Solid state lighting, especially high-power, high-brightness LEDs, are increasingly competing with traditional light sources. Their energy efficiency and ease of control are contributing to wide adoption in retail signs, automotive applications, streetlights, and indoor lighting.
Higher power and density increase operating temperatures that can modify and eventually destroy the electronic circuits in LED lighting systems, for example. Thus, thermal management is growing in importance. A key part of thermal design is to ensure that heat is released out of the system. Heat transfer to the outside world can be improved by better heat sinks, higher air velocities, and liquid cooling.
But first the heat must reach the surface of the package and transfer efficiently to the heat sink. Efficient heat transfer depends on the package's conductivity and the thermal resistance of the thermal interface material (TIM), plus the contact resistances. The thermal interface between outside cooling structures is far from perfect and can be improved by the type of TIM selected -- the lifetime of a device is directly related to the TIM's long-term reliability in the cooling assembly.
Accurate knowledge of the thermal conductivity and long-term behavior of TIMs is becoming more crucial to designing a reliable product. Resistances of the interface material are responsible for an increasing proportion of the temperature rise between the junction and the ambient. These thin layers are the Achilles heel when trying to design the device so that it will stay within its operating temperature range. The widely used composites have low thermal conductivity, and the minimum thickness is limited by the need to take up thermal mismatch between the materials on either side of the device, such as the silicon and metal die pad.
However, the thermal resistance posed by an interface layer is not simply the thickness and conductivity of the interface material. The contact resistance where the TIM layer wets each of the two surfaces significantly contributes to the overall interfacial resistance. The contact resistances between the TIM and the surfaces on either side undermine the performance of these materials in the operating environment.
The best approach to obtaining the most accurate data for use in a design's thermal simulation should come from measuring the total thermal resistance of the interface material used in a previous similar design. Then experimental characterization can be done to get the data needed to verify an optimal design. This thermal characterization increases confidence values that are used during simulation at all packaging levels, starting from within the package itself, using "structure functions" to identify discrepancies between simulation and reality.
In this example, I conducted repeated reliability tests for TIMs with a special emphasize on their thermal properties. We wanted to measure and monitor the changes of TIM performance with the goal of producing data that could be used to predict long-term performance of thermal-management solutions used for electronic systems and LED-based lighting applications. The special power cycling test we used resembled the real application and enabled tracking the TIM property changes quickly without changing the thermal system.
The tests were conducted on eight different materials (see table 1), and we observed three different behaviors. Some materials became significantly better during the power cycles, showing no sign of reliability problems. A small number of materials either kept their initial characteristics or became slightly worse. Most of the materials showed some clear change in their thermal properties. Thus we concluded that it would be a good idea to track these changes and include the information in datasheets. This information would help design engineers to predict the long-term behavior of their thermal systems.
We used a thermal transient methodology combined with the NID mathematical method1, which works well for this type of reliability characterization. The fine temperature resolution allowed us to generate tendencies, and the structure functions were used to verify whether the temperature changes originated from the TIM. When choosing the right temperature probe, the cycle time is very short -- we measured and recorded 2,500 cycles in less than 14 hours.