validation of the simulation model

In this work, the model is validated by comparing the simulation results of the junction temperature of IGBT chips to those obtained from experiment at stage of IGBT chips power loss. Figure 7 shows the entire schematic diagram of experiment apparatus. Heat is directly given to each diode and IGBT chip by a copper block which contains six cartridge heaters. Thermal grease, TIG780-38 (thermal conductivity

) is applied to improve thermal contact the thermal contact between the heater and the IGBT chips and diodes. The heat block is covered with polyether ether ketone to minimize the heat loss. The heat is powered with six DC power supplies (30-05:HSPY). The baseplate of the vapour chamber is kept being cooled with a water-cooling plate connected to a chiller. Three T-type thermocouples are mounted and used to measure junction temperature. Data of the junction temperature is collected by data acquisition system (DataTaker DT 800). It is impractical to directly measure junction temperature of semiconductor module[23]. In this study, thermocouples are mounted as close as possible to IGBT chip to measure the junction temperature and positions are shown in the insert picture of figure 8 . Figure 8 shows that the simulated junction temperature is lower than the experimental results at all three positions, most likely due to variation in material properties.

In general, the simulation is in good agreement with the experiment and the average error of junction temperature is

Figure 7 Schematic diagram of experiment apparatus

Figure 8 Comparison of junction temperature for validation

4. Discussion and results

4.1 Temperature distribution

Chip temperature is a prerequisite for precursor extraction as it is related to many IGBT temperature dependent parameters and failure mechanisms[24]. Figure 9 and 10 show temperature distribution of IGBT module with copper and VC substrate respectively. At the stage of IGBT chip power loss, hot spots mainly locate on IGBT chips; at the other stage the
hot spots move to the diodes. The maximum temperature at the chip is reduced from

at the stage of chip power loss stage after the copper substrate is replaced by vapour chamber. It's because that by means of phase change, vapour chamber transfers heat much more efficiently than the copper substrate. It is also found that the temperature variance in one cycle of IGBT with copper substrate is greater than that of vapour chamber. At the stage of IGBT chip power loss, the temperature difference of IGBT with copper baseplate is

, but the temperature variance is only

when the vapour chamber is used. This wider temperature range can cause higher thermal stress, which may lead to deformation within each layer of IGBT module and eventually damage the thermal reliability of module.

(a) CU model at stage of IGBT chip power loss

(b) CU model at stage of diode power loss

(d) VC model at stage of diode power loss

Figure 10 Temperature distribution in IGBT module

High temperature uniformity is another important requirement for thermal management of power module. If a cooling device generates a low-uniformity cooling performance, large temperature variation in packaging components with different properties and size cause forces at the interfaces between adjacent materials[25]. Thus, high temperature nonuniformity within material layers is also responsible for the high stresses experienced by the packaging. This will impair the reliability and efficiency of the IGBT module. Therefore, thermal management
system is required to supply a uniform cooling so that the difference between the highest and lowest temperature of a chip,

is within a few degrees. Figure 11 shows the change of

of the hottest IGBT chip as a function of time within a power loss cycle. It is obvious that ambient temperature has less effect on temperature uniformity. The

is reduced from

after copper substrate is replaced by vapour chamber, which demonstrates vapour chamber can improve temperature uniformity of IGBT power module. This is because that with phase change, heat spreads from evaporation surface to the entire chamber rapidly and also thermal conductivity of vapour chamber is more than 10 times higher than copper.

Figure 11 Temperature difference

against time within one power loss cycle

4.2 Thermal stress

The mismatch between two adjacent layers of module is not uniform due to the coefficient of thermal expansions difference and mismatch of length of each layer. These generate thermal stress which leads to deformation, solder delamination and bond wire lift off. In this case, to identify the region with highest mechanical load, von mises stress is used for evaluating the thermal stress distribution of IGBT module. Figure 12 shows the von mises stress distribution of IGBT power module at heat transfer coefficient of

. In order to have a clear sight of thermal stress distribution in DBC, the left switch IGBT chips, diodes and solder layer are hided. DBC layer is subjected to highest von mises stress especially for ceramic layer as shown in figure 10. This is because the thermal expansion coefficient of copper is around 3 times higher than that of

, which is at the origin of thermal fatigue. IGBT with copper and vapour chamber substrates experience maximal stress of

and 233 MPa respectively, which both occurs at the edge of ceramic layer due to the distance to neutral point (DNP) effect
and geometrical singularity. This shows that there is a

reduction in thermal stress. The central region of DBC layers also suffer high stress and this is attributed to mismatch in length of DBC layer and chips and higher temperature at the central. Also, the stress is higher at the interface between DBC and IGBT chips than at the interface between diodes and DBC because of higher temperature gradient across IGBT chips. In particular, the highest thermal stress occurs at the edge of upper ceramic/copper interface. This indicates that the crack or peeling starts from the edge of ceramic layer and grow along the interface finally bifurcate and break ceramic layer, which is in agreement with other studies [2, 26-28]

Figure 12 Von mises stress distribution. Bottom bar: IGBT with copper substrate. Top bar:

IGBT with vapour chamber substrate

Sn-3.5Ag solder has lower melting temperature, thus yielding larger stress and higher coefficient of thermal expansion compared with other packaging layers, therefore it is more easily damaged. Figure 13 shows peaks of component stress in Y direction (SY, also referred as peeling stress) and first principal stress (S1) are quite close to chip solder layer. This indicates that the peeling stress is the dominant factor in inducing crack initiation or failure at solder interface[29]. There is an apparent increase in peeling stress and first principal stress at 70 minutes because of temperature variation. For IGBT with vapour chamber, the maximum principal stress is 47 MPa which are not larger than the fracture strength of solder layer, which is 57.6 MPa reported by Hwang and Vargas[30] .

Figure 13 Peek peel (SY) and first principal (S1) stresses of Sn-3.5Ag

Under the thermal stresses, each component of IGBT power module with copper or vapour chamber substrate have a specific amount of deformation and displacement relative to their original shape and position. Figure 14 shows total displacement and deformation of chip solder layer. As shown in figure 14b, 14c and 14d, the Sn-3.5Ag solder undergoes a displacement toward the edge of the IGBT power module and bend into a convex shape under tensile force generated by thermal stress. This is attributed to the fact that top surface of solder layer is hotter than bottom surface, which forced the chip solder layers to deform in that way. The maximum displacement of

and

occur at the solder which locates at the central for vapour chamber case and copper case respectively.

(a)Solder layer 1

(b) Solder layer 2

Figure 14 Total displacement of solder layer

4.3 Energy disspation density

Generally, with thermo-mechanical fatigue, Sn-3.5ag chip solder layers are subjected to creep and plastic which play significant roles in deformation behaviour of materials. As shown in figure 15,

continuiously increases from

and from

for copper case and vapour chamber case respectivily. It indicates that creep dissipation density drops around

when vapour chamber is integrated. As predicted by Morrow's life prediction model described in equation 10, the fatigue lifetime reduces when inelastic dissipation energy density increase. The

is about 2.6 and 3.7 times larger than

for both vapour chamber and copper case respectively. It is found that creep is principal damage in power cycling.

Figure 15 Energy dissipation within two thermal cycles

4.4 Fatigue

In consequence, the weakest part of the module, Sn-3.5Ag solder layer, can be predicted. There are several fatigue life prediction models such as strain, stress and energy-based fatigue models. Strain distribution is more complex because each location has different strain and energy dissipation. It is difficult to obtain a represent position[31]. Moreover, the stress or strain is not sufficient to identify the fatigue properties for some cases. To improve the accuracy, the cycle to failure is estimated by using energy-based model combining the effect of stress and strain into energy. In some researches [16, 32-34], only creep or plastic alone was considered. Based on the analysis of energy dissipation density, plastic and creep play significant in energy dissipation. Therefore, in this simulation, inelastic energy dissipation including creep and plastic energy is used to predict the life time. Figure 16 shows the cycles to failure of solder
layer. The minimum simulated number of cycles to failure of vapour chamber case is 2640 which is

longer than the lifetime of copper case with value of 2402 .

Figure 16 Cycle to failure of solder layer

5. Conclusion

A detailed 3-D FEM simulation of thermo-mechanical performance in high power IGBT modules is presented. This work also proposed a novel structure of IGBT power module integrated with vapour chamber and thermal management. The Direct Copper Bond (DCB) layer is directly soldered on lid of a vapour chamber. To evaluate the performance of whole IGBT module integrated with vapour chamber, a thermo-mechanical coupling cyclic model including creep and plastic constitutive equation is established. Based on simulation results, the junction temperature can maintain at a lower value and decrease from

at the stage of chip power loss stage comparing with that of IGBT with copper baseplate. IGBT module integrated with vapour chamber has a higher temperature uniformity in a single chip. The temperature difference in a chip decreases from

. In this study, thermomechanical performance is also studied. It helps us to have a good understanding of failure mechanism in such operating mode. The maximum thermal stress is decreased

compared with copper baseplate case. It is also found that the outermost solder layer always suffered the highest thermal stress due to its distance from neutral point effect and mismatch in length between solder and DBC substrate. Cracks induced by thermal loading are expected to initiate at the edge. The failure mechanism is also investigated by comparing with creep and
plastic strain energy dissipation density. Creep is principle damage in the thermal cycling. The lifetime of the solder joint is calculated by the Morrow's life prediction model. The minimum simulated number of cycles to failure of vapour chamber case is

longer than the lifetime of copper case.

6. Reference

Medjahed, H., P.-E. Vidal, and B. Nogarede, Thermo-mechanical stress of bonded wires used in high power modules with alternating and direct current modes. Microelectronics Reliability, 2012. 52(6): p. 1099-1104.
Dupont, L., Contribution à l'étude de la durée de vie des assemblages de puissance dans des environnements haute température et avec des cycles thermiques de grande amplitude. 2006, École normale supérieure de Cachan-ENS Cachan.
Kandlikar, S.G. and C.N. Hayner, Liquid Cooled Cold Plates for Industrial High-Power Electronic Devices-Thermal Design and Manufacturing Considerations. Heat Transfer Engineering, 2009. 30(12): p. 918-930.
Wang, P., P. McCluskey, and A. Bar-Cohen, Two-Phase Liquid Cooling for Thermal Management of IGBT Power Electronic Module. Journal of Electronic Packaging, 2013. 135(2): p. 021001.
Oguntala, G.A., et al., Effects of particles deposition on thermal performance of a convectiveradiative heat sink porous fin of an electronic component. Thermal Science and Engineering Progress, 2018. 6: p. 177-185.
Koo, J.-M., et al., Integrated microchannel cooling for three-dimensional electronic circuit architectures. Journal of heat transfer, 2005. 127(1): p. 49-58.
Jörg, J., et al., Direct single impinging jet cooling of a MOSFET power electronic module. IEEE Transactions on Power Electronics, 2017.
Bhunia, A. and C. Chen, On the scalability of liquid microjet array impingement cooling for large area systems. Journal of Heat Transfer, 2011. 133(6): p. 064501.
Mudawar, I., et al., Two-Phase Spray Cooling of Hybrid Vehicle Electronics. IEEE Transactions on Components and Packaging Technologies, 2009. 32(2): p. 501-512.
Zilio, C., et al., Active and passive cooling technologies for thermal management of avionics in helicopters: Loop heat pipes and mini-Vapor Cycle System. Thermal Science and Engineering Progress, 2018. 5: p. 107-116.
Avenas, Y., et al. On the use of flat heat pipes as thermal spreaders in power electronics cooling. in Power Electronics Specialists Conference, 2002. pesc 02. 2002 IEEE 33rd Annual. 2002. IEEE.
Ivanova, M., et al., Heat pipe integrated in direct bonded copper (DBC) technology for cooling of power electronics packaging. IEEE Transactions on Power electronics, 2006. 21(6): p. 15411547.
Li, J., P. Agyakwa, and C. Johnson, Effect of trace Al on growth rates of intermetallic compound layers between Sn-based solders and Cu substrate. Journal of Alloys and Compounds, 2012. 545: p. .
Xu, Z., et al. Si IGBT phase-leg module packaging and cooling design for operation at 200 °C in hybrid electrical vehicle applications. in 2012 Twenty-Seventh Annual IEEE Applied Power Electronics Conference and Exposition (APEC). 2012.
Remsburg, R., Direct Integration of IGBT Power Modules to Liquid Cooling Arrays.
Barbagallo, C., et al., Thermal fatigue life evaluation of solder joints in a multi-chip power module.
Wen, S.S. and G.-Q. Lu. Finite element modeling of heat transfer and thermal stresses for three-dimensional packaging of power electronics modules. in Power Electronics and Motion Control Conference, 2000. Proceedings. IPEMC 2000. The Third International. 2000. IEEE.
Wang, D., Y. Yuan, and L. Luo. Failure analysis of Sn-3.5 Ag solder joints for FCOB using 2D FEA model. in Electronic Packaging Technology & High Density Packaging (ICEPT-HDP), 2010 11th International Conference on. 2010. IEEE.
Busca, C. Modeling lifetime of high power IGBTs in wind power applications-An overview. in Industrial Electronics (ISIE), 2011 IEEE International Symposium on. 2011. IEEE.
Choi, U.-M., F. Blaabjerg, and K.-B. Lee, Study and handling methods of power IGBT module failures in power electronic converter systems. IEEE Transactions on Power Electronics, 2015. 30(5): p. 2517-2533.
Morrow, J.D., ASTM STP 378, 1964. Philadelphia: ASTM(Philadelphia: ASTM): p. 45.
COMSOL Multiphysics v. 5.2. COMSOL AB.
Avenas, Y., L. Dupont, and Z. Khatir, Temperature measurement of power semiconductor devices by thermo-sensitive electrical parameters-A review. IEEE Transactions on Power Electronics, 2012. 27(6): p. 3081-3092.
Ji, B., et al., In situ diagnostics and prognostics of wire bonding faults in IGBT modules for electric vehicle drives. IEEE Transactions on Power Electronics, 2013. 28(12): p. 5568-5577.
Yun, C.-S., et al. Static and dynamic thermal characteristics of IGBT power modules. in Power Semiconductor Devices and ICs, 1999. ISPSD'99. Proceedings., The 11th International Symposium on. 1999. IEEE.
Pietranico, S., et al., Thermal fatigue and failure of electronic power device substrates. International Journal of Fatigue, 2009. 31(11): p. 1911-1920.
Cotterell, B. and J. Rice, Slightly curved or kinked cracks. International journal of fracture, 1980. 16(2): p. 155-169.
Jouini, Z., Z. Valdez-Nava, and D. Malec, Failure Analysis of Ceramic Substrates Used in High Power IGBT Modules. Engineering, 2016. 8(09): p. 561.
Chong, D.Y.R., et al., Evaluation on Influencing Factors of Board-Level Drop Reliability for Chip Scale Packages (Fine-Pitch Ball Grid Array). IEEE Transactions on Advanced Packaging, 2008. 31(1): p. .
Hwang, J. and R. Vargas, Solder joint reliability-Can solder creep? Soldering & Surface Mount Technology, 1990. 2(2): p. 38-45.
Kim, I. and S.-B. Lee. Fatigue life evaluation of lead-free solder under thermal and mechanical loads. in Electronic Components and Technology Conference, 2007. ECTC'07. Proceedings. 57th. 2007. IEEE.
Syed, A. Accumulated creep strain and energy density based thermal fatigue life prediction models for SnAgCu solder joints. in Electronic Components and Technology Conference, 2004. Proceedings. 54th. 2004. IEEE.
Schubert, A., et al. Fatigue life models for and solder joints evaluated by experiments and simulation. in Electronic Components and Technology Conference, 2003. Proceedings. 53rd. 2003. IEEE.
Shi, X., et al., A modified energy-based low cycle fatigue model for eutectic solder alloy. Scripta Materialia, 1999. 41(3): p. 289-296.

Highlights

A novel structure of IGBT integrated with vapour chamber was developed.
Thermal and thermo-mechanical models were used for evaluating performances of IGBT integrated with vapour chamber.
Junction temperature reduced by and the temperature variance dropped from to .
The maximum thermal stress was decreased by .
The creep of the solders was the principle damage and cracks induced by thermal loading initiate at the edge.