On Fri, 2 Aug 1996, Eric Smith wrote:

> If X-ray erasure worked, wouldn't the semiconductor manufacturers either
> warn you not to expose parts to X-rays (in addition to the warnings to
> put stickers opaque to both UV and visible light over the windows), or brag
> about how you can reuse their OTP parts, or both?
>
> In summary, I don't know the answer but am doubtful that it will work.  But
> perhaps you have some actual experience?

I knew there was a reason I kept some of those old files.  :-)

From an Intel-authored article that was published in Electronics in the
August 14, 1980 issue... I'll summarize, mostly:

"The quantum yield and erasure rate per incident photon follows the
square-law dependence upon photon energy shown in figure 5."  The figure
actually looks more like two straight lines with a slight elbow, but
perhaps it was a lot harder to get accurate curves into timely
publication back then, long before DTP and laser printers and all that.
The curve only covers the visible to UV region anyway: the point here
being that normal room light or even sunlight will erase a MOS EPROM
orders of magnitude more slowly than the designed-for UV source.

Oh, BTW, I believe the reason this is alwasy spec'd for UV with a
wavelength of 2,537 angstoms is simply that that's the principal emission
line of mercury-vapor UV tubes.  The curve, however flawed, makes it
perfectly clear that this is NOT even remotely a tuned absorption phenomenon.

A bit later:  "An interesting exception exists for very energetic X rays
and gamma rasy, which can readily penetrate the select gate and then be
absorbed throughout the gate oxides, the floating gate, or the
substrate.  However, a broadband flux on the order of 10**13 photons per
cm [should be cm**2, I imagine] from an X-ray tube operated at 100 KV
with a tungsten target is necessary to fully erase an EPROM; hence X-rays
are not regarded as a special reliability concern for EPROMs."

I can't find any reference for the flux density for normal UV erasure,
but I suspect the situation is essentially that the number of X-ray
photons required to erase an EPROM (at least for the MOS technology Intel
was using around 1980) would be so great as to make the EPROM's erasure
of secondary concern.  Seems to me silicon semiconductors don't like
significant X-ray flux - betas degrade, gate thresholds shift, who knows
what happens to bulk resistance...  From what's in this article I can't
easily tell if this would be a serious concern in erasing EPROMs using
X-rays or not.