4 ms·
Is there any good evidence of vlsi ageing? What chemical or electro chemical reaction would be causative? Capacitor ageing is pcb dielectric leakage. Coils rus
by ggm 7y ago
Is there any good evidence of vlsi ageing? What chemical or electro chemical reaction would be causative?
Capacitor ageing is pcb dielectric leakage. Coils rust. Chips? Sealed, layered devices?
Thermal stress, I could believe
- anticodon 7y agoI've heard that diffusion can disrupt work of diodes and transistors by corrupting p-n junctions. Don't know what time is required for diffusion to destroy p-n junction and couldn't find an answer via Google.
- _ph_ 7y agoYes, that is a major factor in aging of chips. But giving a number for this is very difficult, as it varies on a lot of parameters. Quite a lot depends on the process the chip was manufactured with. Older processes with larger feature sizes usually are less affected than modern processes with smaller structures. But almost all details of processes are confidential and available under NDA only, so not openly published. Also, it depends on the concrete designs. That is, why the carrier migration is simulated as part of the modern design process. Of course, also the operation conditions play a role. The hotter the chip is and the higher the operation voltage, the quicker the diffusion happens.
- PhaseLockk 7y agoThere are many effects that contribute to semiconductor aging. Thermal stress and diffusion, as you and the child comment suggested, are definitely part of it. There is also electromigration, which can cause the interconnect to fail. In advanced nodes there are effects related to the magnitude of the electric field and the various material interfaceds, such as hot-carrier injection (HCI). Here's an article which gives explanation of a few aging effects: https://spectrum.ieee.org/semiconductors/processors/transistor-aging https://spectrum.ieee.org/semiconductors/processors/transist...
- Kirby64 7y agoAs others have noted, yes. It definitely impacts things. Semiconductor evaluation includes a significant time spent on long life expectancy (look up HTOL, or BHAST testing). HTOL and BHAST aren't just to evaluate for thermal stress (in fact, generally, there's ramp rates associated with the temperature changes to make sure you don't thermally shock something). They do it because diffusion, electromigration, etc, get much worse at higher bias voltages and high temperatures. This was significantly less of an issue on older stuff (or on stuff that is run on really large nodes, which tends to be older), since you have to diffuse a LOT more material to actually cause problems compared to the 7nm stuff today.