Thermal cycling testing is a very widely used especially in electronics. It exposes devices and components to fluctuating temperature, which causes fatigue in the components and especially in their interconnections. For example, cracking due to fatigue is one of the most common failure mechanisms in electronics, and therefore, it is a very important consideration in reliability analysis.
Fatigue failures are caused by different thermal expansion coefficients (CTE) of the materials used in electronics devices. For example, the CTE of silicon is very small (about 3ppm) while the CTE of polymer parts can be very high (more than 100ppm) especially, if the polymer materials need to be used unfilled. Even filled polymer materials tend to have rather high CTEs, as shown in the picture below, which can cause formation of very high stresses in the structures. Even though fluctuating stresses due CTE differences are the main reason for failures, in thermal cycling several other potential failure mechanisms are also present. For example, diffusion and relaxation of materials due to high temperature may affect the failure modes.
As mentioned above, thermal cycling testing is a very common test method. Due to this there are numerous test standards and recommendation for these tests and how they should be conducted. However, many standards give lots of options for test parameters or even mention that the parameters should be tailored according to the application. Although it is common to choose a test which has been widely used earlier, before testing it is useful to consider whether the parameters of the test really are suitable, or the most efficient ones for the studied component or device. There are several parameters to consider and picking the best combination is not straightforward.
The main parameters of thermal cycling testing include temperature limits, dwell time at both limits and change rate between the limits.
The main parameters of thermal cycling testing include temperature limits, dwell time at both limits and the change rate between the limits. In the picture below the main test parameters are shown. All of them are important and affect the stresses formed during testing. Of course, the number of test cycles is also a critical factor and the duration of testing should always be carefully considered.
The temperature limits are critical for the acceleration level of testing. The greater the difference between the limits is, the higher the stresses caused by them will be. However, if the limits are too extreme, there is a marked risk that overstress failures occur, leading to early failures which would never really occur in use conditions. A typical example of a critical limit is the glass transition temperature, Tg, of polymer materials. A temperature limit above Tg may lead to catastrophic failure which is easily seen but it may also just change the failure mechanisms to unrealistic ones. Then again, it is useful to use as high temperature limit as possible, since, if the difference between the limits is not great enough, the test has very small acceleration factor and the test time becomes very long.
In addition to the stresses caused by the differences between the temperature limits, the exposure to either high or low temperature may cause degradation leading to failures. For example, high temperature accelerates many harmful processes including for example diffusion, migration, and oxidation. An example of such processes is the growth of intermetallic layers in solder joints which typically reduces the mechanical robustness of the joints. High temperature also causes degradation and oxidation of polymers and permanently weakens their properties. To take these factors into account, it is important to consider how long exposure time is suitable at each temperature limit i.e. the dwell time at each limit.
If a long dwell time is used, the test duration increases unless the number of cycles is reduced. In the picture below the effect of cycle time to the test duration is shown. If the aim is to do 500 cycles with a 30 min cycle, we need approximately 250h or 1.5 weeks of testing. With 120 min cycle the test time increases to 1,000h or to 6 weeks. Then again, sometimes long dwell time may even accelerate testing, if it causes changes in the structure which increase the stresses during the temperature changes. For example, at high temperature polymer materials relax or creep – the polymer chains in the material move to reduce the stresses caused by the high temperature. When the temperature is lowered, these changes may significantly increase the stresses formed in the structures. However, long enough dwell time is required for these changes to occur. Often it is difficult to optimize the dwell time, but it is good to consider if critical changes may occur at high temperatures and would an extended dwell time be required.
Futhermore, the change rate of the temperature is critical. Very fast cycling testing (or shock testing) may cause thermal gradients to form in the tested structures. This means that different parts of a device or component heat up at different rates and warping of the system may occur. If such rapid changes of temperature may occur at use conditions, it is important to test their effects. However, typically such shocks are not present and in testing they cause incorrect failure mechanisms. Consequently, it is typically better to use slower change rates which allows different materials to warm up at similar rate. However, slower change rate naturally increases the testing time. Maximum change rate depends greatly on the structure tested. Small structures or components warm quickly and can be normally tested with very fast change rates but large devices typically require slow change rates and, also longer dwell times should be used.
When the test profile has been chosen, it is good to remember that the actual temperature within the test chamber or more importantly within the tested component may be something quite different than the programmed temperature. In the picture below the test temperature measured from a tested component and the programmed profile are shown. As can be seen, the actual change rate is clearly slower than the programmed one causing the dwell time to be shorter. Due to this effect with large components, there is a substantial risk that the components do not reach the temperature limits especially when a short cycle time with a fast change rate is used. Therefore, it is useful regularly to measure what the test sample is really exposed to in the test conditions, and to adjust the test parameters if needed.
Finally, the number of cycles is a critical parameter to determine. It depends on several other parameters, for example on the test parameters, the use conditions, and expected use life. For some structures, such solder joints, several formulas to calculate optimal test durations exist. However, there is no easy answer how to determine the duration of the test and it should always be considered on basis of the tested components and structures.