Thermal Resistance and Reliability of High-Power LED ... - IEEE Xplore

89 downloads 0 Views 2MB Size Report
Dec 22, 2010 - Ming-Yi Tsai, Member, IEEE, Chun-Hung Chen, and Wan-Lin Tsai. Abstract—The light emitting diode (LED) packaging prob- lems associated ...
738

IEEE TRANSACTIONS ON COMPONENTS AND PACKAGING TECHNOLOGIES, VOL. 33, NO. 4, DECEMBER 2010

Thermal Resistance and Reliability of High-Power LED Packages Under WHTOL and Thermal Shock Tests Ming-Yi Tsai, Member, IEEE, Chun-Hung Chen, and Wan-Lin Tsai

Abstract—The light emitting diode (LED) packaging problems associated with high cost, high junction temperature, low luminous efficiency, and low reliability have to be resolved before the LED gaining more market acceptance. In this paper, chip-on-plate (CoP) LED packages with and without phosphors are evaluated in terms of thermal resistance and reliability under wet and high-temperature operation life (WHTOL) and thermal shock tests. The WHTOL test is with the condition of 85 °C/85%RH and 350 mA of forward current for 1008 h, while thermal shock test is with 200 cycles at temperature ranging from −40 °C to 125 °C. The thermal behavior of the CoP packages was analyzed by 1-D thermal resistance circuit (1-D TRC) with and without spreading angle, 3-D TRC method, and 2-D axisymmetric finite element method. The feasibility of these analyses was evaluated and discussed in detail by comparing those results with experimental measurements. The reliability results indicated that all CoP packages with phosphors in the silicone encapsulant failed after 309 h in the WHTOL test, but all those without phosphors still survived after 1008 h. The failure modes were found to be the debonding of the aluminum wire from the chip or copper pad of the substrate. However, after the aluminum wire was replaced by gold wire, all the packages with and without phosphors passed after 1008 h. For these survival packages in the WHTOL test, their thermal resistances of junction-to-air and junction-to-aluminum substrate increased by about 12 and 9 °C/W, respectively. Moreover, it was also found that there is a difference of 38 °C/W in the junction-to-air thermal resistances for the packages between under natural and forced convections in the chamber during the WHTOL test. This might yield the different reliability data, unless the flow conditions in the test chamber are specified in this standard test. Furthermore, all the packages with and without phosphors could pass 200 cycles in thermal shock test, with minor changes in the thermal resistances. However, the degradation of luminous flux in the packages with phosphors was found to be greater than those without phosphors by 14% vs. 9%. Index Terms—Humidity, light emitting diode (LED) package, reliability, thermal resistance, thermal shock.

Manuscript received June 16, 2009; revised November 18, 2009, January 28, 2010, and June 4, 2010; accepted July 22, 2010. Date of publication November 22, 2010; date of current version December 22, 2010. This work was supported by the Lei Yueh Company and the National Science Council, Taiwan, under the NSC96-2628-E-182-005-MY2 Project. Recommended for publication by Associate Editor M. Arik upon evaluation of reviewers’ comments. The authors are with the Department of Mechanical Engineering, Chang Gung University, Kwei-Shan, Tao-Yuan 333, Taiwan (e-mail: [email protected]; [email protected]; m9622021@stmail. cgu.edu.tw). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TCAPT.2010.2065805

I. Introduction N ORDER TO reduce the electricity consumption for alleviating the global warming problem, the high-power light emitting diode (LED), which features low power consumption, long lifetime, and short response time, has the potential to replace conventional lighting, such as incandescent and fluorescent lamps [1]. However, the LED issues associated with high cost, high junction temperature, low luminous efficiency, and low reliability have to be solved before gaining market acceptation. It is well known that high junction temperature in the LEDs would lead to the reliability problems, such as low quantum efficiency, wavelength shifts, short lifetime, and even catastrophic failure [2]–[4]. In the literature, the effect of die-attach properties on the junction temperature for high-power LEDs was studied by in-situ junction temperature measurement [5]. Moreover, the 1-D equivalent thermal circuit method associated with thermal transient tester was employed to determine the thermal resistance of a multichip LED package [6]. Recently, the reliability of GaNbased LED caused by degradation mechanisms in an active layer, ohmic contacts, and package/phosphor system was thoroughly reviewed and a set of experiments for evaluating those degradation mechanisms were further proposed [7]. Thermal and electrical aging conditions were found to cause similar mechanisms of LED degradations such as light output decay, spectral property change, and epoxy encapsulant browning. In addition, the degradation of optical output induced by thermal aging can be correlated by an exponential law [8]. The use of ceramic packages can enhance thermal performance, but should be given more attention during chip mounting process to avoid delamination failure [9]. Furthermore, a novel encapsulation process for an LED array associated with wafer level package was proposed for reducing the manufacturing cost and precisely controlling the geometry of encapsulations for optimizing optical performance [10], [11]. From the literature, there is a lack of data for the reliability of LED packages under high-humidity and high-temperature aging and thermal shock. In this paper, a novel package of high-power LED, so called chip-on-plate (CoP) package with the special features of low-junction-temperature and low-cost design, is proposed. Its thermal resistances and reliability under wet high temperature operation life (WHTOL) test and thermal shock test are evaluated. The configuration of the CoP package is shown in Fig. 1 with the details of geometries and materials. The thermal

I

c 2010 IEEE 1521-3331/$26.00 

TSAI et al.: THERMAL RESISTANCE AND RELIABILITY OF HIGH-POWER LED PACKAGES UNDER WHTOL AND THERMAL SHOCK TESTS

739

TABLE I Parameters for Convection Coefficient on Natural Convection

Horizontal plate facing up Horizontal plate facing down Vertical plate

Lch WL/2(W + L) WL/2(W + L) H

f 1.0 0.5 1.22

n 0.33 0.33 0.35

W: width of the plate (m), L: length (m), H: height (m), Lch : characteristic length (m).

Fig. 1.

(a) Configuration of CoP LED package. (b) Detail of geometry.

behavior of the CoP package is investigated by 1-D TRC (1-D thermal resistance circuit), 1-D TRC with spreading angle, 3D TRC method, and 2-D axisymmetric finite element method (FEM) as well as experimental measurements. The thermal resistances and optical characterizations of the CoP packages are also discussed before and after thermal shock test. II. Fundamental Heat Transfer The heat transfer from a high-temperature spot to lowtemperature one is mainly by three ways: conduction, convection, and radiation. The heat conduction occurs within or between solids, while the heat convection takes place within fluids or between solid and fluid [12]. The heat conduction equation based on Fourier’s law is described by (T1 − T2 ) dT q = −kA (1) = kA L dx (T2 − T1 ) q = −k (2) L where q is a heat flow from high temperature T1 to low temperature T2 , q heat flux, A an area of surface, k thermal conductivity, L a distance, and dT is the thermal gradient. The dx natural convection equation can be shown as q = hA (Ts − T∞ ) = hAT

(3)

q = h (Ts − T∞ ) = hT

(4)

where h is a convection coefficient, Ts the surface temperature of the solid, and T∞ is the ambient temperature. Based on Ellison [12] for a small-size geometry like the LED package, the convection coefficient in natural convection (hE ) can be expressed as hE = 0.83f (T/Lch )n

(5)

where T is the temperature difference between on package surface and in ambient, Lch is a characteristic length, and f and n are the constants, defined in Table I. III. Thermal Resistance Circuit Method For the electronic or optoelectronic packaging, the thermal resistance is one of the main parameters generally used for evaluating the performance of heat dissipation of packages

[13]. The lower the thermal resistance, the better the heat dissipation of the package. The thermal resistance is defined as   THigh − TLow T R= = (6) P P where R is a thermal resistance, P power dissipation, THigh high temperature, and TLow low temperature. Thermal resistances for thermal conduction (Rk ) and convection (Rh ) can be described, respectively, in terms of convection coefficient (h) and thermal conductivity (k) by L Rk = (7.1) kA 1 . (7.2) hA The heat flow q can be expressed in terms of Rk , Rh , and T as T1 − T2 T (8.1) q= = L/kA Rk Rh =

q=

Ts − T∞ T . = 1/ hA Rh

(8.2)

Equations (8.1) and (8.2) are analogous to Ohm’s law in an electric circuit by q and T corresponding to I and V . Three-dimensional thermal resistance circuit (3-D TRC) model, analogous to an electric circuit, is shown in Fig. 2 for thermal analysis of the CoP package. The heat flow starting from the chip splits into two flows through Paths A and BCD. One flow along the Path A directly passes through silicone (encapsulant) and dissipates into the air, and the other one through the Path BCD has a complicated path. The flow along Path B passes through sapphire (chip substrate) and Ag-epoxy (die attach) to aluminum (package substrate) and is divided into two paths (C and D paths). Furthermore, the flow along Path C goes from aluminum substrate to air in the four (front, back, left, and right) directions. In addition to the 3-D TRC, a 1-D TRC model with or without consideration of spreading angle in [13] is also implemented by ignoring Paths C and D from the 3-D TRC for detailed comparisons. The thermal conductivities for LED packaging materials in Table II and the convection coefficients in (5) are used in these analyses. During the analyses, the calculation of surface temperatures has to be iterated until temperature data converge, due to the convection coefficients as a function of the surface temperature shown in (5). Two input powers (0.5W and 1.22 W, corresponding to input currents 160 mA

740

IEEE TRANSACTIONS ON COMPONENTS AND PACKAGING TECHNOLOGIES, VOL. 33, NO. 4, DECEMBER 2010

Fig. 3. 2-D axisymmetric FEM model and detailed mesh for the CoP LED package. Fig. 2. 3-D thermal resistance circuit (TRC) model for thermal analysis of the CoP LED package (note: Cx4 means four C paths in parallel).

TABLE II Thermal Material Properties for LED Package Material Properties Silicone (encapsulant) LED chip Sapphire (chip substrate) Ag epoxy (die attach) Aluminum substrate (pkg. substrate)

Thermal Conductivity (W/m°C) 0.3 42 35 1.5 220

and 350 mA, respectively) are used in this paper. Note that 86% input power is considered to convert into heat, which is based on the experimental measurements.

IV. Finite Element Method ANSYS 9.0, a commercial finite element code, is used in this paper. The 2-D axisymmetric model is employed in the analysis with assumptions including the following. 1) Aluminum wires and the holes in aluminum substrate are neglected. 2) Perfect bonding exists between the material interfaces. 3) The heat transfer is considered only by thermal conduction and natural convection but not thermal radiation. 4) Eighty-six percept of input power converting into heat is uniformly distributed over the entire chip volume. Material thermal conductivities for the LED package in Table II are also used in this analysis. The 2-D axisymmetric model and detailed mesh are shown in Fig. 3. The convection coefficients for the package are from (5) associated with Table I. The calculation has to be iterated until temperature converges, due to the convection coefficients as a function of the surface temperature. The mesh-dependent solution and its coverage have been further confirmed by using a total of 700 to 15 000 elements in the model. The obtained temperature values only yield 0.4% variation. Two input powers of 0.5 W and 1.22 W are applied in this paper.

V. Temperature Measurements According to the electrical property of the diode, the relation between electrical current (I) and voltage (V ) at a given temperature [14] is shown as QV   T = (9) σ ln IIs where T is absolute temperature (K), Q electrical charge (Coulomb), σ Boltzmann constant (J/atom-K), and Is is the junction reverse saturation current. The above equation would give a linear relationship between T and V if I is relatively small. The junction temperature measurement [14], [15] used here is based on this principle by applying a small current (I = 0.1 mA) to the LED. This linear relation can be calibrated in an oven with constant temperatures and the calibration constant (ß) can be determined in Fig. 4(a) and given as    Thigh − Tlow   . β= (10)  V −V low

high

The pulse-current input is required for measuring the junction temperature. Before the junction temperature measurement, the driving current has to keep the LED in constant power about 20 min. so as to ensure that the LED reaches to a thermal steady state. The pulse current (IM = 0.1 mA) later applied for temperature measurement has only 1 ms duration for generating the LED voltage drop. During the measurement, the voltage difference (V , voltage drop) between the initialstate voltage (VMi ) and measurement-state voltage (VM ) is caused by junction temperature increasing. Then, the junction temperature (Tj ) can be described and determined by the following equations: Tj = Ta + Tj-a

(11)

Tj-a = (VMi − VM ) × β    Thigh − Tlow   Tj = Ta + (VMi − VM ) ×  Vlow − Vhigh 

(12) (13)

where β is a calibration factor, Ta an air temperature, Tj-a temperature difference between the junction and air temperatures. The junction temperature tester used in this paper shown

TSAI et al.: THERMAL RESISTANCE AND RELIABILITY OF HIGH-POWER LED PACKAGES UNDER WHTOL AND THERMAL SHOCK TESTS

Fig. 4.

741

(a) Relation of junction temperature (Tj ) and forward voltage (VF ) in LED. (b) LED junction temperature tester.

Fig. 5. Setup of LED junction and surface temperature measurement, based on EIA/JEDEC Standard 51-2 [16].

Fig. 6. Temperature field of CoP LED package, determined from 2-D axisymmetric FEM (Ansys) under the power of 1.22 W.

in Fig. 4(b) is designed based on the aforementioned principle. There are three specimens at each group under two input powers (0.5 W and 1.22 W) and their Tj are recorded individually. In addition, the thermocouples are employed to measure the air temperature and surface temperatures of aluminum substrate. The setup of the temperature measurement is shown in Fig. 5 as per JEDEC specification [16].

leading companies [17]. In the current thermal shock test, LED packages without any driven current are subjected to a temperature change from −40 °C to 125 °C within 20 s in 30 min a cycle. Meanwhile, the functionality of LED packages is examined after various cycles during the thermal shock test.

VII. Results and Discussion VI. Reliability Tests In order to evaluate the thermal performance of the LED package under the high-temperature and high-moisture condition, the WHTOL test based on JEDS22 Method A101-B is carried out in this paper. In this test, LEDs are driven with a forward current of 350 mA (power = 1.22 W) and are simultaneously subject to 85 °C/85%RH condition for 1008 h, based on the qualification test of one of the LED leading companies [17]. Moreover, the junction and surface temperatures of LED packages are measured at the room temperature after various periods of time during the WHTOL test. For determining the reliability of LED packages due to an abrupt temperature change, the thermal shock test based on MIL-STD-202G Method 107 G with 200 cycles at temperatures from −40 °C to 125 °C is performed in this paper, also referring to the qualification requirement of one of the LED

The 1-D TRC, 1-D TRC with spreading angle, 3-D TRC and 2-D axisymmetric FEM models were employed to analyze the thermal behavior of the CoP package. The temperature field of the CoP package, obtained from 2-D axisymmetric FEM under a power of 1.22 W, is shown in Fig. 6. It indicates that the temperatures at the chip junction and the aluminum substrate are uniform, but not for the temperatures at silicone encapsulant due to its low thermal conductive coefficient. Furthermore, the temperature data at each location from 1-D TRC, 1-D TRC with spreading angle, 3-D TRC method, 2-D axisymmetric FEM, and experimental measurements can be converted to the thermal resistances, shown in Fig. 7, using (6). The results show that the thermal resistances from junction to air and to aluminum substrate, obtained from 3-D TRC and 2-D axisymmetric FEM, are in good agreement with experimental data, but not for those from 1-D TRC and 1-D TRC with spreading angle. The reason is that some heat flow paths are ignored in both 1-D thermal models. Moreover,

742

IEEE TRANSACTIONS ON COMPONENTS AND PACKAGING TECHNOLOGIES, VOL. 33, NO. 4, DECEMBER 2010

Fig. 7. Comparison of thermal resistances for CoP LED packages, determined from 1-D TRC, 1-D TRC with spreading angle, 3-D TRC, 2-D axisymmetric FEM and experiments under the power of 1.22 W (note: experimental data include the average over 3 specimens and deviation).

Fig. 8. Effect of the normalized thermal conductive coefficient (k/kinitial ) of die attach on the thermal resistance for the CoP LED package from 2-D axisymmetric FEM, under the power of 1.22 W (note: kinitial = 1.5 W/m°C).

TABLE III Portion of Heat Flow in CoP LED Package, Determined from 1-D TRC, 1-D TRC with Spreading Angle, 3 D TRC, and 2-D Axisymmetric FEM Under the Power of 1.22 W

Fig. 9. Survival numbers of the blue-light and white-light (phosphorconverted) CoP packages during WHTOL test.

the heat flows through various paths of the CoP package, determined from these analyses, are listed in Table III for detailed comparisons. It can be seen that the heat flows at each path from 3-D TRC and FEM are quite consistent, while the heat flows at each path from 1-D TRC with and without spreading angle have similar results. Moreover, the heat flows through Path D (the bottom surface of aluminum substrate) from 3-D TRC and FEM analyses are only about one quarter of those from 1-D TRC with or without spreading angle. Based on that, the 1-D TRC model could produce a fairly large error in terms of the heat flow of the CoP package. As a result, the entire thermal performance of the CoP package would be underestimated using both 1-D TRC models. At the mean time, after the verification of the 2-D axisymmetric FEM model, the study of material parameter of die attach was carried out for understanding its effect on the thermal resistance of the CoP package. The effect of the normalized thermal conductive coefficient (k/kinitial ) of die attach on the thermal resistance of the CoP package is shown in Fig. 8 from 2-D axisymmetric FEM analysis. Note that k/kinitial = 1 representing the current case in which the k of the die attach is 1.5 W/m °C. The results indicate that the thermal resistance is quite dependent

Fig. 10. Failure modes of phosphor-converted white-light CoP packages during WHTOL test.

on the thermal conductivity of the die attach as k/kinitial 1. This also implies that even higher thermal conductivity of the die attach used would not provide an extra benefit to the entire thermal performance of this package. Two groups of CoP packages, with blue-light LEDs and phosphor-converted white-light LEDs, have been evaluated under the WHTOL test. The survival numbers of the specimens for both groups are shown in Fig. 9. The results indicate that all the blue-light CoP packages could pass the 1008 h duration during the test, but all the phosphor-converted whitelight CoP packages failed after 309 h. Subsequently, the failure modes of these white-light CoP packages were examined by microscopy and further confirmed by means of utilizing a

TSAI et al.: THERMAL RESISTANCE AND RELIABILITY OF HIGH-POWER LED PACKAGES UNDER WHTOL AND THERMAL SHOCK TESTS

Fig. 11. Maximum failure forces of aluminum wire bonding and gold wire bonding by a wire pull test (note: Sd stands for standard deviation).

Fig. 12. Survival number of the blue-light and white-light (phosphorconverted) CoP package with gold wire bonding during WHTOL test.

small tube to press the silicone of the powered CoP package to see if the failed specimens can be lighted on again. When these failed white-light CoP packages can still be lighted on by this method, it implied that the failure of the whitelight CoP packages was caused by aluminum wire debonding from the chip or copper pad of the substrate, shown in Fig. 10. It might be due to extra pulling stress or corrosion of aluminum wire bonding, resulting from moisture absorption for phosphor-containing silicone encapsulant. But, the exact failure mechanism has not been found yet. In general, the strength of the gold wire bonding was larger than that of the aluminum wire bonding. Two groups of CoP LED packages (without silicone), one with aluminum wire bonding and the other with gold wire bonding, were tested by a wire pull test. The obtained results, shown in Fig. 11, indicate that the average maximum failure force with gold wire bonding is larger than that with aluminum wire bonding by 3.9 g (about 53%) increasing. Thus, in order to further improve the reliability of the white-light CoP packages during WHTOL test, the aluminum wire bonding of the CoP packages was replaced with gold wire bonding. All blue and white-light CoP packages with the gold wire bonding eventually passed for the 1008 h duration as shown in Fig. 12.

743

In order to thoroughly understand the thermal performance of CoP packages during the WHTOL test, the junction temperatures (T j ) and thermal resistances (from junction to air, Rj-Air , and from junction to aluminum substrate, Rj-Al Sub ) for the blue-light CoP packages during the test were measured at the room temperature after each test step (in 0, 107, 195, 309, 447, 591, 735, and 1008 h) as shown in Fig. 13(a) and (b), respectively. It is shown that the average junction temperature increases with time duration at the WHTOL by up to 19 °C. Meanwhile, the thermal resistances for Rj-Air and Rj-AlSub increase by about 12 °C/W (16%) and 9 °C/W (47%), respectively, after 1008 h at the WHTOL. After this test, all the packages were baked out at 125 °C for 2 h to eliminate the diffused moisture, and then their thermal resistances were measured again. It was found that their thermal resistances for Rj-Air and Rj-AlSub return to the initial values, shown in Fig. 13(b). The similar phenomenon has been observed in the case of LED packages without silicone encapsulant (which is not shown here). That implies that the increase of the junction temperatures and thermal resistances with time duration during the WHTOL maybe is due to the decrease of the thermal conductivity of the die attach (as shown in Fig. 8) or possibly the increase of contact resistance, resulting from the moisture absorption. However, the effect of the moisture absorption on the thermal conductivity of the die attach material has been studied and found to be negligible. But the case about the increase of contact resistance at the interfaces between sapphire and die attach or between aluminum substrate and die attach during the WHTOL is still being working on and has not been concluded yet at this moment. For understanding the temperature variation in the CoP package under three different test conditions (25 °C/nature convection, 85 °C/85%RH/nature convection, and 85 °C/ 85%RH/forced convection) in a humidity chamber during the WHTOL test, the experimental and FEM analyses were carried out to evaluate the junction temperatures and thermal resistances for the CoP package at these three test conditions. The average junction temperatures and junction-to-air thermal resistances from experiments and FEM were found to be fairly consistent at each condition of 25 °C/natural convection and 85 °C/85%RH/natural convection, as shown in Fig. 14(a) and (b). By contrast, these values at the 85 °C/85%RH/forced convection condition (with 1 m/s flow) are lower than those at 85 °C/85%RH/natural convection condition by about 19 °C for the average junction temperature and about 38 °C/W for junction-to-air thermal resistance. However, the thermal resistances from junction to aluminum substrate at both conditions are in a good agreement. This significant decrease of the thermal resistances from junction to air is attributed to the forced convection in the chamber providing the higher thermal convection coefficient on the surface of CoP packages. As a result, during the WHTOL test the air flow condition in the humidity chamber would possibly affect the variation of reliability data. Therefore, flow conditions in the test chamber should be specified in the standard WHTOL test. For the CoP packages (with aluminum wire bonding) under the thermal shock test, two groups of the LED packages with 30 specimens in each group: one with blue-light LEDs and

744

IEEE TRANSACTIONS ON COMPONENTS AND PACKAGING TECHNOLOGIES, VOL. 33, NO. 4, DECEMBER 2010

Fig. 13. Variation of (a) junction temperature and (b) thermal resistance with time increasing for CoP LED package during WHTOL test.

the other with white-light (phosphor-converted) LEDs, have been tested in the thermal shock test at temperatures between −40 °C and 125 °C. Their functions for each group were examined after every 50 cycles (from 0 to 200 cycles). The survival rate of the specimens for both groups is shown in Fig. 15. It can be seen that all blue- and white-light CoP packages can pass 200 cycles during the test. That implies that the CoP packages with aluminum wire bonding could sustain the impact of the thermal shock better than the moisture absorption in the silicone encapsulant. The junction temperature and thermal resistances for the CoP packages after every 50 cycles of thermal shock were measured in the room temperature and shown in Fig. 16. The results show that, unlike the WHTOL test, the thermal shock gives only a negligible increase in both junction temperature and thermal resistance. In the meantime, the degradation of luminous flux for the LEDs was also evaluated after thermal shock test. The luminous flux for both groups of CoP packages with and without phosphor was measured by an integrating sphere system before and after the 200-cycle thermal shock test. Note that the integrating sphere is for distributing the light uniformly over the entire internal surface of the sphere with a LED light reflecting and scattering several times. This integrating sphere plus a photodetector (so called an integrating sphere system) can be used for measuring the luminous flux and light extraction efficiency of an LED package. The luminous flux degradation vs. survival rate is shown in Fig. 17. It can be found that the luminous flux degradation of LED packages with phosphors is more than those without phosphor by 14% vs. 9%. Moreover, if the

Fig. 14. Comparison of (a) junction temperatures, Tj , and (b) thermal resistances from junction to air, Rj-Air, and from junction to aluminum substrate, Rj-Al Sub, for CoP package from 2-D axisymmetric FEM and experiment under 25 °C/nature convection, 85 °C/85%RH/nature convection, and 85 °C/85%RH/forced convection (1 m/s) (note: experimental data include the average over three specimens and deviation).

Fig. 15. Survival rate of the blue-light and white-light (phosphor-converted) CoP package with aluminum wire bonding during thermal shock test (note: 30 specimens for each package).

luminous flux degradation of the 7% is defined as LED failure based on customer’s request, 90% of blue-light CoP packages pass, while only 60% of white-light CoP packages do after the 200-cycle thermal shock test.

TSAI et al.: THERMAL RESISTANCE AND RELIABILITY OF HIGH-POWER LED PACKAGES UNDER WHTOL AND THERMAL SHOCK TESTS

Fig. 16. Variation of the junction temperature and thermal resistances for the CoP packages with test cycles during thermal shock test.

745

natural and forced convections in the chamber in the WHTOL test exist the 38 °C/W difference in the junction-to-air thermal resistances, which might result in different reliability data. The results also showed that the junction-to-air thermal resistances were sensitive to the flow conditions of the chamber, but not for junction-to-aluminum substrate thermal resistances. Therefore, the standard test of the WHTOL should specify flow conditions in the test chamber. Furthermore, all CoP packages with and without phosphors were found to pass 200 cycles in thermal shock. The data of thermal resistances of all the CoP packages did not experience any apparent change. However, the luminous flux degradation of CoP packages with phosphors was found to be larger than that of those without phosphors by 14% vs. 9%.

References

Fig. 17. Luminous flux degradation of blue-light and white-light (phosphorconverted) CoP packages after 200-cycle thermal shock test (note: 30 specimens for each package).

VIII. Conclusion The thermal resistance and reliability of high-power CoP LED package were evaluated under WHTOL test and thermal shock test, specified in JESD22 Method A101-B and MILSTD-202G Method 107G, respectively. Before the WHTOL test, the thermal behavior of the CoP package was investigated by TRC methods (including 1-D, 1-D with spreading angle, and 3-D), 2-D axisymmetric FEM, and experimental measurements. The thermal resistances (from junction to air and from junction to aluminum substrate) of the CoP package from the 3-D TRC and 2-D axisymmetric FEM analyses were in good agreement with experimental observations, but not for those from 1-D TRC and 1-D TRC with spreading angle, due to the lack of some heat flow paths in 1-D models. The results showed that all CoP packages with phosphors in the silicone encapsulant failed after 309 h in the WHTOL test, but all those without phosphors passed for 1008 h. The failure was caused by the debonding of aluminum wires from the chip or copper pad of the substrate. However, as the aluminum wires were replaced with gold wires, all CoP packages with and without phosphors passed for 1008 h. For the survival packages, the thermal resistances from junction to air (Rj-Air ) and junction to aluminum substrate (Rj-Al Sub ) increase by about 12 °C/W (16%) and 9 °C/W (47%), respectively, after the WHTOL test. Moreover, the CoP packages under the

[1] F. M. Steranka, J. Bhat, D. Collins, L. Cook, M. G. Craford, R. Fletcher, N. Gardner, P. Grillot, W. Goetz, M. Keuper, R. Khare, A. Kim, M. Krames, G. Harbers, M. Ludowise, P. S. Martin, M. Misra, G. Mueller, R. Mueller-Mach, S. Rudaz, Y.-C. Shen, D. Steigerwald, S. Stockman, S. Subramanya, T. Trottier, and J. J. Wierer, “High power LEDs-technology status and market applications,” Phys. Stat. Sol. (a), vol. 94, no. 2, pp. 380–388, 2002. [2] N. Narendran and Y. Gu, “Life of LED-based white light sources,” IEEE/OSA J. Display Technol., vol. 1, no. 1, pp. 167–171, Sep. 2005. [3] N. Narendran, Y. Gu, J. P. Freyssinier, H. Yu, and L. Deng, “Solid-state lighting: Failure analysis of white LEDs,” J. Cryst. Growth, vol. 268, nos. 3–4, pp. 449–456, 2004. [4] M. Arik, C. Becker, S. Weaver, and J. Petroski, “Thermal management of LEDs: Package to system,” in Proc. SPIE, vol. 5187. 2004, pp. 64–75. [5] J. P. You, Y. He, and F. G. Shi, “Thermal management of high power LEDs: Impact of die attach materials,” in Proc. IMPACT, Oct. 2007, pp. 239–242. [6] L. Kim and M. W. Shin, “Thermal resistance measurement of LED package with multichips,” IEEE Trans. Compon. Packag. Technol., vol. 30, no. 4, pp. 632–636, Dec. 2007. [7] M. Meneghini, L. Trevisanello, G. Meneghesso, and E. Zanoni, “A review on the reliability of GaN-based LEDs,” IEEE Trans. Device Mater. Reliab., vol. 8, no. 2, pp. 323–331, Jun. 2008. [8] L. Trevisanello, M. Meneghini, G. Mura, M. Vanzi, M. Pavesi, G. Meneghesso, and E. Zanoni, “Accelerated life test of high brightness light emitting diode,” IEEE Trans. Device Mater. Reliab., vol. 8, no. 2, pp. 304–311, Jun. 2008. [9] J. Hu, L. Yang, and M. W. Shin, “Thermal and mechanical analysis of high-power LEDs with ceramic packages,” IEEE Trans. Device Mater. Reliab., vol. 8, no. 2, pp. 297–303, Jun. 2008. [10] R. Zhang and S. W. R. Lee, “Wafer level encapsulation process for LED array packaging,” in Proc. Int. Conf. Electron. Mater. Packag., 2007, pp. 1–8. [11] R. Zhang and S. W. R. Lee, “Wafer level LED packaging with integrated DRIE trenches for encapsulation,” in Proc. Int. Conf. Electron. Packag. Technol. High Density Packag., Jul. 2008, pp. 1–6. [12] G. N. Ellison, Thermal Computations for Electronic Equipment. Malabar, FL: Krieger, 1989. [13] J. Sergent and A. Krum, Thermal Management Handbook for Electronic Assemblies. New York: McGraw-Hill, 1998. [14] R. Frank, “Semiconductor junction thermometers,” in The Measurement Instrumentation and Sensor Handbook, J. G. Webster, Ed. Boca Raton, FL: CRC/IEEE Press, 1999. [15] B. Siegal, “Measurement of junction temperature confirms package thermal design,” Laser Focus World, Penn Well Corp., 2003. [16] Integrated Circuits Thermal Test Method Environmental Conditions Natural Convection (Still Air), JEDEC Standard EIA/JESD51-2, 1995. [17] Cree, Inc. Cree Xlamp XR family ad 4550 LED reliability. Technical Data Sheet of Reliability [Online]. Available: http://www.cree.com/ products/XLamp− Reliability.pdf

746

IEEE TRANSACTIONS ON COMPONENTS AND PACKAGING TECHNOLOGIES, VOL. 33, NO. 4, DECEMBER 2010

Ming-Yi Tsai (M’01) received the B.S. degree from National Cheng Kung University, Tainan City, Taiwan, in 1982, and the M.S. and Ph.D. degrees in engineering science and mechanics from Virginia Polytechnic Institute and State University (Virginia Tech), Blacksburg, in 1988 and 1990, respectively. Currently, he is a Professor with and Head of the Department of Mechanical Engineering, Chang Gung University, Kwei-Shan, Tao-Yuan, Taiwan. Prior to joining Chang Gung University in 1997, he was a Post-Doctoral Research Associate from 1990 to 1993 and a Research Scientist from 1993 to 1996 with the Department of Engineering Science and Mechanics, Virginia Tech, where he worked on and conducted a research project with adhesively bonded joints supported by the FAA. Later, he was with the Tjing Ling Research Institute, National Taiwan University, Taipei, Taiwan, for one year to work on electronic packaging research. His current research interests are electronic/optoelectronic packaging, photomechanics, adhesive bonding, and mechanics of composite materials. Dr. Tsai is a member of the International Microelectronics and Packaging Society.

Chun-Hung Chen received the M.S. degree in mechanical engineering from Chang Gung University, Kwei-Shan, Tao-Yuan, Taiwan, in 2004. Currently, he is pursuing the Ph.D. degree from the Department Mechanical Engineering, Chang Gung University. His current research interests include thermal management, thermo-mechanical analysis, optical simulation, and reliability evaluation for high-power light emitting diode packages and modules. Mr. Chen is a member of the International Microelectronics and Packaging Society.

Wan-Lin Tsai received the B.S. and M.S. degrees in mechanical engineering from Chang Gung University, Kwei-Shan, Tao-Yuan, Taiwan, in 2007 and 2008, respectively. He is currently with the Department Mechanical Engineering, Chang Gung University. His current research interests include thermal management and reliability of high-power light emitting diode packages and modules.