-1

Microchips can experience a pretty wide range of temperatures and they don't seem to crack or experience substantial fatigue from constant expansion and contraction. Why is this?

JRE
  • 67,678
  • 8
  • 104
  • 179
Aden
  • 11
  • 3
    A counter question, why do you assume they should fail from thermal cycles? – Justme Jun 18 '23 at 11:14
  • 6
    A counter question, why do you assume they don't fail from excessive thermal cycles? – asdfex Jun 18 '23 at 11:50
  • It's an unfair comparison since I'm looking at this from an automotive point of view where plastics and metal components tend to fail after many years of use and heat cycles in the car (and vibrations and other things of course), especially near the engine. Makes me wonder why my water cooled CPU and graphics card don't seem to have the same issues. Most CPU tend to experience wide temperatures ranging from ambient upwards of 90 C on a constant basis. – Aden Jun 18 '23 at 11:58

2 Answers2

3

Because materials like silicon (of which most chips are made) do not crack, or fracture just because the temperature changes. Chips (or more likely, interconnects like solder joints) fail when they are constrained from being able to expand and contract as the temperature changes.

This can occur when any two objects with differing coefficients of thermal expansion (CTEs) are rigidly attached to one another and they undergo temperature changes. The different CTEs cause the two materials to expand or contract at different rates, which stresses the place where they are rigidly joined.

So a failure can occur at the joint (think solder joint failure in a part attached to a PCB), or in one of the two materials, depending on where the forces are greatest and the weakest point in the assembly.

It's important to note as I pointed out in an earlier answer, How to design electronics to last 40 years or more? it may not be the amount of temperature change that is the problem, but the number of temperature cycles. Your CPU and graphics card are probably on continuously, whereas an automotive engine compartment component may experience temperature swings several times a day. The later may be the more stressing environment.

SteveSh
  • 9,672
  • 2
  • 14
  • 31
3

In fact, chips (as well as other electronic elements like transistors, resistors and so on) sometimes do fail because of repeated thermal cycles.

This failure mode is well-known and usually easy to engineer against by:

  • Using materials with high thermal conductivity (Silicon is impressive in this regard)
  • Using materials with low thermal expansion coefficient (Si is good at this, too)
  • Using a geometry that promotes effective heat dissipation (e.g. thin crystals)
  • Using soft packaging materials and/or leaving expansion gaps
  • Matching adjacent materials by their thermal expansion coefficient *
  • Engineering for less heat production in the first place
  • Placing limits on how quick the element must be heated or cooled down by external means (e.g. when soldered)
  • Distributing the heat produced between different parts of the chip (e.g. the operating system periodically switching the active core in a modern multi-core CPUs)

etc, etc... the list is by no means exhaustive.

* A good example of matching expansion coefficient that I remember is in fact related to incandescent bulbs and thermionic valves - a technology much older than chips. In order to prevent the glass from cracking around the electrical leads, one needs a metal that expands and shrinks exactly as the glass with temperature. Platinum happened to fit the job description so it was used for a while. Later, a much cheaper alternative ("platinite") was invented - an alloy that has the same important property while being orders of magnitude cheaper.

fraxinus
  • 8,726
  • 11
  • 34