MR-12052; No of Pages 5

Microelectronics Reliability xxx (2016) xxxxxx

Contents lists available at ScienceDirect

Microelectronics Reliability
journal homepage: www.elsevier.com/locate/mr

Novel heatsink for power semiconductor module using high thermal

conductivity graphite
Y. Yamada a,, M. Yanase a, D. Miura a, K. Chikuba b
a
b

Department of Electrical and Electronic Engineering, Daido University, Nagoya, Aichi 457-8530, Japan
Thermo Graphitics Co.,Ltd. Osaka, Osaka 551-0031, Japan

a r t i c l e

i n f o

Article history:
Received 23 June 2016
Accepted 6 July 2016
Available online xxxx
Keywords:
Power module
Heatsink
Graphite
Reliability
Laminate

a b s t r a c t
The thermal properties and reliability of novel heatsinks that use high thermal conductivity graphite were
investigated. Graphite plates with different high-thermal-conductivity directions were laminated together
using an Ag-based brazing material, with thin Cu plates on their outer surfaces. The heatsinks were bonded to Si heater chips using Sn-3Ag-0.5Cu solder. Samples with conventional Cu or Cu-65Mo heatsinks
were also fabricated as references. The samples were attached to an active cooling plate subjected to a
constant water ow, and thermal and reliability measurements were conducted. The experimental results
were also compared with the results of a nite element analysis. The novel laminated heatsinks exhibited
a lower thermal resistance than the Cu or Cu-65Mo heatsinks, and the experimental results were in reasonable agreement with those of the nite element analysis. The graphite-based heatsinks had better
power cycle reliability than Cu-based heatsinks. Therefore, these novel graphite heatsinks have potential
for application to power semiconductor modules, it seems to be useful for applications with high heat ux
of power semiconductor devices.
2016 Published by Elsevier Ltd.

1. Introduction
Next-generation power semiconductor devices such as wide bandgap SiC or GaN devices are being actively developed to achieve more efcient and smaller electrical invertors. Important advantages of such
devices are their reduced size and their ability to operate at temperatures in excess of 200 C; however, such features can lead to high heat
ux or signicant thermal stress [16].
Heatsinks for power semiconductor modules generally comprise
materials such as Cu, Al, AlSiC, and CuMo. Cu and Al have excellent thermal conductivity but a large coefcient of thermal expansion (CTE) mismatch with semiconductor materials. AlSiC and CuMo have lower
thermal conductivity but a small CTE mismatch with semiconductor
materials. There is thus a general trade-off between the thermal conductivity and CTE for these heatsink materials for power semiconductor
packaging.
Graphite has excellent thermal conductivity in excess of 1000 W/mK
[710]; however, the thermal conductivity is anisotropic, being high in
two directions and signicantly lower in the third, as shown in Fig. 1.
The CTE is also low for the rst two directions and high for the third,
as shown in Fig. 2.

Corresponding author.
E-mail address: yamada-y@daido-it.ac.jp (Y. Yamada).

Graphite has disadvantages in terms of higher cost, due to

the prolonged high-temperature deposition process required
for its fabrication, and it exhibits brittleness under mechanical
impact.
In this present paper, laminated heatsinks using graphite plates with
different in-plane directions were fabricated, and their thermal properties and power cycle reliability were investigated.

2. Experimental
2.1. Sample structure
Graphite plates were fabricated from hydrocarbon gas using thermal
decomposition at temperatures greater than 2000 C. For one of the
plates (HT-x, Fig. 3(a)), the thermal conductivity was high and the
CTE was low in the Y and Z directions, with the opposite being the
case in the X direction. Another plate (HT-y, Fig. 3(b)), which was rotated by 90 with respect to HT-x, had a high thermal conductivity and a
low CTE in the X and Z directions. A simple laminated heatsink with a
single HT-x and HT-y plate would not exhibit symmetric thermal expansion and may deform with changing temperature. In contrast, a
heatsink with a symmetrical laminated structure (HT3) comprising
three layers of graphite, as shown in Fig. 4, would be expected to be stable with respect to temperature change. Therefore, this structure was
used in the present study.

http://dx.doi.org/10.1016/j.microrel.2016.07.029
0026-2714/ 2016 Published by Elsevier Ltd.

Please cite this article as: Y. Yamada, et al., Novel heatsink for power semiconductor module using high thermal conductivity graphite, Microelectronics Reliability (2016), http://dx.doi.org/10.1016/j.microrel.2016.07.029

Y. Yamada et al. / Microelectronics Reliability xxx (2016) xxxxxx

Fig. 1. Thermal conductivity of graphite (HT-x).

Fig. 2. Coefcient of thermal expansion (CTE) of graphite (HT-x).

Fig. 3. CTE and thermal conductivity of graphite HT-x and HT-y (the values are at 200 C).

Fig. 4. Laminated structure of fabricated heatsink.

Please cite this article as: Y. Yamada, et al., Novel heatsink for power semiconductor module using high thermal conductivity graphite, Microelectronics Reliability (2016), http://dx.doi.org/10.1016/j.microrel.2016.07.029

Y. Yamada et al. / Microelectronics Reliability xxx (2016) xxxxxx

temperature was about 800 C and the pressure was several tens
of megaPascals. The outer plates had a thin Cu coating to enable
easy soldering to a Si heater chip. A conventional Sn-3Ag-0.5Cu
solder was used for this process. The assembled heatsink was attached to an active cooling plate using a thermal interface material
(TIM).
2.2. Thermal properties
To measure the thermal properties of the heatsinks, conventional
thermal grease (4.5 W/mK) was used as the TIM. The thickness of the
grease was maintained at 0.06 mm using metal plates located on either
side of the heatsink. The active cooling plate was cooled with water at a
constant ow rate and a temperature of 65 C, as shown in Fig. 6. A DC
voltage was applied to the Si heater chip, and the chip temperature
was measured using an electrically isolated 0.2 mm diameter thermocouple [11].
A steady-state thermal analysis was also conducted using a three-dimensional nite element method (FEM). In the analysis, a constant heat
ux was applied to the surface of a heater chip attached to a heatsink
and an active cooling water (thermal transfer coefcient of 9000 W/
m2 K). The thermal resistance of the heatsink was calculated based on
the temperature difference between the heater chip and the cooling
plate. The physical properties of the components used for the FEM analysis are shown in Table 1.
Fig. 7(a) compares the measured thermal resistance of the HT3
heatsink sample with that of Cu and Cu-65Mo heatsinks. It can be
seen that HT3 exhibits a lower thermal resistance than the other
types of heatsink. Fig. 7(b) shows the corresponding results obtained by the FEM analysis, which are seen to be in reasonable
agreement with the experimental results. A breakdown of the
thermal resistance of each component is shown in Fig. 8. The
thermal resistance of the HT3 heatsink component is signicantly

Fig. 5. Sample structure.

Fig. 5 shows a schematic diagram of the experimental setup.

Each heatsink was fabricated using laminated graphite plates
bonded together with an Ag-based brazing material. The bonding

Fig. 6. Measurement apparatus.

Table 1
Physical properties for simulation.
Component

Material

Chip
Solder
Heatsink

Si
Sn-3Ag-0.5Cu
Cu
Cu-65Mo
HT-x (200 C)
HT-y (200 C)
Grease, sheet
Al alloy

CTE
[ppm/K]
X

TIM
Cooling plate

3
22
17
8
27
0.3

Young's modulus
[GPa]
Y

0.3
27

0.3
0.3

Thermal conductivity
[W/mK]
X

160
38
120
220
50
50

160
70
400
207
4.5
1150
4.5, 1
140

1150
4.5

1150
1150

Poisson's ratio:0.3.

Please cite this article as: Y. Yamada, et al., Novel heatsink for power semiconductor module using high thermal conductivity graphite, Microelectronics Reliability (2016), http://dx.doi.org/10.1016/j.microrel.2016.07.029

Y. Yamada et al. / Microelectronics Reliability xxx (2016) xxxxxx

Fig. 7. Thermal resistances.

less than the Cu and Cu-65Mo heatsink components. Fig. 9 compares the thermal resistance of various heatsink structures,
where HT1, HT2, HT3, and HT5 represent one, two, three, and
ve graphite layers, respectively. The lower resistance expected
for HT2, HT3, and HT5 than Cu or Cu-65Mo is conrmed in this
gure. Thus, laminated graphite heatsinks exhibit excellent thermal properties.

Fig. 8. Simulated thermal resistances for each layer.

2.3. Power cycle reliability

The samples were subjected to power cycle tests with temperature
ranges of 65/150 C and 65/175 C for up to 3000 cycles [5]. The ON
and OFF periods were 20 and 40 s, respectively, and were controlled
by a DC power supply with a computer. A thermal sheet (1 W/mK,
1 mm thick) composed of silicon was used as a TIM between the
heatsink and active cooling plate. The reliability was evaluated based
on the electrical resistance change of the Si heater chip because the Pt
heater had good temperature linearity.
The resistance of a sample with a Cu heatsink increased rapidly in
the 65/150 C test, whereas no change was observed for the sample
with a HT3 heatsink, as shown in Fig. 10(a). A signicant increase in resistance was observed for the HT3 heatsink in the 65/175 C test, and the
heater chip became disconnected; however, no change was observed
for the Cu-65Mo heatsink, as shown in Fig. 10(b). The increases in resistance are attributed to crack propagation in the solder layers due to excess thermal stress during the cycling tests.
Fig. 11 shows the von Mises stress for the solder joint calculated by a
three-dimensional thermo-mechanical analysis. The physical values are
shown in Table 1 and the software used was ANSYS 15.0, as with the
thermal analysis. The temperature matrix calculated by thermal simulation was used for a structural analysis. Fig. 11 also includes the results
calculated for thermal cycles. The order of the simulated results corresponds well with the experimental results.

2.4. Discussion of application

The HT3 shows higher thermal performance than Cu and Cu-65Mo,
therefore it is useful for applications with high heat ux. On the other
hand, the power cycle reliability is less than Cu-65Mo, consequently
the HT3 is not appropriate for applications with high thermal stresses,
such as high temperature operation or large CTE mismatch packages.
Another point of view, thermal conductivities are decreased in higher
temperature range shown in Fig. 1. Practically, the HT3 seems to be useful for applications with high heat ux and usual temperature range.

3. Summary

Fig. 9. Simulated thermal resistances for various structures.

A novel laminated graphite heatsink was found to exhibit excellent

thermal properties and satisfactory power cycle reliability. The estimated thermal conductivity and CTE for the HT3 and other heatsink

Please cite this article as: Y. Yamada, et al., Novel heatsink for power semiconductor module using high thermal conductivity graphite, Microelectronics Reliability (2016), http://dx.doi.org/10.1016/j.microrel.2016.07.029

Y. Yamada et al. / Microelectronics Reliability xxx (2016) xxxxxx

Fig. 10. Results of power cycle test.

materials are shown in Fig. 12. The results indicate that the HT3 heatsink
is in the cutting-edge region.
References

Fig. 11. Calculated stress of thermo-mechanical simulation TCT (Thermal Cycle Test), PCT
(Power Cycle Test).

[1] P.G. Neudeck, R.S. Okojie, L.-Y. Cchen, High-temperature electronics-a role for wide
bandgap, Proc. IEEE 90 (6) (2002) 10651076.
[2] P. Dietrich, Trends in automobile power semiconductor packaging, Microelectron.
Reliab. 53 (2013) 16811686.
[3] P. Dietrich, Joining and package technology for 175 C Tj increasing reliability in automotive applications, Microelectron. Reliab. 54 (2014) 19011905.
[4] T. Ishizaki, M. Yanase, A. Kuno, T. Satoh, M. Usui, F. Osawa, Y. Yamada, Themal simulations of joint with high thermal conductivities for power electronic device,
Microelectron. Reliab. 55 (2015) 10601066.
[5] T. Ishizaki, A. Kuno, A. Tane, M. Yanase, F. Osawa, T. Satoh, Y. Yamada, Reliability of
Cu nanoparticle joint for high temperature power electronics, Microelectron. Reliab.
54 (2014) 18671871.
[6] T. Ishizaki, M. Usui, Y. Yamada, Thermal cycle reliability of Cu-nanoparticle joint,
Microelectron. Reliab. 55 (2015) 18611866.
[7] Y. Yamada, Power semiconductor module using a bonding lm with anisotropic
thermal conduction, Microelectron. Reliab. 52 (2012) 24432446.
[8] J. Kovacik, S. Emmer, J. Bielek, Thermal conductivity of Cu-graphite composites, Int. J.
Therm. Sci. 90 (2015) 298302.
[9] A.A. Balandin, S. Ghosh, W. Bao, I. Calizo, D. Teweldebrhan, F. Miao, C.N. Lau, Superior thermal conductivity of single-layer graphene, Nano Lett. 8 (3) (2008) 902907.
[10] G. Fugallo, A. Cepellotti, L. Paulatto, M. Lazzeri, N. Marzari, F. Mauri, Thermal conductivity of graphene and graphite: collective excitations and mean free paths, Nano
Lett. 14 (11) (2014) 61096114.
[11] T. Ishizaki, T. Satoh, A. Kuno, A. Tane, M. Yanase, F. Osawa, Y. Yamada, Thermal characterizations of Cu nanoparticle joints for power semiconductor devices,
Microelectron. Reliab. 53 (2013) 15431547.

Fig. 12. Estimated property range of HT3.

Please cite this article as: Y. Yamada, et al., Novel heatsink for power semiconductor module using high thermal conductivity graphite, Microelectronics Reliability (2016), http://dx.doi.org/10.1016/j.microrel.2016.07.029