Thermal spreading resistance characteristics of a high power light emitting diode module

Applied Thermal Engineering 70 (2014) 361e368

Contents lists available at ScienceDirect

Applied Thermal Engineering journal homepage: www.elsevier.com/locate/apthermeng

Thermal spreading resistance characteristics of a high power light emitting diode module Kai-Shing Yang a, Chi-Hung Chung b, Cheng-Wei Tu a, Cheng-Chou Wong c, Tsung-Yi Yang b, Ming-Tsang Lee b, * a b c

Green Energy & Environment Research Laboratories, Industrial Technology Research Institute, Hsinchu 310, Taiwan, ROC Department of Mechanical Engineering, National Chung Hsing University, Taichung 402, Taiwan, ROC Material & Chemical Research Laboratories, Industrial Technology Research Institute, Hsinchu 310, Taiwan, ROC

h i g h l i g h t s
We performed an experimental and numerical study for heat transfer in a LED.
The thermal spreading resistance effect is significant in the LED module.
Lateral thermal conductivity of the substrate is critical to the spreading resistance.
The thermal spreading resistance effect increases with increasing of LED power.

a r t i c l e i n f o

a b s t r a c t

Article history: Received 14 February 2014 Accepted 10 May 2014 Available online 20 May 2014

In this study, effects of the dimensions and the thermal conductivity of the substrate on the heat transfer characteristics of a LED module are investigated. The total thermal resistance corresponding to a LED module operating at different power levels is measured using a method following JESD51-1 and JESD5114 standards. In addition, a finite element method (FEM) numerical simulation is carried out to analyze the heat transfer phenomena in the LED module. It is found that, for the current experimental conditions, the importance of the thermal spreading resistance effect increases with decreasing substrate thickness and/or increasing input power of the LED module, which corresponds to an increase in the total thermal resistance and correspondingly a higher chip temperature. Experimental and numerical results show that the thermal spreading resistance and thus the chip temperature can be reduced by increasing the substrate thickness or by utilizing materials with high lateral thermal conductivities (directionallydependent) for the substrate. In consequence, for LED modules with the same substrate thickness, using graphite composite to replace aluminum as the substrate material reduces the spreading resistance by nearly 14% in the current study. © 2014 Elsevier Ltd. All rights reserved.

Keywords: LED Electronic package heat transfer Thermal resistance measurement Spreading resistance

1. Introduction The development of light emitting diodes (LEDs) systems has received a lot of attention due to requirements for energy efficient lighting sources in a variety of applications worldwide. In addition to high energy efficiency, LEDs also are preferred for their fast response and low environmental impact [1,2]. The performance of LEDs in terms of light output quality has improved significantly and efforts have been made to replace traditional lighting sources with

* Corresponding author. E-mail address: [email protected] (M.-T. Lee). http://dx.doi.org/10.1016/j.applthermaleng.2014.05.028 1359-4311/© 2014 Elsevier Ltd. All rights reserved.

LEDs in many countries [3]. However, the light emitted from LEDs degrades over time [4e8]. Possible reasons contributing to the light degradation of LEDs include thermal-induced deterioration of the encapsulation, die-attach [9], reflector and lead wires, as well as impurities and crystalline defects [5,10,11]. The chip temperature has a significant impact on these factors. In addition, nearly 80% of the energy input to a LED can be dissipated as heat [12]. Thus, thermal management is critical to attain high efficient, high power and long lasting LEDs [13e15]. Thermal management of a LED system includes two major factors: packaging and system performance [16]. In addition to the design and optimization of heat sinks such as fin arrays [17e19] and single phase liquid cooling device [20], advanced

362

K.-S. Yang et al. / Applied Thermal Engineering 70 (2014) 361e368

2. Experimental apparatus and measurements Nomenclature A hfc Kf k Pi 000 q_ Q R Rt Ta Tj T t

in-plane surface area [m2] convective heat transfer coefficient [W m1 K1] K factor [K V1] thermal conductivity [W m1 K1] input electrical power [W] volumetric heat source [W m3] input power as heat [W] thermal resistance [ C W1] total thermal resistance of the LED module [ C W1] ambient temperature [ C] junction (chip) temperature [ C] average temperature [ C] thickness [m]

Greek symbols 3 aspect ratio DTj change in the LED's junction (chip) temperature [ C] DVf change in the LED's forward voltage [V] Subscripts p substrate s source

thermal management techniques have been studied for the system and/or the heat sink section for LEDs. For example, Wang et al. [21,22] utilized a vapor chamber based plate to successfully cool a high power LED module. Oscillating heat pipes have also been applied to improve the thermal performance of two-phase flow chamber based heat sinks [23]. Chen et al. [24] applied ionic wind to induce convection and thus promote the rate of heat transfer for a LED module. The package aspect of thermal management involves design of the structure, selection of materials for chip, die-attach and substrate [25]. Heat generated in the chip is transferred through these layered structures. In practice, the thickness of these layers is much thinner than the thickness of the heat sink. However, the package does contribute to a significant fraction of the overall thermal resistance due to low thermal conductivities and small surface areas that result from the fabrication and electrical insulation limitations. In addition, studies of the thermal spreading resistance have emphasized varying the ratio of the chip (heat source) area to the substrate area (heat sink), i.e. the aspect ratio [26e31]. It was found that the thermal spreading resistance effect can be significant for electronic packaging should be included in LED package design. However, there is lack of study for the spreading resistance effects with respect to the power of the heat source (chip), which is one of the issues to be addressed in the present work. In this study, experiments and numerical analysis were carried out to investigate the thermal performance of LED modules, especially for high power LEDs. The effects of substrate materials and thicknesses on the thermal spreading resistances are discussed based on the results. In particular, a new approach to study the thermal spreading resistance effects was designed and conducted on a LED module with an anisotropic medium (directionallydependent thermal conductivities) as the substrate material and adjustable input power; this contrasts with the usual approach where isotropic materials (same thermal conductivity in all directions) are used, and the aspect ratio is the controlling parameter. Recommendations for the design of the LED thermal spreader are provided.

A schematic of the experiment is shown in Fig. 1(a). A thermal resistance testing system (T3Ster®) was utilized in conjunction with a data acquisition system, booster extension box (for boosting high power LEDs) and power supply. A thermoelectric cooler (TEC model: Arroyo instruments 5310-TEC Source and 286-TEC Mount, accuracy: ±0.004  C) was used to maintain the temperature of the LED substrate (bottom surface) at 25  C. The detailed measurement procedure is provided in a previously published work [32]. A chipon-board structured LED module with adjustable input power from 10 to 50 W was used as the testing module. A photograph of the testing LED is shown in Fig. 1(b). Geometrical configurations and materials of different layers of the LED module are presented in Table 1 and in Fig. 3. Four substrates were tested: aluminum plates with 0.9 mm, 1.1 mm, 1.6 mm thicknesses and a graphite composite plate with a 1.6 mm thickness. The graphite composite plate was fabricated at the Industrial Technology Research Institute in Taiwan. The graphite composite plate is an anisotropic medium, where the thermal conductivities in the x, y and z directions are 503.1, 531.4 and 178.3 W m1 K1, respectively (Fig. 3 shows the corresponding directions). Thermal conductivity measurements were carried out using an LFA 447 NanoFlash™ system (range: 0.1 W m1 K1 to 2000 W m1 K1, accuracy: ±5%, repeatability: ±3%) in the Material & Chemical Research Laboratories of Industrial Technology Research Institute in Taiwan. Measurements of the thermal performance of the LED module were carried out for power inputs from 10 to 50 W, with 5 W intervals, for all four substrates. A thermal pad with a thermal conductivity k ¼ 2.8 W m1 K1 is used between the LED and substrate to reduce the thermal contact resistance. The principle of the measurements is that of the Joint Electron Device Engineering Council (JEDEC) standard. The electrical test method (ETM) is based on the JESD51-1 standard [33], and the transient dual interface measurement (TDIM) for the thermal resistance is based on the JESD51-14 standard [34]. First, the temperature sensitivity parameter (TSP) noted as the K-factor was calibrated for every LED module to be tested. The K-factor is defined as:

Kf ¼

DTj DVf

(1)

where DTj is the change in the junction (chip) temperature and DVf is the change in the forward voltage of the LED, respectively. In the present study, K-factors were calibrated with 1 mA bias current from 25 to 115  C, and the results are shown in Fig. 2. Note that in the temperature range of the current study junction temperature varies linearly with the forward voltage. K-factors for those four tested LED-substrate modules are listed in Fig. 2. All four cases have comparable K-factors since the same LED was utilized on all the substrates. This result confirms the repeatability of the K-factor measurement. Based on the calibrated K-factor, the junction temperature variation with respect to time (i.e., the cooling curve) during the thermal resistance measurement can be obtained from the recorded transient forward voltage data. The thermal resistance of each layer of the LED module from the chip to the environment can then be identified and determined from the structure function that is derived from the cooling curve. Uncertainty analysis for the thermal resistance measurements was performed according to the specifications of instrument and the experimental conditions [35]. The uncertainty in the thermal resistance measurement results is estimated to be less than 5.1%. Detailed information for the test method of the thermal resistance of semiconductor devices with heat flow through a single path utilized with the T3Ster® system is

K.-S. Yang et al. / Applied Thermal Engineering 70 (2014) 361e368

363

Fig. 1. (a) Schematic of the experimental setup; (b) picture of the testing LED module.

presented in the JEDEC Standard Document (JESD51-14), technical references [36e40], and in our previously published work [32]. 3. Numerical analysis A numerical analysis for the thermal field was carried out to assist in understanding the heat transfer phenomena in the high power LED module. The LED module is symmetrical in the x and y

directions, as shown in Fig. 1(b). Therefore, a quarter section is analyzed as shown in Fig. 3. As previously noted, the thermal conductivity of the graphite composite utilized as the substrate is directionally-dependent; thus, the equation for the conservation of energy is:




 000 v vT v vT v vT kx þ ky þ kz þ q_ ¼ 0 vx vx vy vy vz vz

(2)

364

K.-S. Yang et al. / Applied Thermal Engineering 70 (2014) 361e368

Table 1 Dimensions and materials of package layers of the testing LED module. Layer

Material

Size (mm)

Thermal conductivity (W m1 K1)

Encapsulant Chip

Silicon InGaN

0.19 33

Die-attach Submount Copper foil

Eutectic AlN Cu

Dielectric layer Substrate

Epoxy

Radius ¼ 4.25 Length*width*thickness ¼ 1.1*1.1*0.17 Thickness ¼ 0.01 9*9*0.5 Length*width*thickness ¼ 32*32*0.035 Length*width*thickness ¼ 32*32*0.08 Length*width ¼ 32*32; thickness ¼ 0.9, 1.1, 1.6 Length*width ¼ 32*32; thickness ¼ 1.6

Al Graphite composite

Thermal pad

Length*width*thickness ¼ 35*35*0.3

58 120 385 1.2 160 kx ¼ 503.1 ky ¼ 531.4 kz ¼ 178.3 2.8

Thermal conductivities of the materials used in this study are 000 given in Table 1. Note that the volumetric heat source term, q_ , applies only on the 4  4 chip array (cf. Fig. 3) and is assumed to be uniformly distributed in each chip unit. As previously noted, a thermoelectric cooler (TEC) composed of an electrical cooler and a fan was used to cool the LED module during the experiment. Therefore, the forced convection boundary condition is specified on the bottom surface of the substrate. Natural convective heat transfer boundary conditions are applied for the other surfaces of the LED module with the convective heat transfer coefficient obtained from Ref. [41]. 4. Results and discussion 4.1. Experimental results Experimental results for thermal resistances of the aluminum substrate with different thicknesses are shown in Fig. 4. The thermal resistance Rt measured with the T3Ster® testing system is defined as:

  Rt ¼ Tj  Ta Pi

(3)

Fig. 3. Schematic of the modeling domain of the LED module.

system [32]. From the experimental results, thermal resistance increases with decreasing thickness of the substrate because of the spreading resistance effect. For example, at an input power of 10 W, the thermal resistance of the aluminum substrate with 1.6 mm thickness is 1.3 K W1, which is 41% lower than that of the 0.9 mm thick aluminum substrate. Chen et al. [28,29] investigated the effect of thermal resistance in a typical electronic packaging that includes a rectangular thin-plate heat source and a rectangular heat spreader plate. In their analysis, an aspect ratio (3 ) is defined as pffiffiffiffiffiffiffiffiffiffiffiffiffi 3 ¼ As =Ap , where As and Ap are the in-plane surface areas of the heat source (i.e. chip) and the spreader plate (i.e. substrate), respectively. For a small (concentrated) heat source attached to a relatively large substrate as shown in Fig. 5(a), two effects contribute significantly to the overall thermal resistance of the spreader plate: (1) the “one-dimensional” conduction resistance in the z direction and, (2) the spreading resistance in the lateral direction (xey planes) [27e30]. The thermal spreading resistance, Rspreading, is defined by the temperatures on the heated surface as shown in Eq. (4) [28]:

  Rspreading ¼ Tða; b; tÞ  Tðz ¼ tÞ Q

(4)

Note that results of Rt were obtained from the structure function measured by using the standard procedure of the T3Ster® testing 3.5 3.0

Thickness:1.6mm Thickness:1.1mm Thickness:0.9mm

2.5 2.0 1.5 1.0 0.5 0.0 10

20

30

40

50

Input power (W)

Fig. 2. Results for the K-factor calibration.

Fig. 4. Thermal resistances for aluminum substrate with different thicknesses (experimental results).

K.-S. Yang et al. / Applied Thermal Engineering 70 (2014) 361e368

365

Fig. 5. (a) Schematic of a heat spreader plate with a relatively small heat source placed in the center. (b) Heat flux plot (adiabats) of a heat source on a substrate with an aspect ratio ¼ 1. (c) Heat flux plot of a heat source on a substrate with an aspect ratio 3 < 1. (d) Heat flux plot of a heat source on a substrate with an aspect ratio 3 < 1 and the substrate thicker than the substrate in (c).

3

where T(a,b,t) is the temperature at the center on the surface of the substrate attached to the heat source (cf. Fig. 5(a)). Fig. 5(b)e(d) schematically illustrates the spreading resistance effect. The side walls in the lateral directions are assumed to be adiabatic (omitting the small heat transfer associated with the small area and small natural convection effect). For an aspect ratio of unity (cf. Fig. 5(b)), the spreading resistance is approximately zero because the temperature distribution on the top surface of the substrate (z ¼ t) is uniform. The characteristic length for conduction in the substrate then equals the thickness of the substrate (t); thus, the overall thermal resistance equals the one-dimensional (z-direction) thermal resistance. For an aspect ratio less than one, i.e. a concentrated heat source, the “paths for the heat transfer” (adiabats) in the substrate are distorted as shown in Fig. 5(c) and (d) [31]. The adiabats are closely spaced near the heat source and more broadly spaced near the cooler surface on the substrate. The temperature distribution on the top surface of the substrate is non-uniform; thus the spreading resistance is non-zero according to its definition by Eq. (4). The characteristic lengths for conduction in the substrate are larger than the thickness of the substrate (t), thus increasing the overall thermal resistance [28,31]. Vermeersch et al. [42] studied the dependence of thermal resistance on the convective heat transfer coefficient (h) in typical electronic packaging. Their results indicated that, for common single phase cooling technologies used in electronic packaging systems, the thermal spreading effect should be considered to prevent underestimating the overall thermal resistance. Furthermore, the one-dimensional conduction resistance increases proportional to the substrate thickness, while the spreading resistance decreases with increasing substrate thickness because the adiabats are less distorted and shorter as illustrated in Fig. 5(d). Previous studies [28,29,31] indicated that, for a concentrated heat source, the overall thermal resistance of the substrate first decreases with an increase in its thickness when the spreading resistance is dominant and before a critical thickness is reached. As illustrated in Fig. 5(c) and (d), the spreading resistance

effect is more significant for thinner substrates because the adiabats extend more in the lateral direction. This profile of the adiabats results from the small convection on the bottom surface of the substrate which causes more heat to be conducted in the lateral direction. The extended lateral profile of the adiabats and the correspondingly increased spreading resistance can be reduced by increasing the thickness of the substrate as shown in Fig. 5(d) [29]. The thermal spreading resistance reaches a constant value at a critical thickness of the substrate so that further increasing the substrate thickness leads to no further change in the lateral extension of the adiabats [28,31]. Thus after the critical thickness is reached, further increases in the substrate thickness will increase the overall thermal resistance because the one-dimensional

2.0 1.8

Aluminum Graphite

1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0 10

20

30

40

50

Input power (W) Fig. 6. Thermal resistances for aluminum and graphite substrates (experimental results).

366

K.-S. Yang et al. / Applied Thermal Engineering 70 (2014) 361e368 400 Aluminum_Experiment Aluminum_Simulation Graphite_Experiment Graphite_Simulation

Temperature (K)

380

360

340

320

300 0

10

20

30

40

50

60

Input power (W) Fig. 7. Comparison of the numerical results for the chip temperature with the experimental results for aluminum and graphite composite substrates.

thermal resistance is now dominant. For the LED module in the present study, the overall thermal resistance decreases with increasing thickness, which indicates that the spreading resistance is dominant. In this case, the spreading resistance and thus the overall thermal resistance can be reduced by increasing the substrate thickness or by increasing the thermal conductivity of the substrate, especially in the lateral directions. The effect of increasing the lateral direction thermal conductivity will be discussed in a later section. In the present study the input electrical power is increased from 10 to 50 W. From the experimental results as shown in Fig. 4, increasing input power increases the overall thermal resistance, which is attributed to the increasing spreading resistance since it is the dominant resistance. Thus, it is concluded that, for the current experimental condition, the spreading resistance increases (adiabats extend more laterally) with increasing input power. To investigate the effect of thermal conductivity on the thermal resistance of the substrate, the aluminum plate of 1.6 mm thickness was replaced by a graphite plate of the same dimensions (same

width, length and thickness). As previously discussed, the thermal conductivities of the graphite composite are directionally dependent (cf. Table 1). The lateral thermal conductivities (kx and ky) are approximately three times the thermal conductivity in the z-direction (kz). As can be seen in Fig. 6, thermal resistances of the graphite composite substrate are lower than that of the aluminum substrate for all input power levels used in the present study. The reductions in thermal resistance at input powers of 10 W and 50 W are 11.8% and 13.6%, respectively. Note that the thermal conductivity in the z-direction of the graphite composite substrate is similar to that of the aluminum substrate. Therefore, the onedimensional thermal resistances of these two substrates are similar. The reduction in the overall thermal resistance is thus attributed to the reduction in spreading resistance associated with the increased lateral thermal conductivities. This result is consistent with the results of Chen et al. [28], where the ratio of kz to kx (assuming kx ¼ ky) was varied from 0.1 to 1. Their results indicated that the spreading resistance can be reduced by increasing the lateral thermal conductivities. This is achieved by effectively spreading the heat from the source with high lateral thermal conductivity of the substrate. The temperature distribution is more greatly extended in the lateral direction, and thus the spreading resistance is decreased. One limiting case would be for very large lateral thermal conductivity, so that the heat would be very effectively spread laterally in the substrate. The temperature profile in the substrate then becomes essentially one dimensional (z-direction) and the spreading resistance becomes very small. 4.2. Numerical results A multi-physics finite element computational software program (COMSOL®) was utilized to further understand the transport phenomena in the LED module. The computed domain is shown in Fig. 3. Testing of the stability of the numerical solution to mesh size was performed. It was found that the numerical results of highest (chip) temperature are 384.89 K, 384.71 K and 384.72 K for a total number of 420,298 elements, 986,817 elements and 2,634,036 elements, respectively. The latter two cases differ by less than 0.0026%. Thus, the mesh system with 986,817 elements was used for the simulation. Natural convection boundary conditions with

Fig. 8. Top view of the temperature profile for the LED module at 50 W input power (a) aluminum substrate. (b) Graphite composite substrate.

K.-S. Yang et al. / Applied Thermal Engineering 70 (2014) 361e368

the environment temperature of 25  C were used on the outer surfaces of the LED module except for the bottom surface of the substrate, where a TEC was utilized to extract heat from the substrate. The TEC is composed of an electrical cooler and a fan. Therefore, an effective forced convection boundary condition should be used to simulate the strong cooling effect. To obtain the convective heat transfer coefficient (hfc) for this boundary, a computation was carried out for the case of 10 W input power with an aluminum substrate of 1.6 mm thickness. With hfc of 3500 W m2 K1, the computed chip temperature is very close to the measured result, with an error of 0.4%. Therefore, this value of hfc was used. In addition, 70% of the input electrical power is assumed to be converted as heat and dissipated from the chip [32]. Fig. 7 shows a comparison of the numerical results for the chip temperatures with the experimental results for aluminum and graphite composite substrates at different input powers. There is generally good agreement among the results. For input powers from 10 to 30 W, the differences between the numerical and experimental results are less than 0.5%. The numerical results are slightly lower than the experimental results for high input powers (40 W and 50 W). It is possible that by increasing the input power and thus the operating temperature, the luminance efficiency of the LED decreases, and more input electrical power is converted to heat. Therefore, the waste heat dissipated from the chip was increased to 75% and 83% for 40 W and 50 W of the input power, respectively, and then the numerical analysis was repeated. Note that this assumption is within a reasonable range for common LED modules [32,43]. Consequently, the difference between numerical results and experimental data for the chip temperatures could be reduced to 0.5%. In future studies the luminance efficiency of the LED should be measured to attain better estimations of the waste heat. As shown in Fig. 7, the chip temperature increases with an increase of input power. At input powers from 10 to 20 W, the chip temperature of the LED module with the aluminum substrate is close to that of the graphite composite substrate. At higher input powers the difference between chip temperatures with different substrates becomes significant. Fig. 8 shows the temperature profile on the top surface of the substrate (cf. Fig. 3). The temperature distribution is more uniform for the graphite composite substrate due to its larger lateral thermal conductivities. Consequently, the spreading resistance is smaller for the graphite substrate, as shown in Fig. 9. Note that the thermal spreading resistance (cf. Eq. (4)) is calculated using the numerical results for the temperature. This

1.0 Aluminum Graphite 0.8

result is consistent with the experimental results as shown in Fig. 6, i.e., the thermal spreading resistance and thus the overall thermal resistance for the graphite substrate is lower than that for the aluminum substrate. It should be emphasized, again, that the onedimensional thermal resistances in the z-direction are similar for both materials. Thus, the thermal spreading resistance effect is important in the present cases, and yields large difference in the chip temperatures of these two substrates. 5. Conclusions In this study, experiments and numerical analyses were conducted for a LED module mounted on four types of substrate. The effects of substrate thickness and thermal conductivity are studied based on the measured and computed results. For aluminum substrates, the thermal resistance increases with decreasing thickness of the substrate. For the same thickness, the thermal resistance for graphite composite with anisotropic substrate is approximately 12e14% smaller than that of aluminum substrate, depends on the input power of the module. Consequently, the chip temperature of an LED module with a graphite substrate is lower than one with an aluminum substrate. In addition, the aspect ratio of the heat source to the substrate was fixed and the input power of the LED module was varied to directly investigate the effects of input power on the thermal spreading resistance. It is found that the spreading resistance and thus the chip temperature increase with increasing input power. Numerical results are in reasonably good agreement with the experimental data. At the same input power, the lateral temperature distribution for the larger lateral thermal conductivity graphite substrate is more uniform than the lateral temperature distribution for the smaller lateral conductivity aluminum substrate. These results indicate that the thermal spreading resistance effect is significant in the present LED configurations, and should be taken into account for the LED module design. It is also found that the effect of the thermal conductivity of the substrate is more important for high power LEDs. Thus, advanced materials with higher thermal conductivities (especially in the lateral directions) should be considered for high power LEDs. Acknowledgements Authors would like to acknowledge the supports from the Bureau of Energy, Ministry of Economic Affairs, Republic of Taiwan, and the Ministry of Science and Technology of Taiwan, R.O.C. (under project number NSC 103-2218-E-005-004). Dr. Ming-Tsang Lee wishes to express great thanks to Professor Ralph Greif at the University of California at Berkeley for discussions and suggestions to this study. References

0.6

Rspreading

367

0.4

0.2

0.0 10

20

30

40

50

Input power (W) Fig. 9. Thermal spreading resistances of the LED module (numerical results).

[1] H.-H. Wu, K.-H. Lin, S.-T. Lin, A study on the heat dissipation of high power multi-chip COB LEDs, Microelectron. J. 43 (2012) 280e287. [2] M.Y. Tsai, C.H. Chen, C.S. Kang, Thermal measurements and analyses of lowcost high-power LED packages and their modules, Microelectron. Reliab. 52 (2012) 845e854. [3] R.D. Dupuis, History, development, and applications of high-brightness visible light-emitting diodes, J. Lightwave Technol. 26 (2008) 1154e1171. [4] M. Dal Lago, M. Meneghini, N. Trivellin, G. Meneghesso, E. Zanoni, Degradation mechanisms of high-power white LEDs activated by current and temperature, Microelectron. Reliab. 51 (2011) 1742e1746. [5] R.-H. Horng, R.-C. Lin, Y.-C. Chiang, B.-H. Chuang, H.-L. Hu, C.-P. Hsu, Failure modes and effects analysis for high-power GaN-based light-emitting diodes package technology, Microelectron. Reliab. 52 (2012) 818e821. [6] J. Hu, L. Yang, W.J. Hwang, M.W. Shin, Thermal and mechanical analysis of delamination in GaN-based light-emitting diode packages, J. Cryst. Growth 288 (2006) 157e161.

368

K.-S. Yang et al. / Applied Thermal Engineering 70 (2014) 361e368

[7] J. Hu, L. Yang, M. Whan Shin, Mechanism and thermal effect of delamination in light-emitting diode packages, Microelectron. J. 38 (2007) 157e163. [8] M.-H. Chang, D. Das, P.V. Varde, M. Pecht, Light emitting diodes reliability review, Microelectron. Reliab. 52 (2012) 762e782. [9] Z. Li, Y. Tang, X. Ding, C. Li, D. Yuan, Y. Lu, Reconstruction and thermal performance analysis of die-bonding filling states for high-power light-emitting diode devices, Appl. Therm. Eng. 65 (2014) 236e245. [10] J.H. Choi, M.W. Shin, Thermal investigation of LED lighting module, Microelectron. Reliab. 52 (2012) 830e835. [11] M. Meneghini, M. Dal Lago, N. Trivellin, G. Mura, M. Vanzi, G. Meneghesso, E. Zanoni, Chip and package-related degradation of high power white LEDs, Microelectron. Reliab. 52 (2012) 804e812. [12] T. Cheng, X. Luo, S. Huang, S. Liu, Thermal analysis and optimization of multiple LED packaging based on a general analytical solution, Int. J. Therm. Sci. 49 (2010) 196e201. [13] A. Christensen, S. Graham, Thermal effects in packaging high power light emitting diode arrays, Appl. Therm. Eng. 29 (2009) 364e371. [14] H.K. Fu, Y.P. Peng, S.P. Ying, T.T. Chen, C.P. Wang, C.L. Chen, P.T. Chou, The evaluation for the chromatic characteristics of LED module under electrical and thermal coupling analysis, Microelectron. Reliab. 53 (12) (2013) 1916e1921. [15] K.C. Yung, H. Liem, H.S. Choy, W.K. Lun, Thermal performance of high brightness LED array package on PCB, Int. Commun. Heat Mass Transfer 37 (2010) 1266e1272. [16] C.-J. Weng, Advanced thermal enhancement and management of LED packages, Int. Commun. Heat Mass Transfer 36 (2009) 245e248. [17] D. Jang, S.-H. Yu, K.-S. Lee, Multidisciplinary optimization of a pin-fin radial heat sink for LED lighting applications, Int. J. Heat Mass Transfer 55 (2012) 515e521. [18] J.-C. Shyu, K.-W. Hsu, K.-S. Yang, C.-C. Wang, Thermal characterization of shrouded plate fin array on an LED backlight panel, Appl. Therm. Eng. 31 (2011) 2909e2915. [19] K.C. Yung, H. Liem, H.S. Choy, Z.X. Cai, Thermal investigation of a high brightness LED array package assembly for various placement algorithms, Appl. Therm. Eng. 63 (2014) 105e118. [20] B. Ramos-Alvarado, B. Feng, G.P. Peterson, Comparison and optimization of single-phase liquid cooling devices for the heat dissipation of high-power LED arrays, Appl. Therm. Eng. 59 (2013) 648e659. [21] J.-C. Wang, R.-T. Wang, T.-L. Chang, D.-S. Hwang, Development of 30 Watt high-power LEDs vapor chamber-based plate, Int. J. Heat Mass Transfer 53 (2010) 3990e4001. [22] J.-C. Wang, Thermal investigations on LED vapor chamber-based plates, Int. Commun. Heat Mass Transfer 38 (2011) 1206e1212. [23] Z. Lin, S. Wang, J. Huo, Y. Hu, J. Chen, W. Zhang, E. Lee, Heat transfer characteristics and LED heat sink application of aluminum plate oscillating heat pipes, Appl. Therm. Eng. 31 (2011) 2221e2229. [24] I.Y. Chen, M.-Z. Guo, K.-S. Yang, C.-C. Wang, Enhanced cooling for LED lighting using ionic wind, Int. J. Heat Mass Transfer 57 (2013) 285e291.

[25] C.-H. Chung, K.-S. Yang, K.-H. Chien, M.-S. Jeng, M.-T. Lee, Heat Transfer characteristics in high power LED packaging, Smart Sci. 2 (2014) 1e6. [26] D.P. Kennedy, Spreading resistance in cylindrical semiconductor devices, J. Appl. Phys. 31 (1960) 1490. [27] X. Luo, Z. Mao, S. Liu, Analytical thermal resistances model for eccentric heat source on rectangular plate with convective cooling at upper and lower surfaces, Int. J. Therm. Sci. 50 (2011) 2198e2204. [28] Y.-S. Chen, K.-H. Chien, Y.-S. Tseng, Y.-K. Chan, Determination of optimized rectangular spreader thickness for lower thermal spreading resistance, J. Electron. Packag. 131 (2009) 011004. [29] Y.-S. Chen, K.-H. Chien, C.-C. Wang, T.-C. Hung, Y.-M. Ferng, B.-S. Pei, Investigations of the thermal spreading effects of rectangular conduction plates and vapor chamber, J. Electron. Packag. 129 (2007) 348. [30] T.Q. Feng, J.L. Xu, An analytical solution of thermal resistance of cubic heat spreaders for electronic cooling, Appl. Therm. Eng. 24 (2004) 323e337. [31] C.T. Yang, W.C. Liu, C.Y. Liu, Measurement of thermal resistance of first-level Cu substrate used in high-power multi-chips LED package, Microelectron. Reliab. 52 (2012) 855e860. [32] K.-S. Yang, C.-H. Chung, M.-T. Lee, S.-B. Chiang, C.-C. Wong, C.-C. Wang, An experimental study on the heat dissipation of LED lighting module using metal/carbon foam, Int. Commun. Heat Mass Transfer 48 (2013) 73e79. [33] EIA/JEDEC Standard, EIA/JESD51-1 Integrated Circuits Thermal Measurement Method e Electrical Test Method (Single Semiconductor Device), 1995. [34] JEDEC Standard, JESD 51-14 Transient Dual Interface Test Method for the Measurement of the Thermal Resistance Junction to Case of Semiconductor Devices with Heat Flow Trough a Single Path, 2010. [35] J.P. Holman, Experimental Methods for Engineers, eigh ed., 2011. [36] V. Szekely, A new evaluation method of thermal transient measurement results, Microelectron. J. 28 (1997) 277e292. [37] V. Szekely, T.V. Bien, Fine structure of heat flow path in semiconductor devices: a measurement and identification method, Solid-State Electron. 31 (1988) 1363e1368. [38] L. Yang, J. Hu, S. Jang, M.W. Shin, Thermal design of ceramic packages for high power light-emitting diodes, Semicond. Sci. Technol. 22 (2007) 705e708. [39] H.-K. Fu, C.-P. Wang, H.-C. Chiang, T.-T. Chen, C.-L. Chen, P.-T. Chou, Evaluation of temperature distribution of LED module, Microelectron. Reliab. 53 (2013) 554e559. [40] R.-H. Horng, J.-S. Hong, Y.-L. Tsai, D.-S. Wuu, C.-M. Chen, C.-J. Chen, Optimized thermal management from a chip to a heat sink for high power GaN based LED, IEEE Trans. Electron Devices 57 (2010) 2203e2207. [41] F.P.D. Incropera, D.P. DeWitt, Fundamentals of Heat and Mass Transfer, fourth ed, John Wiley & Sons, 1996. [42] B. Vermeersch, G. De Mey, Dependency of thermal spreading resistance on convective heat transfer coefficient, Microelectron. Reliab. 48 (2008) 734e738. [43] H.H. Cheng, D.-S. Huang, M.-T. Lin, Heat dissipation design and analysis of high power LED array using the finite element method, Microelectron. Reliab. 52 (2012) 905e911.