On Mon, Mar 29, 2010 at 10:04:49AM +0100, Steve wrote:
> Roger wrote:
> > Steve Thompson wrote:
> >>>> Some semiconductors I worked with years ago were rated for 25
> >>>> years continuous operation at maximum temperature. It was said
> >>>> that the tracks inside the part migrated to the next nearest
> >>>> track over time and stress.
> >>> I doubt the tracks themselves migrate.
> >> The metal does migrate, dependent on temperature and current
> >> density.
> > True. Two pieces of gold placed together will eventually bond. IIRC
> > The migration of gold through silicon is very slow except at very
> > high temperatures which would be high enough to create problems with
> > the doping migration and minority carriers on their own. Of course in
> > today's devices with very small trace spacing and extremely small
> > traces a number of problems crop up including conductor erosion from
> > electron flow, particularly when looking at devices with 40 nanometer
> > traces and equal spacing and millions or billions of transistors per
> > die.
> I think it's this last bit that David and I were referring to -
> electromigration. There's references to it in the old Motorola app notes
> AN1024, AN1025 and AN1027 on the Freescale web site (I think these are
> actually TRW originally and renumbered after the Motorola takeover). The
> effects are predicted by Black's equation.
>
> The ones I was thinking of were from Mike Flahie at TRW - App note #1
> 'Reliability and MTF - The Long and Short of it'. and #5 'A study of the
> advantages of gold metallization in the manufacture of microwave
> transistors'. All of this dates back to the late 60s and 70s.
IBM started finding out about electromigration the hard way when they
got involved in ... 100 nm traces, IIRC, and did a very intensive study
of the problem. Articles showed up in the IBM J. of R&D 80-ish.
--
Mike Andrews, W5EGO
mikea@mikea.ath.cx
Tired old sysadmin
_______________________________________________
Amps mailing list
Amps@contesting.com
http://lists.contesting.com/mailman/listinfo/amps
|