- Email: [email protected]
Optimization of the thermal contact resistance within press pack IGBTs
MR-12273; No of Pages 12 Microelectronics Reliability xxx (2017) xxx–xxx
Contents lists available at ScienceDirect
Microelectronics Reliability journal homepage: www.elsevier.com/locate/microrel
Optimization of the thermal contact resistance within press pack IGBTs Erping Deng a,b,⁎, Zhibin Zhao a, Peng Zhang b, Yongzhang Huang a, Jinyuan Li b a b
State Key Laboratory of Alternate Electrical Power System with Renewable Energy Sources, North China Electric Power University, Changping District, Beijing 102206, China State Grid Global Energy Interconnection Research Institute, Changping District, Beijing 102211, China
a r t i c l e
i n f o
Article history: Received 25 August 2016 Received in revised form 8 January 2017 Accepted 8 January 2017 Available online xxxx Keywords: Press pack IGBTs Thermal contact resistance Temperature Clamping force Nanosilver sintering technology
a b s t r a c t The research of thermal contact resistance between multi-layers within press pack IGBTs (PP IGBTs) is significant for optimizing the PP IGBTs' thermal resistance to improve reliability, as the thermal contact resistance accounts for approximately 50% of the total thermal resistance of PP IGBTs. In this paper, thermal contact resistance between multi-layers is analysed via a finite element model (FEM) of a single fast recovery diode (FRD) submodule. Most importantly, the influence of temperature and clamping force on the thermal contact resistance is also discussed, and findings are verified by submodule thermal resistance experiments. Based on the FEM and experimental results, nanosilver sintering technology is proposed to fill the gap between the contact interfaces to reduce thermal contact resistance. The fabrication of a sintered single FRD submodule is also investigated in this paper, and the results of the sintered sample indicate that the thermal resistance is reduced by approximately 18.8% compared to a direct contact sample. © 2017 Published by Elsevier Ltd.
1. Introduction To meet the growing requirements of applications of IGBT devices, capacity and reliability have become great challenges for IGBT devices. Press pack IGBTs (PP IGBTs) have gradually been applied to high-voltage and high-power-density applications, such as electric locomotives and HVDC transmission, because of their high reliability, double-side cooling, high power density and ease of laying out in series [1] compared to typical wire-bonded IGBT modules. The simplified internal structure of the studied PP IGBT is shown in Fig. 1, as it may be observed that the PP IGBT has a multi-layered structure. Two copper electrodes (the collector and emitter pole) provide the electrical and thermal paths for the silicon chips, and the silicon IGBT chips are sandwiched by two molybdenum plates, which help with the uniform distribution of the clamping force. A silver shim plate, together with a silicon IGBT chip and two molybdenum plates, is used to form a chip assembly. An external clamping force is required to maintain the electrical and thermal contact of all components within PP IGBTs. Thermal contact resistance is a very important parameter for thermal resistance optimization and reliability improvement, as it accounts for approximately 50% of the total thermal resistance [2]. When two rough surfaces are bought into contact, actual contact only occurs at certain discrete spots or micro-contacts, while the noncontacting areas form vacuums or are filled with some medium (such as air, water or oil). The actual contact area accounts for approximately 0.01%–0.1% of the nominal contact area, and the proportion only ⁎ Corresponding author at: No. 2, Beinong Road, Changping District, 102206 Beijing, China. E-mail address: [email protected] (E. Deng).
increases to 1%–2% under a contact pressure of 10 MPa [3]. Because the actual contact surface area is relatively small, as is the thermal conductivity of the interfacial gases, heat flow across the interface experiences a relatively large thermal resistance, commonly referred to as thermal contact resistance [4,5]. Thermal contact resistance is a complex parameter affected by the material properties, surface morphology, contact pressure, temperature and so on [6,7], and thus the precision of the test bench will be a significant challenge for measurement. At present, steady-state and transient methods are used to predict the thermal contact resistance, and the steady-state method is most commonly used. The temperature difference between the interfaces can be obtained by a linear fitting after the temperature has been measured between two samples in the steady-state method [8]. Not only will the temperature distribution of the specimen be disturbed by the embedded thermocouple, but the accuracy of the adjacent measured temperature will also be influenced since this steady method requires a thermocouple to measure the temperature. Most importantly, many thermocouples should be located in the samples, and thus the steady-state method is not suitable for a specimen whose geometry is small (in the millimetre ranger). To cover some disadvantages of the thermocouple method, an infrared imaging system with an accuracy of 0.1 °C, instead of a thermocouple, is put forward by various scholars to record the temperature of the two-dimensional interface. Although the accuracy is greatly improved, an error of approximately 23% is still presented [9]. The optical-thermal method is the widely used method among transient thermal contact resistance measurement methods [10,11]. Thermal contact resistance is obtained by the phase difference of the heat wave and modulation wave after encountering the interface. However, the accuracy of the optical-thermal method is affected by the interface characteristic, as the heat wave is diffused at the contact interface, and it
http://dx.doi.org/10.1016/j.microrel.2017.01.003 0026-2714/© 2017 Published by Elsevier Ltd.
Please cite this article as: E. Deng, et al., Optimization of the thermal contact resistance within press pack IGBTs, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.01.003
2
E. Deng et al. / Microelectronics Reliability xxx (2017) xxx–xxx
Fig. 1. A cross-section schematic diagram for press pack IGBTs.
destroys their phase relationship [12]. All the methods mentioned thus far are for the measurement of thermal contact resistance between power semiconductor devices and heatsinks or among copper bus bar junctions. Namely, all the methods are for the thermal contact resistance outside power semiconductors rather than within them. Although T. Poller et al. [13] measured the thermal contact resistance within PP IGBTs through a combination of experiments and a finite element method, the deviation of the thermocouple still exists, and the testing process is relatively complicated. Thermal contact resistance within PP IGBTs has not been identified for its special packaging style and working conditions likewise thyristors. And the measurement of thermal contact resistance within PP IGBTs is still impossible with the limits that no probes can be placed between two test samples and the clamping force system. Furthermore, the optical-thermal or other non-contact methods are not suitable for PP IGBTs because the chips are clamped and encapsulated. Thus, the theoretical model with the measured data to make it more precise is proposed to predict the behaviour of thermal contact resistance in this paper. Meanwhile, an indirect measurement of junction to case thermal resistance is proposed to predict the thermal contact resistance within PP IGBTs and to identify the theoretical values. The point of this indirect measurement is to use the variation of junction to case thermal resistance to deduce the variation of thermal contact resistance. According to the theoretical and experimental results, nanosilver sintering technology is introduced to reduce the thermal contact resistance, and a nanosilver-sintered sample is manufactured to verify this. The structure of this paper is set as follows. Characteristics of PP IGBTs and the status of thermal contact resistance calculation and measurement are presented in the introduction. The finite element model based on theoretical model is used to calculate the thermal contact resistance within PP IGBTs, as well as the influence of temperature and clamping force in the second section. And some parameters needed in the calculation
are measured through experiment to make the model more precise. A single fast recovery diode (FRD) chip submodule is fabricated to measure the thermal resistance change and compare it with the theoretical values in the experiment part. The optimization section presents a nanosilver sintering technology for optimizing the thermal contact resistance between the collector side of the chip and molybdenum, and a sintered sample is also manufactured to make a comparison with the direct contact submodule. Conclusions and outlook are provided in the last section. 2. Theoretical 2.1. Basic principle A micro-structure diagram of the contact interface is shown in Fig. 2 [14]; the contact spot is always named micro-contact, and the non-contact area is filled with air. Because the thermal conductivity of the interfacial gases is relatively low (0.023 W/(m·K)), heat flow across the interface experiences a relatively large thermal resistance, commonly referred to as thermal contact resistance. Micro-contacts, which have a relatively long distance among them, are general distributed randomly when two rough surfaces came into contact. Only thermal conduction is considered between the micro-contact and the interfacial gaps, which are usually filled with air in this situation [15]. In most practical situations concerning thermal contact resistance, the gap thickness between two contacting bodies is quite small (b10 μm), and thus the Grashof number based on the gap thickness is b 2000. Consequently, in most instances, the heat transfer of convection through the interstitial gas in the gap is neglected [16]. Thermal radiation across the interfacial gaps is generally considered as insignificant as the surface temperature is less than approximately 700 K [17]. In conclusion, the thermal contact conductance hc of the contact interfaces, which can be calculated by
Fig. 2. Micro-contact interface schematic diagram.
Please cite this article as: E. Deng, et al., Optimization of the thermal contact resistance within press pack IGBTs, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.01.003
E. Deng et al. / Microelectronics Reliability xxx (2017) xxx–xxx
3
Fig. 3. The roughness surfaces equivalent schematic diagram.
formula (1), consists of the micro-contact thermal contact conductance hs and interfacial gap thermal contact conductance hg. Furthermore, thermal contact resistance Rc can be obtained through formula (2), as the thermal contact conductance hc and nominal contact area A is given. hc ¼ hg þ hs
ð1Þ
Rc ¼ A hc
ð2Þ
2.1.1. Micro-contact conductance hs Williamson et al. [18] have shown experimentally that many of the techniques used to produce engineering surfaces give a Gaussian
distribution of surface heights. The contact between Gaussian rough surfaces can be considered as the contact between a single Gaussian surface, having the effective surface characteristics, placed in contact with a perfectly smooth flat surface, as shown in Fig. 3 [19]. The equivalent roughness and surface slope can be calculated from formula (3). σ¼
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi σ 21 þ σ 22 and m ¼ m21 þ m22
ð3Þ
where subscripts 1 and 2 represent the two contact surfaces, respectively. The surface roughness of the rough contact interface is largely influenced by the machining precision and should be measured via a surface
Fig. 4. Contact configurations.
Please cite this article as: E. Deng, et al., Optimization of the thermal contact resistance within press pack IGBTs, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.01.003
4
E. Deng et al. / Microelectronics Reliability xxx (2017) xxx–xxx
roughness tester. The surface slope m affected by the manufacture craft and surface roughness, etc., can be approximately predicted through a function of surface roughness with theoretical calculation or experimental data. The reason is that the surface topography is determined by the manufacture craft or different process [20] and four typical surfaces are shown in Fig. 4. Meanwhile, the surface roughness and slope is depended on the surface topography to a large extent as shown in Fig. 3. Correlations for surface slope are summarized in Table 1. As the measured surface roughness of those interfaces in PP IGBTs is much lower than 1.6 μm, the surface slope is identified by [22] shown in Table 1. Thermal resistance associated with micro-contact (known as thermal constriction/spreading resistance) is generally modelled by applying the so-called flux cube solution to each micro-contact with the assumption that the asperities undergo plastic deformation. According to the theory of Bahrami et al., thermal contact conductance of the contact interface can be calculated as long as a single micro-contact thermal model is given [24,25]. hs ¼ 1:25ks
m p 0:95 σ Hc
ð4Þ
where p is contact pressure and ks is the harmonic mean of solid thermal conductivity of the two bodies, which is shown in formula (5). ks ¼
2k1 k2 k1 þ k2
ð5Þ
Table 1 Correlations for surface slope. Reference
Correlation
Tanner [21] Antonetti et al. [22] Lambert [23]
m = 0.152σ0.4 m = 0.124σ0.743 , σ≤ 1.6 μm m = 0.076σ0.52
Different from the IGBT chip used in most wire bond packaging, the gate of the IGBT chip used in press pack packaging is located in the corner of the chip. Thus, grooves are shaped into the pedestals, as well as other components on the emitter side, for the gate pin of the housing. Meanwhile, a constant voltage is needed to apply to the IGBT chip's gate so the device channel is open and current flows through during the measurement. To exclude interference factors and reduce the complexity of the experiment, a single FRD chip submodule is designed in this paper to predict the behaviour of thermal contact resistance within PP IGBTs because the heat flux of IGBT chip and FRD chip are the same. The three-dimensional structure diagram and equivalent thermal network are shown in Fig. 6. Both sides of the FRD chip are covered with several μm of aluminium to realize the connection. The thermal contact resistance of the collector side and emitter side should be calculated separately because the surface roughness of the aluminium layer is quite different. According to the description above, a single FRD chip submodule contains four contact interfaces, which is shown in Fig. 6(a), and the material properties are listed in Table 2.
Hc is the surface microhardness of the softer of the two contacting solids. The microhardness is generally complex because it depends on several geometric and physical parameters, such as the Vickers microhardness correlation coefficients, surface roughness and contact pressure. As shown in formula (6), Song and Yovanovich related Hmic to the surface parameters and nominal pressure [26]. 0
1 11þ0:071c
p p B C ¼@ c2 A H mic c1 1:62 σσ0 m
2
ð6Þ
2.1.2. Micro-gap conductance hg Many theoretical models of the gap conductance hg are proposed by scholars, and the most commonly used is given by the approximation of Yovanovich [27], which is shown in (7). hg ¼
kg Y þM
ð7Þ
where kg is the thermal conductivity of the gap substance and the gas parameter M accounts for rarefaction effects at high temperatures and low gas pressures. The effective gap thickness Y, shown in Fig. 3, can be calculated accurately by means of the simple power-law correlation equation proposed by Antonetti and Yovanovich [28]. Y ¼ 1:53σ
p Hc
−0:097
ð8Þ
2.2. Finite element model The IGBT chip surface is quite rough from a microscopic view, as are other components within PP IGBTs. Fig. 5 shows a microscopic sectional view of the IGBT chip for press pack packaging. An aluminium thickness of several μm is needed to cover the chip surface to realize the connection with the external electrodes. As for the rough surfaces, the contact interface will introduce both electrical and thermal contact resistance.
Fig. 5. Micro cross-section schematic diagram of an IGBT chip.
Please cite this article as: E. Deng, et al., Optimization of the thermal contact resistance within press pack IGBTs, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.01.003
E. Deng et al. / Microelectronics Reliability xxx (2017) xxx–xxx
Only thermal conduction is considered in this paper according to the discussion mentioned before, and the basic differential equation is shown in (9), where ρ is the material density, Cp is the constant pressure specific heat, t is the time, k is the thermal conductivity and Q is the volumetric heat source. ρC p
∂T −∇ ðk∇T Þ ¼ Q ∂t
ð9Þ
5
The boundary conditions for the thermal model are set as follows. The temperatures of the collector and emitter pole outer faces (case temperature) have been set to 60 °C, with the assumption that the external water-cooling system can sustain a constant case temperature. The heat generation in the active area of the FRD chip is set as 100 W. The equivalent clamping force of the single FRD chip submodule is assumed to be 1 kN according to PP IGBTs' work condition (1.2 kN/cm2) and the contact area.
Fig. 6. Single FRD chip submodule.
Please cite this article as: E. Deng, et al., Optimization of the thermal contact resistance within press pack IGBTs, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.01.003
6
E. Deng et al. / Microelectronics Reliability xxx (2017) xxx–xxx
2.3. Results
Table 2 Material parameters.
Density [kg/m3] Thermal conductivity [W / (m ∗ K)] Specific heat [J / (kg ∗ K)] Elasticity modulus [GPa] Surface roughness [μm] Surface slope
Copper
Moly
Alum
8930 394 385 110 ※ ※
10,200 145 217 330 ※ ※
2700 210 900 68 ※ ※
where ※ denotes that the material property should be measured via experiment.
As shown in the thermal contact resistance calculation formulas, thermal contact resistance is mainly affected by surface roughness, surface slope, microhardness and contact pressure. In this paper, thermal contact resistance within all the interfaces are calculated through different interface parameters based on the formulas aforementioned and the interface between the collector side of the FRD chip and molybdenum is used as an example to show the influence of the previously mentioned parameters. Surface roughness ranges from 0.1 μm to 1 μm according to the manufacture craft for PP IGBTs. Surface slope ranges from 0.02 to 0.2
Fig. 7. Thermal contact resistance of the interface micro-contact.
Please cite this article as: E. Deng, et al., Optimization of the thermal contact resistance within press pack IGBTs, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.01.003
E. Deng et al. / Microelectronics Reliability xxx (2017) xxx–xxx
according to the surface roughness. Microhardness ranges from 1 GPa to 10 GPa according to material property, and the equivalent clamping force ranges from 200 N to 2000 N based on the clamping force of a single FRD chip within PP IGBTs. The calculation results are shown in Figs. 7 and 8. The unit of thermal contact resistance is K/W. It is observed that as expected, a higher contact pressure (equivalent clamping force) and a lower microhardness Hmic, which accelerate the plastic deformation of the asperities and the formation of micro-contacts with a larger contact surface area, give arise to a lower value of the thermal micro-contact resistance. Likewise, a decrease in the surface roughness and an increase in the roughness slope, which result in higher values of the micro-contact surface area, also give rise to a lower thermal micro-contact resistance. The micro-gap thermal
7
resistance is mainly affected by contact pressure, surface roughness and microhardness, with the exception of surface slope. 2.3.1. Influence of the temperature Temperature is the most important parameter affecting many material properties, such as microhardness and thermal conductivity. In general, thermal contact resistance between contact interfaces always decreases with the increase of temperature because a temperature increment, which lowers the microhardness (i.e., it softens), gives rise to a larger contact surface area. Microhardness changes with temperature eventually lead to a change of surface roughness. However, accurate measurement of microhardness is quite difficult, and the measurement would cause irreversible damage to the chip. In this paper, the influence
Fig. 8. Thermal contact resistance of the interface gap.
Please cite this article as: E. Deng, et al., Optimization of the thermal contact resistance within press pack IGBTs, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.01.003
8
E. Deng et al. / Microelectronics Reliability xxx (2017) xxx–xxx
Table 3 Comparison of the surface roughness. Surface roughness σ [μm]
Before
After
Deviation
Collector Collector-side molybdenum Chip collector side Chip emitter side Emitter-side molybdenum Emitter
0.226 0.344 0.495 0.147 0.334 0.638
0.151 0.268 0.147 0.057 0.307 0.617
33.18% 22.09% 70.30% 61.22% 8.08% 3.29%
of temperature on material properties, especially the surface properties, is represented by its final appearance (surface roughness decrement). Micro-contacts undergo an elasticity deformation or even a plastic deformation under the external clamping force as the microhardness is decreased with the augment of the temperature of the chip and other components during the thermal resistance test. Although the temperature of the chip and other components is reduced after the measurement is completed, the deformation is still unable to be fully restored to the state before measurement under the external clamping force and eventually leads to a decrease of surface roughness. The surface properties, such as surface roughness, would no longer change after several repeated thermal resistance measurements (equivalent to temperature cycling). The surface roughness (arithmetic average of absolute values Ra) of every contact surface is measured through a surface roughness tester (Mitutoyo SURFTEST SJ-310) after several repeated measurements, as shown in Table 3, and compared with the value before the thermal resistance test. It is observed that, as expected, the surface roughness of the components after testing is much smaller than before testing, especially for the components of the collector side. This is because the thermal resistance test only measures the junction to case thermal resistance of the collector-side cooling with the emitter-side adiabatic, and the heat flow only goes through the collector side. The reason for why we just measure the collector-side cooling is that thermal contact resistance under double-side cooling is the parallel relationship of the one under collector-side cooling and emitter-side cooling. So it is very difficult to distinguish the relationship of each interface and hard to verify the theoretical values. The calculated thermal contact resistance before and after the thermal resistance test based on the theoretical model mentioned before is shown in Table 4 with the clamping force of 1 kN.
Fig. 9. The relationship between the clamping force and thermal contact resistance between four interfaces of the FRD chip submodule.
As shown in Fig. 9, the thermal contact resistance decreases quickly with the increase of the clamping force at an early stage; the rate then slows down in the middle stage and finally tends toward stability. Due to the double-side cooling the heat flow can go through both the collector and emitter sides, and the thermal contact resistance of each side is connected in series, as shown in Fig. 6(b) and (c). Fig. 10 shows the difference of thermal contact resistance of both the collector and emitter sides under different clamping forces with the rated clamping force of 1 kN. We can see that thermal contact resistance sharply increases when the clamping force is lower. And the increment of clamping force can reduce it to some extent, but with the clamping force continuing to increase the rate of thermal contact resistance reduction slows down. Meanwhile, the failure rate of the chip sharply increases. The thermal contact resistance increment caused by lower clamping force gives rise to a high junction temperature and eventually leads to low reliability. Thus, the strictly controlled distribution of clamping force within PP IGBTs has a significant effect on the electrical and thermal characteristics and reliability of PP IGBTs.
2.3.2. Influence of the clamping force Though thermal contact resistance can be reduced to some extent through the refinement of machining craft to decrease surface roughness, the effect is limited. The reason is that the surface slope, which is a function of surface roughness, reduces with the surface roughness decrement and leads to an increment of thermal contact resistance, therefore finally suppresses the decrement of thermal contact resistance. Most importantly, the refinement of machining craft will inevitably result in a sharp increase in production cost. Microhardness is influenced by many external factors, such as temperature, but mainly determined by internal properties. Thus the effect of microhardness to reduce thermal contact resistance seems very small due to the material of PP IGBTs that is assigned. As discussed above, the only parameter that can greatly change the thermal contact resistance in PP IGBTs is the clamping force. The thermal contact resistance of each interface under different clamping forces is calculated based on the formulas aforementioned and is plotted in Fig. 9. Table 4 Thermal contact resistance comparison (@ 1 kN).
Before [K/W] After [K/W] Deviation [%]
Collector_Mo
Chip_Mo_C
Chip_Mo_E
Emitter_Mo
0.1024 0.086 16.02
0.1192 0.09 24.50
0.1211 0.117 3.39
0.1584 0.156 1.52
Fig. 10. The difference of the thermal contact resistance verus clamping force, with respect of the value at the nominal clamping force of 1 kN
Please cite this article as: E. Deng, et al., Optimization of the thermal contact resistance within press pack IGBTs, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.01.003
E. Deng et al. / Microelectronics Reliability xxx (2017) xxx–xxx
9
3. Experiments 3.1. Test bench Neither experimental equipment nor a fixture for PP IGBT thermal resistance measurement are available for purchasing or use, as PP IGBTs are a special packaging style for IGBT devices. In this paper, two fixtures for PP IGBTs are designed, and a thermal resistance tester (Phase11) is used to measure the thermal resistance of PP IGBTs. According to the principle of thermal resistance measurement, the K factor [29,30], which is short for the relationship between the junction temperature and voltage drop, should be measured for the junction temperature prediction during the test, and then the thermal resistance is recorded. The fixtures for the K factor and the thermal resistance of PP IGBTs are shown in Fig. 11.
Fig. 12. Comparison between voltage-temperature calibration curves (K factor) measured with a clamping force of 5 kN and 8 kN.
Four nuts are used to constrain the displacement to sustain the desired clamping force, which is given by a standard pressure machine. A disc spring is needed to compensate the physical movements during the process of clamping and thermal expansion. The clamping force has little influence on the K factor as long as the studied PP IGBT is in good contact because the measurement current is very small [13]. Fig. 12 shows the results of the 3300-V 360-A PP IGBT manufactured by westcode under 5 kN and 8 kN with the small constant sense current of 20 mA. The results show that the fixture can fully meet the measurement requirements because the slopes (°C/V) are exactly the same, except for a difference of 0.02 (°C) in intercept. The thermal resistance measurement fixture consists of dual heatsinks, a DUT, a disc spring, a clamping force sustaining plate, a hydraulic pump and a pressure sensor. Dual heatsinks are used to supply the needed clamping force and heat dissipation path. This fixture can easily realize the function of collector/emitter side and double side cooling for the DUT via managing the dual heatsinks. The clamping force uniformity on the surface is also very important for the results. Thus, a hemisphere is proposed to in this fixture because it is very good at the surface flatness adjustment. As mentioned above, the disc spring is needed to compensate the physical movements during the process of clamping and thermal expansion. The clamping force sensor is used to record the clamping force of the DUT and to deliver to display in real-time. The displayer is not shown in the picture above. The
Fig. 11. Measurement fixtures.
Fig. 13. The junction to case thermal resistance of the direct contact submodule under a clamping force of 1 kN.
Please cite this article as: E. Deng, et al., Optimization of the thermal contact resistance within press pack IGBTs, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.01.003
10
E. Deng et al. / Microelectronics Reliability xxx (2017) xxx–xxx
the collector-side cooling Rc_collector is the sum of Rc_Collector_Mo and Rc_Chip_Mo_C as shown in Fig. 6(c). 3.2.1. Influence of the temperature The junction to case thermal resistance of the submodule under the rated clamping force of 1 kN is repeatedly measured by the thermal resistance tester. From the data shown in Fig. 13, we can see that the thermal resistance reduces quickly with the repeated measurements and tends to be approximately stable after the 11th test. 3.2.2. Influence of the contact pressure The junction to case thermal resistance of the FRD chip submodule is measured under different clamping forces, ranging from 500 N to 1800 N. As shown in Fig. 14, the thermal resistance decreases along with the augment of clamping force. 4. Optimization Fig. 14. The relationship between the clamping force and the junction to case thermal resistance of the direct contact resistance.
hydraulic pump is used to provide a continuously adjustable clamping force, ranging up to 50 kN, which fully meets the requirements of all current ratings of PP IGBTs. The thermocouple located between the case and heatsink to measure the case temperature will be destroyed under such a high clamping pressure (1.2 kN/cm2), which is needed to assure the submodule electric and thermal contact. Thus, a channel for the thermocouple location is designed in both the collector- and emitter-side electrode poles to meet the measurement requirements, and the submodule in the test bench is marked red, as shown in Fig. 11. 3.2. Experimental results Because the geometry of components in PP IGBTs is quite small, especially the molybdenum plates and chips, which are several hundred μm, the traditional thermocouple and transient optical-thermal methods are not suitable for the measurement of thermal contact resistance within PP IGBTs. In this paper, the relationship between the external parameters, such as the clamping force and temperature, and the junction to case thermal resistance is considered to be equivalent to the relationship of the thermal contact resistance. The junction to case thermal resistance of the FRD chip submodule contains the bulk thermal resistance and thermal contact resistance. The variation of thermal contact resistance under different conditions will lead to a variation of junction to case thermal resistance with the assumption that the bulk thermal resistance is not sensitive to the clamping force or temperature (the temperature is not very high). Meanwhile, the accurate measurement of thermal contact resistance within PP IGBTs is very difficult. So the junction to case thermal resistance of the collector-side cooling with the emitter-side adiabatic is measured to deduce the behaviour of thermal contact resistance. Thus, the thermal contact resistance of
According to the simulation and experiments data, it is observed that the thermal contact resistance between multi-layers within the submodule accounts for a large proportion of the total thermal resistance. Although thermal contact resistance can be improved via surface roughness refinement or increasing the clamping force, the effect of improvement is limited by the machining craft and production cost. As shown in Fig. 2(b), the thermal interface material is usually used to fill up micro-gaps to increase the contact surface area and reduce the thermal contact resistance because the heat flow mainly goes through micro-contacts rather than micro-gaps. Nanosilver has been successfully applied to the semiconductor chip's connection for its high electric conductivity, good thermal conductivity, high melting point and low sintering temperature [31,32]. Nanosilver sintering technology, which can be applied to the FRD chip and molybdenum connection, is good for the semiconductor chip's reliability improvement. The direct contact and sintered (collector side of the FRD chip and molybdenum) submodules used in the thermal resistance test are shown in Fig. 15, and the bulk thermal resistance of the nanosilver is neglected because the thickness of sintered nanosilver is only 50 μm. All external conditions, such as the power dissipation, cooling conditions and clamping force, of the sintered submodule thermal resistance test remain the same with the direct contact submodule, with the consideration of just one experimental variable during the measurement. The experimental result of the sintered submodule is shown in Fig. 16. Similar to the direct contact submodule, the sintered submodule junction to case thermal resistance reduces with the repeated measurements and then tends toward a stable value, but it reaches stability more quickly than the direct contact one. The final stable thermal resistance is much lower than the direct contact one, with the decrement of approximately 18.8%. The result given above shows that nanosilver sintering technology can greatly reduce the thermal contact resistance or even eliminate it. The thermal contact resistance Rc_Chip_Mo_C between the collector side of the chip and molybdenum can be obtained and compared with the theoretical one, as shown in Table 5.
Fig. 15. Submodules for the thermal resistance measurement.
Please cite this article as: E. Deng, et al., Optimization of the thermal contact resistance within press pack IGBTs, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.01.003
E. Deng et al. / Microelectronics Reliability xxx (2017) xxx–xxx
11
increases when the clamping force is reduced, which, in turn, leads to a high junction temperature and eventually leads to low reliability. Most importantly, sequential increment of the clamping force will increase the failure rate of the chip but with a slight thermal contact resistance reduction. Thus, the strictly controlled distribution of the clamping force within PP IGBTs has a significant effect on the electrical and thermal characteristics and reliability of PP IGBTs. 3) Thermal contact resistance within PP IGBTs can be significantly reduced as the thermal interface material can effectively fill the gaps of the interface and increase the contact area. Furthermore, we can conclude that the decrease of thermal contact resistance reduces the junction temperature. Further power cycling tests will be carried out on the direct and sintered submodule to identify its reliability.
Acknowledgements
Fig. 16. The junction to case thermal resistance of the sintered submodule under a clamping force of 1 kN.
The FEM results before and after indicate the calculated thermal contact resistances derived from the measured values for the surface roughness before the thermal resistance measurement and after repeated measurements, respectively. And in experimental results, before indicates the first thermal resistance measurement and after indicates the value which is constant with repeated measurements. From the data comparison we can see that the theoretical model of thermal contact resistance is suitable for PP IGBTs as the theoretical results agree with the experimental data well. The reason the experimental data are slightly larger is that the nanosilver layer cannot totally eliminate the thermal contact resistance, and the nanosilver layer will yield a bulk thermal resistance. Totally different from the IGBT modules, the high temperature will reduce the thermal contact resistance and thus reduce the total thermal resistance of PP IGBTs. This characteristic will improve the reliability of PP IGBTs because the increment of thermal resistance after power cycling is one of the main failure modes of IGBT modules, which is caused by the degradation of the die attach. Meanwhile, the nanosilver sintering technology can greatly reduce the thermal contact resistance within PP IGBTs and improve the reliability of die attach with its high electric and thermal conductivity [33]. 5. Conclusions Various preliminary conclusions can be drawn via the calculation and measurement of the thermal contact resistance of the single FRD chip submodule. 1) The surface topography, such as the surface roughness and surface slope, has a great impact on the thermal contact resistance. Though reducing the surface roughness can greatly improve the thermal characteristics, the surface slope also decreases with the roughness which, in turn, increases the thermal contact resistance. At the same time, the cost will sharply increase. Thus, in actual applications, a comprehensive consideration should be taken regarding the reduction of thermal contact resistance. 2) The clamping force is the most important parameter that influences the thermal contact resistance. Thermal contact resistance sharply Table 5 Thermal contact resistance comparison.
Before [K/W] After [K/W]
FEM
Measurement
Deviation [%]
0.1192 0.093
0.142 0.112
16.06 16.96
The work was supported by the National Natural Science Foundation of China [51477048], the National Key Research and Development Program of China [2016YFB0901803], and the Science and Technology Project of State Grid [5455DW150005]. References [1] F. Wakeman, D. Hemmings, W. Findlay, G. Lockwood, Pressure Contact IGBT, Testing for Reliability, Westcode Semiconductors Ltd., March 2012 [2] C. Busca, R. Teodorescu, F. Blaabjerg, et al., Dynamic Thermal Modelling and Analysis of Press-pack IGBTs Both at Component-level and Chip-level [C]//Industrial Electronics Society, IECON 2013 - 39th Annual Conference of the IEEE, IEEE, 2013 677–682. [3] Z. Ping, Y. Xuan, L.I. Qiang, Development on thermal contact resistance [J], CIESC J. 4 (2012) 11–20. [4] M. Yovanovich, E. Marotta, Thermal Spreading and Contact Resistances [M], Wiley, New Jersey, 2003 261–395. [5] M.A. Lambert, L.S. Fletcher, Thermal contact conductance of spherical, rough metals [J], Trans. Asme Serie C J. Heat Transf. 119 (4) (1997) 684–690. [6] Y.Z. Li, C.V. Madhusudana, E. Leonardi, On the enhancement of the thermal contact conductance: effect of loading history [J], J. Heat Transf. 122 (1) (2000) 46–49. [7] T. Mcwaid, E. Marschall, Thermal contact resistance across pressed metal contacts in a vacuum environment [J], Int. J. Heat Mass Transf. 35 (11) (1992) 2911–2920. [8] M. Bahrami, Thermal Resistances of Gaseous Gap for Conforming Rough Contacts [C]//42nd AIAA Aerospace Sciences Meeting and Exhibit, 2004 (2002–0317). [9] Astm, Astm D5470-06: Standard Test Method for Thermal Transmission Properties of Thermally Conductive Electrical Insulation Materials [S], Astm International, America, 2006. [10] Y. Osone, T. Kubo, N. Nakazato, Optical Measurement of Thermal Contact Conductance Between Wafer-Like Thin Solid Samples [J], Trans. Asme Serie C J. Heat Transf. 121 (4) (1999) 954–963. [11] L. Shi, Investigation of Heat Transport on the Solid-Solid Contact Interface at Low Temperature [D], Huazhong University of Science & Technology, Wuhan, 2006. [12] O. Kwon, L. Shi, A. Majumdar, Scanning thermal wave microscopy [J], J. Heat Transf. 125 (2003) 156–163. [13] T. Poller, J. Lutz, S. D'Arco, et al., Determination of the thermal and electrical contact resistance in press-pack IGBTs [J], Mikrobiologiia 75 (3) (2006) 305–311. [14] M. Grujicic, C.L. Zhao, E.C. Dusel, The effect of thermal contact resistance on heat management in the electronic packaging [J], Appl. Surf. Sci. 246 (1–3) (2005) 290–302. [15] M. Bahrami, Thermal Resistances of Gaseous Gap for Conforming Rough Contacts [M], 2004. [16] W.H. McAdams, Heat Transmission, McGraw-Hill, New York, 1954. [17] A. Bejan, D. Kraus, Heat Transfer Handbook, John Willey, New York, 2003. [18] J.B.P. Williamson, J. Pullen, R.T. Hunt, The shape of solid surfaces, Surface Mechanics, ASME, New York 1969, pp. 24–35. [19] J.A. Greenwood, The area of contact between rough surfaces and flats, J. Lubr. Technol. (January 1967) 81–91. [20] M. Bahrami, J.R. Culham, M.M. Yovanovich, et al., Review of thermal joint resistance models for non-conforming rough surfaces in a vacuum [C]//ASME 2003 heat transfer summer conference, Am. Soc. Mech. Eng. (2003) 411–431. [21] L.H. Tanner, M. Fahoum, A study of the surface parameters of ground and lapped metal surfaces, using specular and diffuse reflection of laser light, Wear 36 (1976) 299–316. [22] V.W. Antonetti, T.D. Whittle, R.E. Simons, An approximate thermal contact conductance correlation, ASME, The 28th National Heat Transfer Conference Minneapolis, Minnesota, HTD-Vol. 170, Experimental/Numerical Heat Transfer in Combustion and Phase Change 1991, pp. 35–42. [23] M.A. Lambert, Thermal Contact Conductance of Spherical, Rough Metals(Ph.D. Thesis) Dept. of Mech. Eng., Texas A&M University, USA, 1995. [24] M. Bahrami, J.R. Culham, M.M. Yovanovich, G.E. Schneider, Thermal contact resistance of non-conforming rough surfaces. Part 1. Mechanical model, Proceedings of
Please cite this article as: E. Deng, et al., Optimization of the thermal contact resistance within press pack IGBTs, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.01.003
12
[25]
[26]
[27]
[28]
E. Deng et al. / Microelectronics Reliability xxx (2017) xxx–xxx the 36th AIAA Thermophysics Conference, AIAA Paper No. 2003-4197, June23–26, Orlando, FL, 2003. M. Bahrami, J.R. Culham, M.M. Yovanovich, G.E. Schneider, Thermal contact resistance of non-conforming rough surfaces. Part 2. Thermal model, Proceedings of the 36th AIAA Thermophysics Conference, AIAA Paper No. 2003-4198, June 23–26, Orlando, FL, 2003. S. Song, M.M. Yovanovich, Relative contact pressure: dependence upon surface roughness and Vickers microhardness, J. Thermophys. Heat Transf. 2 (1) (January 1988) 43–47. M.M. Yovanovich, New Contact and Gap Correlations for Conforming Rough Surfaces, AIAA-81-1164, Presented at AIAA 16th Thermophysics Conference, Palo Alto, CA, June 1981. V.W. Antonetti, M.M. Yovanovich, Thermal Contact Resistance in Microelectronic Equipment, Thermal Management Concepts in Microelectronic Packaging From Component to System, ISHM Technical Monograph Series 6984-003, 1984 135–151.
[29] P. Cova, et al., On the effect of power cycling stress on IGBT modules, Microelectron. Reliab. 38 (6–8) (1998) 1347–1352. [30] U. Scheuermann, R. Schmidt, Investigations on the vce(t) - method to determine the junction temperature by using the chip itself as sensor, PCIM, 2009. [31] G.Q. Lu, J.N. Calata, G. Lei, et al., Low-temperature and Pressureless Sintering Technology for High-performance and High [C]//2007 International Conference on Thermal, Mechanical and Multi-Physics Simulation Experiments in Microelectronics and Micro-Systems. EuroSime, 2007, 2007 1–5. [32] G. Lu, Nanosilver paste: its sintering behavior and applications for chip interconnection, IEEE Int. Symp. 755 (3) (1998) 298-298. [33] N. Heuck, K. Guth, M. Thoben, et al., Aging of new Interconnect-Technologies of Power-Modules during Power-Cycling [C]//Integrated Power Systems (CIPS), 2014 8th International Conference on VDE, 2014 1–6.
Please cite this article as: E. Deng, et al., Optimization of the thermal contact resistance within press pack IGBTs, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.01.003
Contents lists available at ScienceDirect
Microelectronics Reliability journal homepage: www.elsevier.com/locate/microrel
Optimization of the thermal contact resistance within press pack IGBTs Erping Deng a,b,⁎, Zhibin Zhao a, Peng Zhang b, Yongzhang Huang a, Jinyuan Li b a b
State Key Laboratory of Alternate Electrical Power System with Renewable Energy Sources, North China Electric Power University, Changping District, Beijing 102206, China State Grid Global Energy Interconnection Research Institute, Changping District, Beijing 102211, China
a r t i c l e
i n f o
Article history: Received 25 August 2016 Received in revised form 8 January 2017 Accepted 8 January 2017 Available online xxxx Keywords: Press pack IGBTs Thermal contact resistance Temperature Clamping force Nanosilver sintering technology
a b s t r a c t The research of thermal contact resistance between multi-layers within press pack IGBTs (PP IGBTs) is significant for optimizing the PP IGBTs' thermal resistance to improve reliability, as the thermal contact resistance accounts for approximately 50% of the total thermal resistance of PP IGBTs. In this paper, thermal contact resistance between multi-layers is analysed via a finite element model (FEM) of a single fast recovery diode (FRD) submodule. Most importantly, the influence of temperature and clamping force on the thermal contact resistance is also discussed, and findings are verified by submodule thermal resistance experiments. Based on the FEM and experimental results, nanosilver sintering technology is proposed to fill the gap between the contact interfaces to reduce thermal contact resistance. The fabrication of a sintered single FRD submodule is also investigated in this paper, and the results of the sintered sample indicate that the thermal resistance is reduced by approximately 18.8% compared to a direct contact sample. © 2017 Published by Elsevier Ltd.
1. Introduction To meet the growing requirements of applications of IGBT devices, capacity and reliability have become great challenges for IGBT devices. Press pack IGBTs (PP IGBTs) have gradually been applied to high-voltage and high-power-density applications, such as electric locomotives and HVDC transmission, because of their high reliability, double-side cooling, high power density and ease of laying out in series [1] compared to typical wire-bonded IGBT modules. The simplified internal structure of the studied PP IGBT is shown in Fig. 1, as it may be observed that the PP IGBT has a multi-layered structure. Two copper electrodes (the collector and emitter pole) provide the electrical and thermal paths for the silicon chips, and the silicon IGBT chips are sandwiched by two molybdenum plates, which help with the uniform distribution of the clamping force. A silver shim plate, together with a silicon IGBT chip and two molybdenum plates, is used to form a chip assembly. An external clamping force is required to maintain the electrical and thermal contact of all components within PP IGBTs. Thermal contact resistance is a very important parameter for thermal resistance optimization and reliability improvement, as it accounts for approximately 50% of the total thermal resistance [2]. When two rough surfaces are bought into contact, actual contact only occurs at certain discrete spots or micro-contacts, while the noncontacting areas form vacuums or are filled with some medium (such as air, water or oil). The actual contact area accounts for approximately 0.01%–0.1% of the nominal contact area, and the proportion only ⁎ Corresponding author at: No. 2, Beinong Road, Changping District, 102206 Beijing, China. E-mail address: [email protected] (E. Deng).
increases to 1%–2% under a contact pressure of 10 MPa [3]. Because the actual contact surface area is relatively small, as is the thermal conductivity of the interfacial gases, heat flow across the interface experiences a relatively large thermal resistance, commonly referred to as thermal contact resistance [4,5]. Thermal contact resistance is a complex parameter affected by the material properties, surface morphology, contact pressure, temperature and so on [6,7], and thus the precision of the test bench will be a significant challenge for measurement. At present, steady-state and transient methods are used to predict the thermal contact resistance, and the steady-state method is most commonly used. The temperature difference between the interfaces can be obtained by a linear fitting after the temperature has been measured between two samples in the steady-state method [8]. Not only will the temperature distribution of the specimen be disturbed by the embedded thermocouple, but the accuracy of the adjacent measured temperature will also be influenced since this steady method requires a thermocouple to measure the temperature. Most importantly, many thermocouples should be located in the samples, and thus the steady-state method is not suitable for a specimen whose geometry is small (in the millimetre ranger). To cover some disadvantages of the thermocouple method, an infrared imaging system with an accuracy of 0.1 °C, instead of a thermocouple, is put forward by various scholars to record the temperature of the two-dimensional interface. Although the accuracy is greatly improved, an error of approximately 23% is still presented [9]. The optical-thermal method is the widely used method among transient thermal contact resistance measurement methods [10,11]. Thermal contact resistance is obtained by the phase difference of the heat wave and modulation wave after encountering the interface. However, the accuracy of the optical-thermal method is affected by the interface characteristic, as the heat wave is diffused at the contact interface, and it
http://dx.doi.org/10.1016/j.microrel.2017.01.003 0026-2714/© 2017 Published by Elsevier Ltd.
Please cite this article as: E. Deng, et al., Optimization of the thermal contact resistance within press pack IGBTs, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.01.003
2
E. Deng et al. / Microelectronics Reliability xxx (2017) xxx–xxx
Fig. 1. A cross-section schematic diagram for press pack IGBTs.
destroys their phase relationship [12]. All the methods mentioned thus far are for the measurement of thermal contact resistance between power semiconductor devices and heatsinks or among copper bus bar junctions. Namely, all the methods are for the thermal contact resistance outside power semiconductors rather than within them. Although T. Poller et al. [13] measured the thermal contact resistance within PP IGBTs through a combination of experiments and a finite element method, the deviation of the thermocouple still exists, and the testing process is relatively complicated. Thermal contact resistance within PP IGBTs has not been identified for its special packaging style and working conditions likewise thyristors. And the measurement of thermal contact resistance within PP IGBTs is still impossible with the limits that no probes can be placed between two test samples and the clamping force system. Furthermore, the optical-thermal or other non-contact methods are not suitable for PP IGBTs because the chips are clamped and encapsulated. Thus, the theoretical model with the measured data to make it more precise is proposed to predict the behaviour of thermal contact resistance in this paper. Meanwhile, an indirect measurement of junction to case thermal resistance is proposed to predict the thermal contact resistance within PP IGBTs and to identify the theoretical values. The point of this indirect measurement is to use the variation of junction to case thermal resistance to deduce the variation of thermal contact resistance. According to the theoretical and experimental results, nanosilver sintering technology is introduced to reduce the thermal contact resistance, and a nanosilver-sintered sample is manufactured to verify this. The structure of this paper is set as follows. Characteristics of PP IGBTs and the status of thermal contact resistance calculation and measurement are presented in the introduction. The finite element model based on theoretical model is used to calculate the thermal contact resistance within PP IGBTs, as well as the influence of temperature and clamping force in the second section. And some parameters needed in the calculation
are measured through experiment to make the model more precise. A single fast recovery diode (FRD) chip submodule is fabricated to measure the thermal resistance change and compare it with the theoretical values in the experiment part. The optimization section presents a nanosilver sintering technology for optimizing the thermal contact resistance between the collector side of the chip and molybdenum, and a sintered sample is also manufactured to make a comparison with the direct contact submodule. Conclusions and outlook are provided in the last section. 2. Theoretical 2.1. Basic principle A micro-structure diagram of the contact interface is shown in Fig. 2 [14]; the contact spot is always named micro-contact, and the non-contact area is filled with air. Because the thermal conductivity of the interfacial gases is relatively low (0.023 W/(m·K)), heat flow across the interface experiences a relatively large thermal resistance, commonly referred to as thermal contact resistance. Micro-contacts, which have a relatively long distance among them, are general distributed randomly when two rough surfaces came into contact. Only thermal conduction is considered between the micro-contact and the interfacial gaps, which are usually filled with air in this situation [15]. In most practical situations concerning thermal contact resistance, the gap thickness between two contacting bodies is quite small (b10 μm), and thus the Grashof number based on the gap thickness is b 2000. Consequently, in most instances, the heat transfer of convection through the interstitial gas in the gap is neglected [16]. Thermal radiation across the interfacial gaps is generally considered as insignificant as the surface temperature is less than approximately 700 K [17]. In conclusion, the thermal contact conductance hc of the contact interfaces, which can be calculated by
Fig. 2. Micro-contact interface schematic diagram.
Please cite this article as: E. Deng, et al., Optimization of the thermal contact resistance within press pack IGBTs, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.01.003
E. Deng et al. / Microelectronics Reliability xxx (2017) xxx–xxx
3
Fig. 3. The roughness surfaces equivalent schematic diagram.
formula (1), consists of the micro-contact thermal contact conductance hs and interfacial gap thermal contact conductance hg. Furthermore, thermal contact resistance Rc can be obtained through formula (2), as the thermal contact conductance hc and nominal contact area A is given. hc ¼ hg þ hs
ð1Þ
Rc ¼ A hc
ð2Þ
2.1.1. Micro-contact conductance hs Williamson et al. [18] have shown experimentally that many of the techniques used to produce engineering surfaces give a Gaussian
distribution of surface heights. The contact between Gaussian rough surfaces can be considered as the contact between a single Gaussian surface, having the effective surface characteristics, placed in contact with a perfectly smooth flat surface, as shown in Fig. 3 [19]. The equivalent roughness and surface slope can be calculated from formula (3). σ¼
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi σ 21 þ σ 22 and m ¼ m21 þ m22
ð3Þ
where subscripts 1 and 2 represent the two contact surfaces, respectively. The surface roughness of the rough contact interface is largely influenced by the machining precision and should be measured via a surface
Fig. 4. Contact configurations.
Please cite this article as: E. Deng, et al., Optimization of the thermal contact resistance within press pack IGBTs, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.01.003
4
E. Deng et al. / Microelectronics Reliability xxx (2017) xxx–xxx
roughness tester. The surface slope m affected by the manufacture craft and surface roughness, etc., can be approximately predicted through a function of surface roughness with theoretical calculation or experimental data. The reason is that the surface topography is determined by the manufacture craft or different process [20] and four typical surfaces are shown in Fig. 4. Meanwhile, the surface roughness and slope is depended on the surface topography to a large extent as shown in Fig. 3. Correlations for surface slope are summarized in Table 1. As the measured surface roughness of those interfaces in PP IGBTs is much lower than 1.6 μm, the surface slope is identified by [22] shown in Table 1. Thermal resistance associated with micro-contact (known as thermal constriction/spreading resistance) is generally modelled by applying the so-called flux cube solution to each micro-contact with the assumption that the asperities undergo plastic deformation. According to the theory of Bahrami et al., thermal contact conductance of the contact interface can be calculated as long as a single micro-contact thermal model is given [24,25]. hs ¼ 1:25ks
m p 0:95 σ Hc
ð4Þ
where p is contact pressure and ks is the harmonic mean of solid thermal conductivity of the two bodies, which is shown in formula (5). ks ¼
2k1 k2 k1 þ k2
ð5Þ
Table 1 Correlations for surface slope. Reference
Correlation
Tanner [21] Antonetti et al. [22] Lambert [23]
m = 0.152σ0.4 m = 0.124σ0.743 , σ≤ 1.6 μm m = 0.076σ0.52
Different from the IGBT chip used in most wire bond packaging, the gate of the IGBT chip used in press pack packaging is located in the corner of the chip. Thus, grooves are shaped into the pedestals, as well as other components on the emitter side, for the gate pin of the housing. Meanwhile, a constant voltage is needed to apply to the IGBT chip's gate so the device channel is open and current flows through during the measurement. To exclude interference factors and reduce the complexity of the experiment, a single FRD chip submodule is designed in this paper to predict the behaviour of thermal contact resistance within PP IGBTs because the heat flux of IGBT chip and FRD chip are the same. The three-dimensional structure diagram and equivalent thermal network are shown in Fig. 6. Both sides of the FRD chip are covered with several μm of aluminium to realize the connection. The thermal contact resistance of the collector side and emitter side should be calculated separately because the surface roughness of the aluminium layer is quite different. According to the description above, a single FRD chip submodule contains four contact interfaces, which is shown in Fig. 6(a), and the material properties are listed in Table 2.
Hc is the surface microhardness of the softer of the two contacting solids. The microhardness is generally complex because it depends on several geometric and physical parameters, such as the Vickers microhardness correlation coefficients, surface roughness and contact pressure. As shown in formula (6), Song and Yovanovich related Hmic to the surface parameters and nominal pressure [26]. 0
1 11þ0:071c
p p B C ¼@ c2 A H mic c1 1:62 σσ0 m
2
ð6Þ
2.1.2. Micro-gap conductance hg Many theoretical models of the gap conductance hg are proposed by scholars, and the most commonly used is given by the approximation of Yovanovich [27], which is shown in (7). hg ¼
kg Y þM
ð7Þ
where kg is the thermal conductivity of the gap substance and the gas parameter M accounts for rarefaction effects at high temperatures and low gas pressures. The effective gap thickness Y, shown in Fig. 3, can be calculated accurately by means of the simple power-law correlation equation proposed by Antonetti and Yovanovich [28]. Y ¼ 1:53σ
p Hc
−0:097
ð8Þ
2.2. Finite element model The IGBT chip surface is quite rough from a microscopic view, as are other components within PP IGBTs. Fig. 5 shows a microscopic sectional view of the IGBT chip for press pack packaging. An aluminium thickness of several μm is needed to cover the chip surface to realize the connection with the external electrodes. As for the rough surfaces, the contact interface will introduce both electrical and thermal contact resistance.
Fig. 5. Micro cross-section schematic diagram of an IGBT chip.
Please cite this article as: E. Deng, et al., Optimization of the thermal contact resistance within press pack IGBTs, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.01.003
E. Deng et al. / Microelectronics Reliability xxx (2017) xxx–xxx
Only thermal conduction is considered in this paper according to the discussion mentioned before, and the basic differential equation is shown in (9), where ρ is the material density, Cp is the constant pressure specific heat, t is the time, k is the thermal conductivity and Q is the volumetric heat source. ρC p
∂T −∇ ðk∇T Þ ¼ Q ∂t
ð9Þ
5
The boundary conditions for the thermal model are set as follows. The temperatures of the collector and emitter pole outer faces (case temperature) have been set to 60 °C, with the assumption that the external water-cooling system can sustain a constant case temperature. The heat generation in the active area of the FRD chip is set as 100 W. The equivalent clamping force of the single FRD chip submodule is assumed to be 1 kN according to PP IGBTs' work condition (1.2 kN/cm2) and the contact area.
Fig. 6. Single FRD chip submodule.
Please cite this article as: E. Deng, et al., Optimization of the thermal contact resistance within press pack IGBTs, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.01.003
6
E. Deng et al. / Microelectronics Reliability xxx (2017) xxx–xxx
2.3. Results
Table 2 Material parameters.
Density [kg/m3] Thermal conductivity [W / (m ∗ K)] Specific heat [J / (kg ∗ K)] Elasticity modulus [GPa] Surface roughness [μm] Surface slope
Copper
Moly
Alum
8930 394 385 110 ※ ※
10,200 145 217 330 ※ ※
2700 210 900 68 ※ ※
where ※ denotes that the material property should be measured via experiment.
As shown in the thermal contact resistance calculation formulas, thermal contact resistance is mainly affected by surface roughness, surface slope, microhardness and contact pressure. In this paper, thermal contact resistance within all the interfaces are calculated through different interface parameters based on the formulas aforementioned and the interface between the collector side of the FRD chip and molybdenum is used as an example to show the influence of the previously mentioned parameters. Surface roughness ranges from 0.1 μm to 1 μm according to the manufacture craft for PP IGBTs. Surface slope ranges from 0.02 to 0.2
Fig. 7. Thermal contact resistance of the interface micro-contact.
Please cite this article as: E. Deng, et al., Optimization of the thermal contact resistance within press pack IGBTs, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.01.003
E. Deng et al. / Microelectronics Reliability xxx (2017) xxx–xxx
according to the surface roughness. Microhardness ranges from 1 GPa to 10 GPa according to material property, and the equivalent clamping force ranges from 200 N to 2000 N based on the clamping force of a single FRD chip within PP IGBTs. The calculation results are shown in Figs. 7 and 8. The unit of thermal contact resistance is K/W. It is observed that as expected, a higher contact pressure (equivalent clamping force) and a lower microhardness Hmic, which accelerate the plastic deformation of the asperities and the formation of micro-contacts with a larger contact surface area, give arise to a lower value of the thermal micro-contact resistance. Likewise, a decrease in the surface roughness and an increase in the roughness slope, which result in higher values of the micro-contact surface area, also give rise to a lower thermal micro-contact resistance. The micro-gap thermal
7
resistance is mainly affected by contact pressure, surface roughness and microhardness, with the exception of surface slope. 2.3.1. Influence of the temperature Temperature is the most important parameter affecting many material properties, such as microhardness and thermal conductivity. In general, thermal contact resistance between contact interfaces always decreases with the increase of temperature because a temperature increment, which lowers the microhardness (i.e., it softens), gives rise to a larger contact surface area. Microhardness changes with temperature eventually lead to a change of surface roughness. However, accurate measurement of microhardness is quite difficult, and the measurement would cause irreversible damage to the chip. In this paper, the influence
Fig. 8. Thermal contact resistance of the interface gap.
Please cite this article as: E. Deng, et al., Optimization of the thermal contact resistance within press pack IGBTs, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.01.003
8
E. Deng et al. / Microelectronics Reliability xxx (2017) xxx–xxx
Table 3 Comparison of the surface roughness. Surface roughness σ [μm]
Before
After
Deviation
Collector Collector-side molybdenum Chip collector side Chip emitter side Emitter-side molybdenum Emitter
0.226 0.344 0.495 0.147 0.334 0.638
0.151 0.268 0.147 0.057 0.307 0.617
33.18% 22.09% 70.30% 61.22% 8.08% 3.29%
of temperature on material properties, especially the surface properties, is represented by its final appearance (surface roughness decrement). Micro-contacts undergo an elasticity deformation or even a plastic deformation under the external clamping force as the microhardness is decreased with the augment of the temperature of the chip and other components during the thermal resistance test. Although the temperature of the chip and other components is reduced after the measurement is completed, the deformation is still unable to be fully restored to the state before measurement under the external clamping force and eventually leads to a decrease of surface roughness. The surface properties, such as surface roughness, would no longer change after several repeated thermal resistance measurements (equivalent to temperature cycling). The surface roughness (arithmetic average of absolute values Ra) of every contact surface is measured through a surface roughness tester (Mitutoyo SURFTEST SJ-310) after several repeated measurements, as shown in Table 3, and compared with the value before the thermal resistance test. It is observed that, as expected, the surface roughness of the components after testing is much smaller than before testing, especially for the components of the collector side. This is because the thermal resistance test only measures the junction to case thermal resistance of the collector-side cooling with the emitter-side adiabatic, and the heat flow only goes through the collector side. The reason for why we just measure the collector-side cooling is that thermal contact resistance under double-side cooling is the parallel relationship of the one under collector-side cooling and emitter-side cooling. So it is very difficult to distinguish the relationship of each interface and hard to verify the theoretical values. The calculated thermal contact resistance before and after the thermal resistance test based on the theoretical model mentioned before is shown in Table 4 with the clamping force of 1 kN.
Fig. 9. The relationship between the clamping force and thermal contact resistance between four interfaces of the FRD chip submodule.
As shown in Fig. 9, the thermal contact resistance decreases quickly with the increase of the clamping force at an early stage; the rate then slows down in the middle stage and finally tends toward stability. Due to the double-side cooling the heat flow can go through both the collector and emitter sides, and the thermal contact resistance of each side is connected in series, as shown in Fig. 6(b) and (c). Fig. 10 shows the difference of thermal contact resistance of both the collector and emitter sides under different clamping forces with the rated clamping force of 1 kN. We can see that thermal contact resistance sharply increases when the clamping force is lower. And the increment of clamping force can reduce it to some extent, but with the clamping force continuing to increase the rate of thermal contact resistance reduction slows down. Meanwhile, the failure rate of the chip sharply increases. The thermal contact resistance increment caused by lower clamping force gives rise to a high junction temperature and eventually leads to low reliability. Thus, the strictly controlled distribution of clamping force within PP IGBTs has a significant effect on the electrical and thermal characteristics and reliability of PP IGBTs.
2.3.2. Influence of the clamping force Though thermal contact resistance can be reduced to some extent through the refinement of machining craft to decrease surface roughness, the effect is limited. The reason is that the surface slope, which is a function of surface roughness, reduces with the surface roughness decrement and leads to an increment of thermal contact resistance, therefore finally suppresses the decrement of thermal contact resistance. Most importantly, the refinement of machining craft will inevitably result in a sharp increase in production cost. Microhardness is influenced by many external factors, such as temperature, but mainly determined by internal properties. Thus the effect of microhardness to reduce thermal contact resistance seems very small due to the material of PP IGBTs that is assigned. As discussed above, the only parameter that can greatly change the thermal contact resistance in PP IGBTs is the clamping force. The thermal contact resistance of each interface under different clamping forces is calculated based on the formulas aforementioned and is plotted in Fig. 9. Table 4 Thermal contact resistance comparison (@ 1 kN).
Before [K/W] After [K/W] Deviation [%]
Collector_Mo
Chip_Mo_C
Chip_Mo_E
Emitter_Mo
0.1024 0.086 16.02
0.1192 0.09 24.50
0.1211 0.117 3.39
0.1584 0.156 1.52
Fig. 10. The difference of the thermal contact resistance verus clamping force, with respect of the value at the nominal clamping force of 1 kN
Please cite this article as: E. Deng, et al., Optimization of the thermal contact resistance within press pack IGBTs, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.01.003
E. Deng et al. / Microelectronics Reliability xxx (2017) xxx–xxx
9
3. Experiments 3.1. Test bench Neither experimental equipment nor a fixture for PP IGBT thermal resistance measurement are available for purchasing or use, as PP IGBTs are a special packaging style for IGBT devices. In this paper, two fixtures for PP IGBTs are designed, and a thermal resistance tester (Phase11) is used to measure the thermal resistance of PP IGBTs. According to the principle of thermal resistance measurement, the K factor [29,30], which is short for the relationship between the junction temperature and voltage drop, should be measured for the junction temperature prediction during the test, and then the thermal resistance is recorded. The fixtures for the K factor and the thermal resistance of PP IGBTs are shown in Fig. 11.
Fig. 12. Comparison between voltage-temperature calibration curves (K factor) measured with a clamping force of 5 kN and 8 kN.
Four nuts are used to constrain the displacement to sustain the desired clamping force, which is given by a standard pressure machine. A disc spring is needed to compensate the physical movements during the process of clamping and thermal expansion. The clamping force has little influence on the K factor as long as the studied PP IGBT is in good contact because the measurement current is very small [13]. Fig. 12 shows the results of the 3300-V 360-A PP IGBT manufactured by westcode under 5 kN and 8 kN with the small constant sense current of 20 mA. The results show that the fixture can fully meet the measurement requirements because the slopes (°C/V) are exactly the same, except for a difference of 0.02 (°C) in intercept. The thermal resistance measurement fixture consists of dual heatsinks, a DUT, a disc spring, a clamping force sustaining plate, a hydraulic pump and a pressure sensor. Dual heatsinks are used to supply the needed clamping force and heat dissipation path. This fixture can easily realize the function of collector/emitter side and double side cooling for the DUT via managing the dual heatsinks. The clamping force uniformity on the surface is also very important for the results. Thus, a hemisphere is proposed to in this fixture because it is very good at the surface flatness adjustment. As mentioned above, the disc spring is needed to compensate the physical movements during the process of clamping and thermal expansion. The clamping force sensor is used to record the clamping force of the DUT and to deliver to display in real-time. The displayer is not shown in the picture above. The
Fig. 11. Measurement fixtures.
Fig. 13. The junction to case thermal resistance of the direct contact submodule under a clamping force of 1 kN.
Please cite this article as: E. Deng, et al., Optimization of the thermal contact resistance within press pack IGBTs, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.01.003
10
E. Deng et al. / Microelectronics Reliability xxx (2017) xxx–xxx
the collector-side cooling Rc_collector is the sum of Rc_Collector_Mo and Rc_Chip_Mo_C as shown in Fig. 6(c). 3.2.1. Influence of the temperature The junction to case thermal resistance of the submodule under the rated clamping force of 1 kN is repeatedly measured by the thermal resistance tester. From the data shown in Fig. 13, we can see that the thermal resistance reduces quickly with the repeated measurements and tends to be approximately stable after the 11th test. 3.2.2. Influence of the contact pressure The junction to case thermal resistance of the FRD chip submodule is measured under different clamping forces, ranging from 500 N to 1800 N. As shown in Fig. 14, the thermal resistance decreases along with the augment of clamping force. 4. Optimization Fig. 14. The relationship between the clamping force and the junction to case thermal resistance of the direct contact resistance.
hydraulic pump is used to provide a continuously adjustable clamping force, ranging up to 50 kN, which fully meets the requirements of all current ratings of PP IGBTs. The thermocouple located between the case and heatsink to measure the case temperature will be destroyed under such a high clamping pressure (1.2 kN/cm2), which is needed to assure the submodule electric and thermal contact. Thus, a channel for the thermocouple location is designed in both the collector- and emitter-side electrode poles to meet the measurement requirements, and the submodule in the test bench is marked red, as shown in Fig. 11. 3.2. Experimental results Because the geometry of components in PP IGBTs is quite small, especially the molybdenum plates and chips, which are several hundred μm, the traditional thermocouple and transient optical-thermal methods are not suitable for the measurement of thermal contact resistance within PP IGBTs. In this paper, the relationship between the external parameters, such as the clamping force and temperature, and the junction to case thermal resistance is considered to be equivalent to the relationship of the thermal contact resistance. The junction to case thermal resistance of the FRD chip submodule contains the bulk thermal resistance and thermal contact resistance. The variation of thermal contact resistance under different conditions will lead to a variation of junction to case thermal resistance with the assumption that the bulk thermal resistance is not sensitive to the clamping force or temperature (the temperature is not very high). Meanwhile, the accurate measurement of thermal contact resistance within PP IGBTs is very difficult. So the junction to case thermal resistance of the collector-side cooling with the emitter-side adiabatic is measured to deduce the behaviour of thermal contact resistance. Thus, the thermal contact resistance of
According to the simulation and experiments data, it is observed that the thermal contact resistance between multi-layers within the submodule accounts for a large proportion of the total thermal resistance. Although thermal contact resistance can be improved via surface roughness refinement or increasing the clamping force, the effect of improvement is limited by the machining craft and production cost. As shown in Fig. 2(b), the thermal interface material is usually used to fill up micro-gaps to increase the contact surface area and reduce the thermal contact resistance because the heat flow mainly goes through micro-contacts rather than micro-gaps. Nanosilver has been successfully applied to the semiconductor chip's connection for its high electric conductivity, good thermal conductivity, high melting point and low sintering temperature [31,32]. Nanosilver sintering technology, which can be applied to the FRD chip and molybdenum connection, is good for the semiconductor chip's reliability improvement. The direct contact and sintered (collector side of the FRD chip and molybdenum) submodules used in the thermal resistance test are shown in Fig. 15, and the bulk thermal resistance of the nanosilver is neglected because the thickness of sintered nanosilver is only 50 μm. All external conditions, such as the power dissipation, cooling conditions and clamping force, of the sintered submodule thermal resistance test remain the same with the direct contact submodule, with the consideration of just one experimental variable during the measurement. The experimental result of the sintered submodule is shown in Fig. 16. Similar to the direct contact submodule, the sintered submodule junction to case thermal resistance reduces with the repeated measurements and then tends toward a stable value, but it reaches stability more quickly than the direct contact one. The final stable thermal resistance is much lower than the direct contact one, with the decrement of approximately 18.8%. The result given above shows that nanosilver sintering technology can greatly reduce the thermal contact resistance or even eliminate it. The thermal contact resistance Rc_Chip_Mo_C between the collector side of the chip and molybdenum can be obtained and compared with the theoretical one, as shown in Table 5.
Fig. 15. Submodules for the thermal resistance measurement.
Please cite this article as: E. Deng, et al., Optimization of the thermal contact resistance within press pack IGBTs, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.01.003
E. Deng et al. / Microelectronics Reliability xxx (2017) xxx–xxx
11
increases when the clamping force is reduced, which, in turn, leads to a high junction temperature and eventually leads to low reliability. Most importantly, sequential increment of the clamping force will increase the failure rate of the chip but with a slight thermal contact resistance reduction. Thus, the strictly controlled distribution of the clamping force within PP IGBTs has a significant effect on the electrical and thermal characteristics and reliability of PP IGBTs. 3) Thermal contact resistance within PP IGBTs can be significantly reduced as the thermal interface material can effectively fill the gaps of the interface and increase the contact area. Furthermore, we can conclude that the decrease of thermal contact resistance reduces the junction temperature. Further power cycling tests will be carried out on the direct and sintered submodule to identify its reliability.
Acknowledgements
Fig. 16. The junction to case thermal resistance of the sintered submodule under a clamping force of 1 kN.
The FEM results before and after indicate the calculated thermal contact resistances derived from the measured values for the surface roughness before the thermal resistance measurement and after repeated measurements, respectively. And in experimental results, before indicates the first thermal resistance measurement and after indicates the value which is constant with repeated measurements. From the data comparison we can see that the theoretical model of thermal contact resistance is suitable for PP IGBTs as the theoretical results agree with the experimental data well. The reason the experimental data are slightly larger is that the nanosilver layer cannot totally eliminate the thermal contact resistance, and the nanosilver layer will yield a bulk thermal resistance. Totally different from the IGBT modules, the high temperature will reduce the thermal contact resistance and thus reduce the total thermal resistance of PP IGBTs. This characteristic will improve the reliability of PP IGBTs because the increment of thermal resistance after power cycling is one of the main failure modes of IGBT modules, which is caused by the degradation of the die attach. Meanwhile, the nanosilver sintering technology can greatly reduce the thermal contact resistance within PP IGBTs and improve the reliability of die attach with its high electric and thermal conductivity [33]. 5. Conclusions Various preliminary conclusions can be drawn via the calculation and measurement of the thermal contact resistance of the single FRD chip submodule. 1) The surface topography, such as the surface roughness and surface slope, has a great impact on the thermal contact resistance. Though reducing the surface roughness can greatly improve the thermal characteristics, the surface slope also decreases with the roughness which, in turn, increases the thermal contact resistance. At the same time, the cost will sharply increase. Thus, in actual applications, a comprehensive consideration should be taken regarding the reduction of thermal contact resistance. 2) The clamping force is the most important parameter that influences the thermal contact resistance. Thermal contact resistance sharply Table 5 Thermal contact resistance comparison.
Before [K/W] After [K/W]
FEM
Measurement
Deviation [%]
0.1192 0.093
0.142 0.112
16.06 16.96
The work was supported by the National Natural Science Foundation of China [51477048], the National Key Research and Development Program of China [2016YFB0901803], and the Science and Technology Project of State Grid [5455DW150005]. References [1] F. Wakeman, D. Hemmings, W. Findlay, G. Lockwood, Pressure Contact IGBT, Testing for Reliability, Westcode Semiconductors Ltd., March 2012 [2] C. Busca, R. Teodorescu, F. Blaabjerg, et al., Dynamic Thermal Modelling and Analysis of Press-pack IGBTs Both at Component-level and Chip-level [C]//Industrial Electronics Society, IECON 2013 - 39th Annual Conference of the IEEE, IEEE, 2013 677–682. [3] Z. Ping, Y. Xuan, L.I. Qiang, Development on thermal contact resistance [J], CIESC J. 4 (2012) 11–20. [4] M. Yovanovich, E. Marotta, Thermal Spreading and Contact Resistances [M], Wiley, New Jersey, 2003 261–395. [5] M.A. Lambert, L.S. Fletcher, Thermal contact conductance of spherical, rough metals [J], Trans. Asme Serie C J. Heat Transf. 119 (4) (1997) 684–690. [6] Y.Z. Li, C.V. Madhusudana, E. Leonardi, On the enhancement of the thermal contact conductance: effect of loading history [J], J. Heat Transf. 122 (1) (2000) 46–49. [7] T. Mcwaid, E. Marschall, Thermal contact resistance across pressed metal contacts in a vacuum environment [J], Int. J. Heat Mass Transf. 35 (11) (1992) 2911–2920. [8] M. Bahrami, Thermal Resistances of Gaseous Gap for Conforming Rough Contacts [C]//42nd AIAA Aerospace Sciences Meeting and Exhibit, 2004 (2002–0317). [9] Astm, Astm D5470-06: Standard Test Method for Thermal Transmission Properties of Thermally Conductive Electrical Insulation Materials [S], Astm International, America, 2006. [10] Y. Osone, T. Kubo, N. Nakazato, Optical Measurement of Thermal Contact Conductance Between Wafer-Like Thin Solid Samples [J], Trans. Asme Serie C J. Heat Transf. 121 (4) (1999) 954–963. [11] L. Shi, Investigation of Heat Transport on the Solid-Solid Contact Interface at Low Temperature [D], Huazhong University of Science & Technology, Wuhan, 2006. [12] O. Kwon, L. Shi, A. Majumdar, Scanning thermal wave microscopy [J], J. Heat Transf. 125 (2003) 156–163. [13] T. Poller, J. Lutz, S. D'Arco, et al., Determination of the thermal and electrical contact resistance in press-pack IGBTs [J], Mikrobiologiia 75 (3) (2006) 305–311. [14] M. Grujicic, C.L. Zhao, E.C. Dusel, The effect of thermal contact resistance on heat management in the electronic packaging [J], Appl. Surf. Sci. 246 (1–3) (2005) 290–302. [15] M. Bahrami, Thermal Resistances of Gaseous Gap for Conforming Rough Contacts [M], 2004. [16] W.H. McAdams, Heat Transmission, McGraw-Hill, New York, 1954. [17] A. Bejan, D. Kraus, Heat Transfer Handbook, John Willey, New York, 2003. [18] J.B.P. Williamson, J. Pullen, R.T. Hunt, The shape of solid surfaces, Surface Mechanics, ASME, New York 1969, pp. 24–35. [19] J.A. Greenwood, The area of contact between rough surfaces and flats, J. Lubr. Technol. (January 1967) 81–91. [20] M. Bahrami, J.R. Culham, M.M. Yovanovich, et al., Review of thermal joint resistance models for non-conforming rough surfaces in a vacuum [C]//ASME 2003 heat transfer summer conference, Am. Soc. Mech. Eng. (2003) 411–431. [21] L.H. Tanner, M. Fahoum, A study of the surface parameters of ground and lapped metal surfaces, using specular and diffuse reflection of laser light, Wear 36 (1976) 299–316. [22] V.W. Antonetti, T.D. Whittle, R.E. Simons, An approximate thermal contact conductance correlation, ASME, The 28th National Heat Transfer Conference Minneapolis, Minnesota, HTD-Vol. 170, Experimental/Numerical Heat Transfer in Combustion and Phase Change 1991, pp. 35–42. [23] M.A. Lambert, Thermal Contact Conductance of Spherical, Rough Metals(Ph.D. Thesis) Dept. of Mech. Eng., Texas A&M University, USA, 1995. [24] M. Bahrami, J.R. Culham, M.M. Yovanovich, G.E. Schneider, Thermal contact resistance of non-conforming rough surfaces. Part 1. Mechanical model, Proceedings of
Please cite this article as: E. Deng, et al., Optimization of the thermal contact resistance within press pack IGBTs, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.01.003
12
[25]
[26]
[27]
[28]
E. Deng et al. / Microelectronics Reliability xxx (2017) xxx–xxx the 36th AIAA Thermophysics Conference, AIAA Paper No. 2003-4197, June23–26, Orlando, FL, 2003. M. Bahrami, J.R. Culham, M.M. Yovanovich, G.E. Schneider, Thermal contact resistance of non-conforming rough surfaces. Part 2. Thermal model, Proceedings of the 36th AIAA Thermophysics Conference, AIAA Paper No. 2003-4198, June 23–26, Orlando, FL, 2003. S. Song, M.M. Yovanovich, Relative contact pressure: dependence upon surface roughness and Vickers microhardness, J. Thermophys. Heat Transf. 2 (1) (January 1988) 43–47. M.M. Yovanovich, New Contact and Gap Correlations for Conforming Rough Surfaces, AIAA-81-1164, Presented at AIAA 16th Thermophysics Conference, Palo Alto, CA, June 1981. V.W. Antonetti, M.M. Yovanovich, Thermal Contact Resistance in Microelectronic Equipment, Thermal Management Concepts in Microelectronic Packaging From Component to System, ISHM Technical Monograph Series 6984-003, 1984 135–151.
[29] P. Cova, et al., On the effect of power cycling stress on IGBT modules, Microelectron. Reliab. 38 (6–8) (1998) 1347–1352. [30] U. Scheuermann, R. Schmidt, Investigations on the vce(t) - method to determine the junction temperature by using the chip itself as sensor, PCIM, 2009. [31] G.Q. Lu, J.N. Calata, G. Lei, et al., Low-temperature and Pressureless Sintering Technology for High-performance and High [C]//2007 International Conference on Thermal, Mechanical and Multi-Physics Simulation Experiments in Microelectronics and Micro-Systems. EuroSime, 2007, 2007 1–5. [32] G. Lu, Nanosilver paste: its sintering behavior and applications for chip interconnection, IEEE Int. Symp. 755 (3) (1998) 298-298. [33] N. Heuck, K. Guth, M. Thoben, et al., Aging of new Interconnect-Technologies of Power-Modules during Power-Cycling [C]//Integrated Power Systems (CIPS), 2014 8th International Conference on VDE, 2014 1–6.
Please cite this article as: E. Deng, et al., Optimization of the thermal contact resistance within press pack IGBTs, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.01.003