What exactly gets "worn out" and damaged by heat?

Question

It's quite common knowledge that heat is bad for electronics. That constantly high temperature decreases the expected lifetime of computer parts even if they are not overheating per se.

If, for example, there's dust insulating a component in a PC, "cutting it off" from the usual airflow. What is it that experiences higher "wear" on higher temperatures? I've seen liquid capacitors mentioned as parts failing faster the higher their operating temperture is, because of pressure building and resulting leaking. Is that correct? But surely, there are many things else? Could you name some?

Answer 1

There are really two different types of temperature stress, cyclying and sustained heat.

Just about any part is susceptible to failure from large number of temperature cycles. Each different type of material in a part expands and contracts at different rates. Of course packages are designed to accomodate this, and materials are chosen or specifically formulated for common thermal expansion responses, but stresses occur nonetheless. Eventually those stresses being applied back and forth enough times will break something.

Sustained heat is different. Silicon stops being a semiconductor, and silicon transistors therefore stop working, at around 150°C. Heating a IC to that temperature won't directly hurt it, other than it won't work as intended. However, that "not working as intended" could include excessive currents, which then cause more heat. Eventually something melts and the part is irreversibly damaged. Some chips, like modern processors, have such high density that failing to get rid of the heat for even a few seconds from the die can cause something to melt. Consider the size of a high end processor die compared to the end of a soldering iron, and then consider that there can be 10s of Watts dumped into the die, and that the soldering iron gets to solder-melting temperatures at that same power level. Getting rid of heat is a major issue with such chips. That is why they come with integrated heat sinks and fans nowadays. Take off the heat sink and fan, and your processor is toast in short order. Or, it shuts itself down to protect itself. Either way, your PC isn't going to run.

Electrolytic capacitors are different from most other electronic components in that they inherently go bad over time. Heat accelerates this. Running a electrolytic cap at 100°C, even without cycling, will degrade it much more rapidly than at 50°C.

Answer 2

No one has mentioned electromigration so let me add that. Failure of integrated circuit wiring due to electromigration is accelerated by temperature, and is independent of on/off cycles.

Answer 3

If a transistor operates at the same continual temperature, it actually will run reliably for many years. Continual heating and cooling of parts causes micro cracks due to uneven thermal expansion of different materials within the device. This is why tube televisions had evolved to have a constant grid heater at low wattage even when the TV is off. Hot to cold, cold to hot several times a day, 10,000 cycles in a few years....that's what caused TV's to fail.

This fact is not to diss the famous Arrhenius equation though (higher failure rate function of temperature). Most physical parts, like the capacitor you mentioned, obey the Arrhenius equation. It is necessary to point out that, for some devices, cycling is a cause of failure more than temperature.

My only concern, please someone tell this fact to the MTBF guys at Lockheed. Reliability equations there have no number-of-cycles factor so they just "wonder" why some satellites fail and some don't.

Answer 4

I can think of a few examples where heat plays a role in the degradation of parts:

1) Electrolytic capacitors, as you eluded to. The electrolyte slowly evaporates over time, and this evaporation is accelerated by the temperature of the part (both environmental and self-generated from ESR losses).

2) Optocouplers suffer from CTR (current transfer ratio) degradation as they age; this can be reasonably controlled by driving them as weakly as the design will allow and having overhead in the design for loss of CTR.

3) Class-II ceramic capacitors suffer dielectric aging, losing capacitance over time. This can be 'fixed' by heating the parts past their Curie point for a few hours, but this isn't something you can do when the part is in-circuit. (Johansen Dielectrics claims temperature plays a role in this aging, but doesn't provide any hard data)