How stuff dies......
[ QUOTE ]
LEDlightman said:
Doug, when you say "far worse" than keeping the temperature constant, just how worse is it?
I guess what I am asking is are we all slowly destroying our high powered LED flashlights by simply using them?
[/ QUOTE ]
The sorts of failures brought by the sorts of stresses heat cycling causes tend to be fatal rather than degrading. Things flat out die. About as 'worse' as it can get.
For instance, one of the common ways real world materials fail is through the movement of what's called dislocations. Entire layers of atoms simply slip over each other within the crystal. When you bend a tea spoon or paper clip, this is what you're doing. Forcing layers of atoms of iron to break free from each other (and yet remain firmly in place in the crystal WRT their neighbors) and slip past each other one row of atoms at a time. When you release the pressure, the process stops. Once again, the material is stable and has nearly identical properties to what it had before. Do this too many times, however, and the dislocations increasingly don't 'knit up' when they stop and the material breaks. Heat treating the material (i.e.. 'tempering it') can modify the situation by allowing very small amounts of other material mixed in the iron (typically carbon in steels) to migrate through the material and collect in small groups called participates. It's called participate hardening. Fun process where the carbon moves through the hot (but not yet molten) iron by changing places with the iron atom next to it. A solid state solution change. The longer we hold this 'anneal' temperature before 'quenching' (rapid cooling which 'freezes' the atoms in place) the smaller number of larger participates we have and the 'softer' the steel is. Chemically, in fact, there is no difference between the tea spoon and the knife next to it in the drawer.
Dislocations and their actions are a common failure mode in semi-conductors. Heat changes cause them to move through solid matter in waves. It's very commonly observed in what's called 'in situ' (Latin for 'in place') heating experiments done is electron microscopes like I work with. You can watch dislocations move through a material, often controlled by what's called pinning, running into precipitates which act to stop them like nails driven through the pages of a phone book. Grain boundaries (the atomic level change from one grain to another) can perform the same function. Just like the over worked tea spoon, the semi conductor crystal (i.e. the LED) can and does fail.
There are lots of other failure modes in semi conductors of course. Also quite common is 'migration' of contact materials through the crystal (again driven by heat). Some failure modes are actually self healing, at the proper temperatures they mend as fast (or even faster) than they fail..
Of course we're "slowly destroying our high powered LED flashlights by simply using them?". If we never lit them up, they'd last for ever, right? Used reasonably (within maker's ratings) they should last a very long time. Thousands of cycles over thousands of hours. For instance, Lumiled Application Note AB 07, "Lumen Maintenance of White Luxeon Power Light Sources", shows little degradation for thousands of hours *at specifications*. It also states that the lens/liquid scheme is to prevent yellowing, not for thermal stress reductions in bond wires as suggested.
We can see from the spec sheets that we can store (not drive) or use (drive) the part at same maximum temperature limited by *the package*, not the die (which will run 15 or more degrees C hotter). No penalty for (within spec) heat. Cycles are another matter.
So, IMO, keep it between the ditches (run it at or under specs), and it should outlive you. Abuse the part, and you're on your own. And, yes, it's also my opinion that the far lower lifetime of the 5 Watt unit is heat related. Localized stresses not as extreme in the 1 Watt unit. Consider that the dies (and the process to make them) are essentially identical, it's the application that's different.
Cycles are a funny thing WRT reliability. You can fly a 737 nearly for ever, but as we've found pressurize and depressurize the cabin (like you do on each flight) something between 30 and 40,000 times and even a 'fairly new' plane will unexpectedly fall apart in flight. Without warning, just like the one in Hawaii did some years back when the top flew off in flight. "Maintain the even strain" as the saying goes.
Doug Owen
