بازسازی و تجزیه و تحلیل عملکرد حرارتی از حالات پرکردن باندینگ برای دستگاه های دیود با قدرت بالا

This paper proposed a half-experimental model to reconstruct the die-bonding thermal path of high-power light-emitting diodes (HP-LEDs). In this model, the partially insufficient filling of bonding materials and their directional/random distributions (“filling state” for short) have been taken into consideration. Both the silver-paste structure and the Au/Sn-eutectic structure were analyzed and compared. Finite element analysis (FEA) indicated that qualified die-bonding with uniform filled areas would lead to much better thermal performance. Hotspots have been observed above the insufficiently filled regions. The simulated thermal resistances of the defective bonding were 5.4 times and 2.1 times higher than those of the qualified samples under conditions of Au/Sn-eutectic and silver-paste, respectively. Transient thermal resistance measurements further demonstrated that the devices with different filling states would result in distinct thermal resistances. Interestingly, it was noted that although the qualified silver-paste bonding had a larger filled area, the measured thermal resistance remained higher than that of the defective Au/Sn-eutectic bonding because of the high contact thermal resistance caused by poor wetting properties. Furthermore, defectively bonded LED devices revealed a poor maintenance of luminous flux after 500 h of aging, which was consistent with the results of thermal performance analysis on the reconstructed die-bonding models.

With the development of high-power light-emitting diodes (HP-LEDs) [1] and [2], the move to replace general lighting fixtures with solid-state lights (SSLs) has become widespread. To reduce the cost of the rebuilding investment of lighting systems, the power density of LED chips is becoming increasingly higher, reaching a value of 125A/cm2 in the latest Cree® products [3]. This continuously increasing injection current has led to remarkable heat generation in the active layer of LEDs, and overheated p–n junctions have led to a decrease in luminosity [4] and [5], as well as premature failure after aging [6], [7] and [8]. Therefore, thermal management has become one of the most important issues with respect to energy efficiency and reliability. In practice, LEDs are manufactured into single components or compact chip-on-board (COB) packaged devices. Regarding the former, a surface mounting assembly of components on metal-core print circuit boards (MCPCBs) is necessary. This assembly introduces several sources of thermal impedance into lighting systems. COB packaged devices have been reported as an effective solution to this problem, exhibiting a significantly lower thermal resistance by reducing the distance of the thermal path [9] and [10]. In both cases, LED chips are inevitably die-bonded to a large leadframe substrate with high thermal conductivity by using a layer of bonding material. Moreover, from the optical point of view, a roughened leadframe substrate was preferred for improving light extraction [11] and [12]. Therefore, it is easily understood that it was difficult to form a perfect contact with the die-bonding material [13] and [14]. Although die-bonding materials such as silver-paste [14] and [15], carbon nanotube (CNT)-embedded silver-paste [16], SAC305 solder [15], [16] and [17], and Au/Sn-eutectic [14], [15] and [18] have been studied as potential solutions, the die-bonding layer is still one of the greatest conductive barriers in the thermal path of LED devices. The bonding layer was usually simplified as a regular contact structure in conventional design or analysis models. Thus, the thermal resistance directly calculated based on these models was much lower than the measured values [19] and [20]. Different mechanisms have been reported in the literature to explain such inconsistencies. From the thermal properties point of view, the die-bonding thermal resistance could be decomposed into one component representing the bulk thermal resistance and another component representing the contact thermal resistance [13] and [14]. Some researchers believed that the thermal conductivities of die-bonding materials were significantly lower than the theoretical values (effective thermal conductivity argument with respect to the bulk thermal resistance) [19], [20] and [21]. Others have indicated that the increased thermal impedances observed were mainly caused by poor contact conditions (contact thermal resistance argument) [14]. Generally, voids exist in the die-bonding layer in the forms of top/bottom, middle, and through modes [22]. Only in the bonding layers, wherein there is a large amount of middle voids, the thermal conductivities of bonding materials would notably decrease. However, in a real LED device, the die-bonding layer is extremely thin and not too large. In addition, die-bonding materials are prepared under sufficient deaeration conditions. It is reasonable to assume that the die-bonding materials are dense. Accordingly, it is believed that the contact thermal resistance argument is the most appropriate in these cases. T. Chung et al. calculated the thermal resistance based on an improved model with periodical stripe-like voids embedded in bonding structures, which was based on the real filled area ratio of bonding materials observed in ultrasound images [14]. Clearly, in his calculations, the material distribution was neglected, whereas in recent works, it was observed that the filling states with distinct distributions of die-bonding materials would not only affect the overall thermal resistance [22] but also create a significantly nonuniform junction temperature [23]. This paper extends the previous investigations and proposes a novel half-experimental model based on X-ray transmission images. The models are submitted to finite element analysis (FEA) as well as thermal resistance networks analysis and finally utilized to discuss the luminous properties of LED devices. This work presents a more precise method for predicting HP-LED die-bonding quality compared with other previously reported methods.

A considerably improved half-experimental model based on X-ray transmission images has been proposed to reconstruct the die-bonding thermal path of HP-LEDs. Both silver-paste and Au/Sn-eutectic structures were modeled and analyzed. Due to the effects of the bonding materials, interface structures, and bonding conditions, the die-bonding quality could not be fully guaranteed, which caused the insufficient filling of chip corners when silver-paste was used, as well as directional distributions of the bonding material in Au/Sn-eutectic structures. This information regarding the filling states was taken into consideration in the models developed in this study. FEA simulations indicated that the qualified Au/Sn-eutectic-bonded LED devices exhibited the best thermal performance with the lowest junction temperature of 74.5 °C. Moreover, the temperature uniformity of the epitaxial layer was highly similar to that of the die-bonding filled area, leading to hotspots above the insufficiently bonded regions. Transient thermal resistance measurements demonstrated that the devices with different bonding materials and qualities led to the generation of distinct thermal resistances. Interestingly, it was noted that although the qualified silver-paste bondings exhibited a larger filled area, the measured thermal resistance was still higher than that of the defective Au/Sn-eutectic bondings. By comparing the FEA results and the measured values, this contradiction could be ascribed to the contact thermal resistances of different die-bonding structures. This analysis indicated that the contact thermal resistances constituted more than 50% of the total thermal resistances in the die-bonding layers, especially in the silver-paste structures. Furthermore, LED devices with different bonding structures were submitted luminous measurements and long-term aging. The defectively silver-paste-bonded LED devices showed a poor maintenance of luminous flux, which was in agreement with the results obtained from the thermal performance analysis based on the reconstructed die-bonding thermal path. A useful finding is that the contact thermal resistance is the major heat conduction barrier in LED die-bonding structures. Improving the wetting conditions of die-bonding materials at the contact interface is one of the most effective approaches to resolve this issue. Die-bonding thermal models for HP-LEDs play important roles in components manufacture and failure analysis. To more closely simulate real physical structures, more actual filling information about the bonding layer must be taken into consideration, as demonstrated in this study. Yet, the findings presented in this report still have limitations. For example, the parasitic heat-conduction path was neglected under the conditions of the current measurement technology, which requires further study.