Article intended for FPGA-types has a nice summary (for dummies like
me) of the state of the wire-delay/interconnect problem:

http://www.us.design-reuse.com/articles/article5786.html

"James Meindl at the Georgia Institute of Technology, who has become
an expert in predicting the impending impact of physical parameters on
future IC generations, has taken on the interconnect problem. His
analysis predicts 80 levels of metal by 2014 if no architectural
changes are made in circuit design."

Won't happen of course.  Possible solutions: get away from Manhattan
routing (25% savings in wire delay--yawn), copper interconnect (been
done of course, discovered from the same article that copper atoms
like to diffuse into silicon, maybe it's just as well that chips
become obsolete every few years), repeaters every few atoms or so (why
do we keep trying to do it with electrons?  Because we're *electrical*
engineers, that's why), and then

"Another possibility in this direction is the introduction of photonic
waveguides for long interconnect lines. There is some hope here. With
recent work on building photonic bandgap structures into silicon
circuits, this might become a practical option for designers. Photonic
structures can now be defined on-chip with the same lithographic
processes used in CMOS manufacturing. Photonic interconnects do not
carry the RC delay penalty that creates so many problems for wire
inter connects."

RM

On Thu, 26 Jun 2003 13:04:09 +0200, Bernd Paysan <bernd.paysan@gmx.de>
wrote:

>
>Repeater/buffer insertion: nothing new, and trivial to do automatically.

At some scale, this strategy has to stop being sensible, because your
circuit will be nothing but repeaters and buffers.  They take up space
and consume power, if nothing else.

I posted the article in part because of discussions elsewhere to the
effect that rising mask costs (now in the multimillion dollar range)
will lead to the use of FPGA's where once ASICs might have been used.
That means that the interconnect issue (both on chip and getting from
the chip to the motherboard are problems, but I mean the interconnects
on the chip) become more and more important as feature sizes decrease.

The conclusion of the article sounds right to me: at some point, chip
designers will have to figure out how to do it with photons.

RM

In article <lntlfvschqhod2c45ikqqnmp5q43t511q8@4ax.com>,
 Robert Myers <rmyers@rustuck.com> writes:
snip
|> The conclusion of the article sounds right to me: at some point, chip
|> designers will have to figure out how to do it with photons.
|> 
|> RM
|> 
Yep, just like "they" said in articles that we would all be using x-ray
lithography to do .25 micron parts.  IBM even spend 1 super size (billion
dollars) to build a synchrotron in East Fishkill so it could do x-ray
lithography.  Anyone need a syncrotron?  I know where you can get time on one. :-)

-- 

Del Cecchi  
cecchi@us.ibm.com
Personal Opinions Only

On 26 Jun 2003 16:03:02 GMT, cecchi@signa.rchland.ibm.com (Del Cecchi)
wrote:

>Yep, just like "they" said in articles that we would all be using x-ray
>lithography to do .25 micron parts.  IBM even spend 1 super size (billion
>dollars) to build a synchrotron in East Fishkill so it could do x-ray
>lithography.  Anyone need a syncrotron?  I know where you can get time on one. :-)

Not my turf, as you know Del.  The limits to which optical lithography
have been pushed have taken most by surprise, even those who know a
good bit about some of the really strange things you can do with
optics.

All the same, my memberships in SPIE, ACM, and IEEE Computer Society
are all up to date.  No full membership in IEEE.  The future of
photons in computing is much more interesting than the future of
electrons, and people like you have the electrons business completely
covered, anyway.

RM

RM> At some scale, this strategy has to stop being sensible, because your
RM> circuit will be nothing but repeaters and buffers.  They take up space
RM> and consume power, if nothing else.

Yep.  I dug up something I wrote about this a couple of years ago, and
stuck it on my web site:

http://www.mcclatchie.com/articles/fo_vs_wires.html

Followup to:  <bdf5fm$uhm$1@news.rchland.ibm.com>
By author:    cecchi@us.ibm.com
In newsgroup: comp.arch
>
> In article <lntlfvschqhod2c45ikqqnmp5q43t511q8@4ax.com>,
>  Robert Myers <rmyers@rustuck.com> writes:
> snip
> |> The conclusion of the article sounds right to me: at some point, chip
> |> designers will have to figure out how to do it with photons.
> |> 
> |> RM
> |> 
> Yep, just like "they" said in articles that we would all be using x-ray
> lithography to do .25 micron parts.  IBM even spend 1 super size (billion
> dollars) to build a synchrotron in East Fishkill so it could do x-ray
> lithography.  Anyone need a syncrotron?  I know where you can get time on one. :-)
> 

It seems to me that electron beam lithography would be a lot easier to
deal with than X-ray lithography.  We're still a few generations away;
current fabs use mostly DUV (deep ultraviolet), and I think the 65 nm
generation will need EUV (extreme ultraviolet)... but at some point it
*will* run out of steam regardless of improvements.  However, it's not
clear that going to sticking with photons soft X-rays is the next
step... X-ray optics is messy, whereas charged particules can always
be dealt with electromagnetically.

	-hpa
-- 
<hpa@transmeta.com> at work, <hpa@zytor.com> in private!
"Unix gives you enough rope to shoot yourself in the foot."
Architectures needed: ia64 m68k mips64 ppc ppc64 s390 s390x sh v850 x86-64

On Thu, 26 Jun 2003, Robert Myers wrote:

> On Thu, 26 Jun 2003 13:04:09 +0200, Bernd Paysan <bernd.paysan@gmx.de>
> wrote:
>
> >
> >Repeater/buffer insertion: nothing new, and trivial to do automatically.
>
> At some scale, this strategy has to stop being sensible, because your
> circuit will be nothing but repeaters and buffers.  They take up space
> and consume power, if nothing else.
>
> I posted the article in part because of discussions elsewhere to the
> effect that rising mask costs (now in the multimillion dollar range)
> will lead to the use of FPGA's where once ASICs might have been used.
> That means that the interconnect issue (both on chip and getting from
> the chip to the motherboard are problems, but I mean the interconnects
> on the chip) become more and more important as feature sizes decrease.
>
> The conclusion of the article sounds right to me: at some point, chip
> designers will have to figure out how to do it with photons.
>

Biggest problem with optical is that high speed recievers are non-trivial,
and it would take a "long" signal path length to justify conversion from
electrical to optical and back. That is unless you want to go all-optical.

I'd be willing to bet that the first application for optical signaling
would be the clock trees.

Erik

Bernd Paysan <bernd.paysan@gmx.de> writes:

> Robert Myers wrote:
> > Possible solutions: get away from Manhattan
> > routing (25% savings in wire delay--yawn)
> 
> He even said this was "novel". Looking at the 25 years old die of the 8086, 
> 45 angle wiring wasn't novel back then - when layout was made by hand, and 
> digitized afterwards. What's novel is chip place&route tools that can 
> handle 45 angle wiring.
> 

One thing I've never understood is what makes IC P&R different to
PCBs?  PCB autorouters have been doing 45 degree routing for
years... Interested minds want to know.

Cheers!

Martin

-- 
martin.j.thompson@trw.com
TRW Conekt, Solihull, UK
http://www.trw.com/conekt

In article <klfjfvkej6gjid66vjumdopjoiau97kslc@4ax.com>,
Robert Myers  <rmyers@rustuck.com> wrote:
>Won't happen of course.  Possible solutions: get away from Manhattan
>routing (25% savings in wire delay--yawn), copper interconnect (been
>done of course, discovered from the same article that copper atoms
>like to diffuse into silicon, maybe it's just as well that chips
>become obsolete every few years) ...

It is, apparently, common knowledge that copper is doom, in terms
of silicon.  So I asked my chip-process-engineer brother how Intel
and IBM can be investigating the use of copper.  My digest of his
answer is that the "copper equals doom" theory appears to be bunk.
It certainly has some ways to go wrong, but from a chemistry and
solid-state physics standpoint, it ought to be able to work.

I may have misinterpreted what he said, but I throw this into the
mix for discussion purposes. :-)
-- 
In-Real-Life: Chris Torek, Wind River Systems (BSD engineering)
Salt Lake City, UT, USA (40�39.22'N, 111�50.29'W)  +1 801 277 2603
email: forget about it   http://67.40.109.61/torek/index.html (for the moment)
Reading email is like searching for food in the garbage, thanks to spammers.

Hi Chris,

Copper has been solved by the process guys.  The trick is that you have to
put a barrier metal (tantalum?) between the copper and the silicon
dioxide/FSG in order to prevent the copper from diffusing into the
dielectric.  This makes copper processes more complicated (dual damascene)
and hence a little more costly, have higher defect rates (tantalum bridging,
via underfill, etc.)  Apparently, things get trickier when you mix copper
with lower-k dielectrics.

There's a nice little paper here that describes things at a pretty high
level: http://www.icknowledge.com/threshold_simonton/techtrends01.pdf

It took a while for the fab guys to get everything worked out, but now adays
I think almost everyone's in copper at .13u and definitely below.  We've
been successfully manufacturing and yielding chips use copper interconnects
since the 0.15u node (in 2001), and all of our recent FPGAs use an
all-copper process (APEX 20KC, APEX II, Stratix, Stratix GX, Cyclone).

Regards,

Paul Leventis
Altera Corp.


"Chris Torek" <nospam@elf.eng.bsdi.com> wrote in message
news:bdh5md$pa1$1@elf.eng.bsdi.com...
> In article <klfjfvkej6gjid66vjumdopjoiau97kslc@4ax.com>,
> Robert Myers  <rmyers@rustuck.com> wrote:
> >Won't happen of course.  Possible solutions: get away from Manhattan
> >routing (25% savings in wire delay--yawn), copper interconnect (been
> >done of course, discovered from the same article that copper atoms
> >like to diffuse into silicon, maybe it's just as well that chips
> >become obsolete every few years) ...
>
> It is, apparently, common knowledge that copper is doom, in terms
> of silicon.  So I asked my chip-process-engineer brother how Intel
> and IBM can be investigating the use of copper.  My digest of his
> answer is that the "copper equals doom" theory appears to be bunk.
> It certainly has some ways to go wrong, but from a chemistry and
> solid-state physics standpoint, it ought to be able to work.
>
> I may have misinterpreted what he said, but I throw this into the
> mix for discussion purposes. :-)
> -- 
> In-Real-Life: Chris Torek, Wind River Systems (BSD engineering)
> Salt Lake City, UT, USA (40�39.22'N, 111�50.29'W)  +1 801 277 2603
> email: forget about it   http://67.40.109.61/torek/index.html (for the
moment)
> Reading email is like searching for food in the garbage, thanks to
spammers.

On Fri, 27 Jun 2003 04:55:00 GMT, Erik Magnuson <erik@cts.com> wrote:

>
>Biggest problem with optical is that high speed recievers are non-trivial,
>and it would take a "long" signal path length to justify conversion from
>electrical to optical and back.

High-speed optical transmitters aren't a walk in the park, either.

>That is unless you want to go all-optical.
>
*I* certainly don't want to do that, that's for sure.

>I'd be willing to bet that the first application for optical signaling
>would be the clock trees.
>
That sounds like a safe bet.  Finding even academic work on the
subject isn't easy.  The most useful document I turned up, with a nice
collection of references, was essentially a class paper:

www.ee.princeton.edu/~gcontrer/optical_clock.pdf 

RM

Paul Leventis wrote:

> Hi Chris,
> 
> Copper has been solved by the process guys.  The trick is that you have to
> put a barrier metal (tantalum?) between the copper and the silicon
> dioxide/FSG in order to prevent the copper from diffusing into the
> dielectric.

http://www.chipscalereview.com/issues/0301/techReport.html

gives an overview of barriers used. Tantalum Nitride is one of them, you 
also can use Titanium or Tungsten Nitride, and optionally add Silicon. 
Basically what you want is a robust crystal that doesn't allow the copper 
to migrate through. There are many available, pick those which can be 
processed easily and don't cost too much. Titanium Nitride is already used 
as barrier for the AlCu alloys, so the question is whether it's good enough 
(there are doubts). Tungsten has the disadvantage of crystalizing into 
nano-needles, which increases the surface resistance (not that Tungsten has 
a good inner resistance, either), but it works. Tungsten is also often used 
as local interconnect, and some processes use Tungsten barriers for AlCu 
(e.g. xfab 0.6u).

-- 
Bernd Paysan
"If you want it done right, you have to do it yourself"
http://www.jwdt.com/~paysan/

In article <bdg5jg$mcj$1@cesium.transmeta.com>,
 H. Peter Anvin <hpa@zytor.com> writes:
|> Followup to:  <bdf5fm$uhm$1@news.rchland.ibm.com>
|> By author:    cecchi@us.ibm.com
|> In newsgroup: comp.arch
|> >
|> > In article <lntlfvschqhod2c45ikqqnmp5q43t511q8@4ax.com>,
|> >  Robert Myers <rmyers@rustuck.com> writes:
|> > snip
|> > |> The conclusion of the article sounds right to me: at some point, chip
|> > |> designers will have to figure out how to do it with photons.
|> > |> 
|> > |> RM
|> > |> 
|> > Yep, just like "they" said in articles that we would all be using x-ray
|> > lithography to do .25 micron parts.  IBM even spend 1 super size (billion
|> > dollars) to build a synchrotron in East Fishkill so it could do x-ray
|> > lithography.  Anyone need a syncrotron?  I know where you can get time on one. :-)
|> > 
|> 
|> It seems to me that electron beam lithography would be a lot easier to
|> deal with than X-ray lithography.  We're still a few generations away;
|> current fabs use mostly DUV (deep ultraviolet), and I think the 65 nm
|> generation will need EUV (extreme ultraviolet)... but at some point it
|> *will* run out of steam regardless of improvements.  However, it's not
|> clear that going to sticking with photons soft X-rays is the next
|> step... X-ray optics is messy, whereas charged particules can always
|> be dealt with electromagnetically.
|> 
|> 	-hpa
|> -- 
If you are talking about direct write e-beam, that was used back a long time ago
to personalize bipolar gate arrays for mainframes on the remnants of the QTAT
line.  The three members of the line were Apsra, Poca, and Hontas.  The main
problem is one of throughput.  It was much lower than conventional masks.  

To use electrons in a high throughput situation one either needs a bunch of beams
or a projection type system, which has mask problems.  

X-ray masks are hard enough.  
|> <hpa@transmeta.com> at work, <hpa@zytor.com> in private!
|> "Unix gives you enough rope to shoot yourself in the foot."
|> Architectures needed: ia64 m68k mips64 ppc ppc64 s390 s390x sh v850 x86-64

-- 

Del Cecchi  
cecchi@us.ibm.com
Personal Opinions Only

In article <bdh5md$pa1$1@elf.eng.bsdi.com>,
 Chris Torek <nospam@elf.eng.bsdi.com> writes:
|> In article <klfjfvkej6gjid66vjumdopjoiau97kslc@4ax.com>,
|> Robert Myers  <rmyers@rustuck.com> wrote:
|> >Won't happen of course.  Possible solutions: get away from Manhattan
|> >routing (25% savings in wire delay--yawn), copper interconnect (been
|> >done of course, discovered from the same article that copper atoms
|> >like to diffuse into silicon, maybe it's just as well that chips
|> >become obsolete every few years) ...
|> 
|> It is, apparently, common knowledge that copper is doom, in terms
|> of silicon.  So I asked my chip-process-engineer brother how Intel
|> and IBM can be investigating the use of copper.  My digest of his
|> answer is that the "copper equals doom" theory appears to be bunk.
|> It certainly has some ways to go wrong, but from a chemistry and
|> solid-state physics standpoint, it ought to be able to work.
|> 
|> I may have misinterpreted what he said, but I throw this into the
|> mix for discussion purposes. :-)
|> -- 

Copper is doom.  IBM shipping large quantities of copper metallized chips.  Hmm,
secret double damascene process to keep copper out of si.  

Change "it ought to be able to work" to "It works"
-- 

Del Cecchi  
cecchi@us.ibm.com
Personal Opinions Only

In article <9vjedb.6m3.ln@miriam>, Bernd Paysan wrote:
> Robert Myers wrote:
>> Possible solutions: get away from Manhattan
>> routing (25% savings in wire delay--yawn)
> He even said this was "novel". Looking at the 25 years old die of the 8086, 
> 45 angle wiring wasn't novel back then - when layout was made by hand, and 
> digitized afterwards. What's novel is chip place&route tools that can 
> handle 45 angle wiring.

I've seen 45 degree routing in cache cells, but nothing else.  Even in the
smaller blocks drawn by hand (no place & route), we're still not allowed to
use 45s.  I was told that it makes the mask too difficult (too complex, too
expensive, too time-consuming to make..?  I don't know).  With all the OPC's
they're adding to the masks, I can't imagine how 45s would be so much harder.
Anyone know if that's true?

(this from the perspective of a CPU designer, not smaller chips)

  - Tim

-- 
Tim Schmidt
Not speaking for Intel.

Timothy Schmidt wrote:

> In article <9vjedb.6m3.ln@miriam>, Bernd Paysan wrote:
> 
>>Robert Myers wrote:
>>
>>>Possible solutions: get away from Manhattan
>>>routing (25% savings in wire delay--yawn)
>>
>>He even said this was "novel". Looking at the 25 years old die of the 8086, 
>>45 angle wiring wasn't novel back then - when layout was made by hand, and 
>>digitized afterwards. What's novel is chip place&route tools that can 
>>handle 45 angle wiring.
> 
> 
> I've seen 45 degree routing in cache cells, but nothing else.  Even in the
> smaller blocks drawn by hand (no place & route), we're still not allowed to
> use 45s.  I was told that it makes the mask too difficult (too complex, too
> expensive, too time-consuming to make..?  I don't know).  With all the OPC's
> they're adding to the masks, I can't imagine how 45s would be so much harder.
> Anyone know if that's true?

There was a time when the cost of rasterizing off-perpendicular 
geometries for mask-making was a substantial issue (and the process 
people freaked out at seeing all those stairsteps, even though it all 
anti-aliased out in the fab).

I wonder whether at these feature sizes there's enough anisotropy with 
respect to lattice direction in the projection, etching and deposition 
processes that characterizing it for arbitrary angles would just be too 
much work.

paul

--=.pi0LPknc6iZkNP
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit

On Thu, 26 Jun 2003 09:47:43 -0400
Robert Myers <rmyers@rustuck.com> wrote:

> I posted the article in part because of discussions elsewhere to the
> effect that rising mask costs (now in the multimillion dollar range)
> will lead to the use of FPGA's where once ASICs might have been used.

My advisor is part of a research group that is looking into an
interesting new way to design chips called "Via Patterned Gate Arrays"
(VPGAs).  The basic structure looks a whole lot like FPGAs: big arrays
of configurable logic blocks.  The big difference is that VPGAs are
programmed by optionally inserting little bits of metal (vias) during
manufacturing, as opposed to setting SRAM bits in the field.  The goal
is for VPGAs to be approximately as dense as standard-cell designed
ASICs, without the exponentially increasing pain of creating an entire
mask set.

Googling for "vpga via" will get you what little there is to read about
the technique on the internet.

Cheers,
Benjamin

--=.pi0LPknc6iZkNP
Content-Type: application/pgp-signature

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (GNU/Linux)

iD8DBQE+/JX7pFlxgi1HM9MRAqH+AJ4iK9DFM1VzLW/lv9MkkX2oCQETGQCfSIfL
J90ecE6lekSrAw05qb7j6YU=
=aTEd
-----END PGP SIGNATURE-----

--=.pi0LPknc6iZkNP--

On 27 Jun 2003 12:44:42 -0700, iain-3@truecircuits.com (Iain
McClatchie) wrote:

>RM> like keeping particles all of which have the same charge in a beam
>RM> isn't?
>
>If you want to pump a certain amount of *energy* into your photoresist,
>can't you improve the beam focus just by reducing the electron count
>and increasing the velocity, with no improvement in the guidance field?
>Forgive me for my naivitee, but don't higher energy electrons have
>shorter wavelengths (fewer issues with diffraction) and shorter-range
>absorption?
>
>Okay, now I'm going to ask a REALLY stupid question: at the limit,
>what stops me from exposing my photoresist by firing a single
>relativistically heavy electron past the mask?  Can a single electron
>couple energy into multiple photoresist molecules?
>

Higher energy electrons have shorter wavelengths, carry more energy
per electron, and reduce the charge density of a beam required to
deliver a certain power density, and are at least theoretically easier
to focus.  There are two problems:

1. The cost of producing very high energy electrons can be very high
(there's at least part of a big circular ditch in Texas as a testament
to that).

2. Getting high energy electrons to give up their energy isn't all
that easy.  Alpha particles from space make it all the way through 15
psi of atmospheric burden with enough energy left to invade RAM and
flip bits.

The DOE, which has even more flexible funding restrictions than either
IBM or Intel, has looked at multiple ways of delivering energy to a
small target for purposes of simulating fusion and always seems to
come back to photons.

See Del's parallel post for someone with experience with the actual
problem at hand.  Multiple beams seem like a much better way to get
high energy densities without high charge densities.

RM







>If the photoresist requires one activation electron per 100 square
>angstroms, then I'll need 100 million electrons to expose each square cm
>of wafer.  If I space my electrons roughly 1 cm apart (1 electron per
>cc), then my relativistic ebeam takes 1/3 of a second to make that
>exposure.  This does not seem like a very long time, and 1 electron per
>cc seems like it would produce a fairly small electrical field.
>
>A 300mm diameter relativistic electron beam with 1 electron per cc is
>3.3 uA, which doesn't seem like it would produce a big magnetic field
>either.
>
>It sure doesn't seem like this beam wants to disassemble over short
>distances.  What else is the problem?

www.xinitiative.org

They're plugging away at it.  One guy I know who runs a we-build-your-ASIC-
for-you shop doesn't think much is going to happen here for a while, just
because so many toolchain components have to be rewritten.  I see a lot of
pressure to pick up the benefit, which is comparable to the gains expected
from organic dielectrics.

I see a 10-20% reduction in overall wiring length.  That's a 20-40% drop in
capacitance and 10-20% drop in resistance.  Since the wires are the majority
of switched capacitance, and a good chunk of the resistance, that's a 15-35%
drop in switched capacitance, and a 20-50% speed improvement!

I think the industry has spent billions on the organic dielectrics.  It
seems to me that X routing could cost less than that to deliver.  I think
the problem is just a matter of time.  Several PhD students need to flow
through and write up brilliant new ways of handling non-manhattan geometry
efficiently.

And BTW, this is the electronics industry.  I don't think there have been
any 2-5x improvements in anything in a long time.  Maybe giant
magnetoresistance.  It's all about minimizing the risk of 2 - 30%
improvements here and there, in an environment where competing companies have
to synchronize their investments or have a very good chance of losing big
time.  Speaking of which, did everyone catch Intel abandoning the 157nm
lithography "node"?  Lots of pissed-off lithography folks, since Intel
effectively bulldozed them into investing in that option 3-4 years ago.

On 27 Jun 2003 17:32:31 -0700, iain-3@truecircuits.com (Iain
McClatchie) wrote:

<snip>
>
>And BTW, this is the electronics industry.  I don't think there have been
>any 2-5x improvements in anything in a long time.  Maybe giant
>magnetoresistance.  It's all about minimizing the risk of 2 - 30%
>improvements here and there, in an environment where competing companies have
>to synchronize their investments or have a very good chance of losing big
>time.  

Unless you and I have been living in parallel universes, losing big
time is as much a time-honored tradition in the electronics as in any
other.  The current climate makes people *very* cautious, which is not
necessarily a bad thing.

But if you're contemplating enterprises that don't seem plausible to
some people, and if the next 30% won't do it for you in the long run
(even if it's good enough to get started), you have a natural desire
to know if there is a brick wall at 31% improvement.

Okay, so I yawned at 25% improvement from x-routing.  That's a
one-time gain.  Who knows where the feature size barrier is, but as
the features get smaller and the masks get more expensive, the playing
field tilts toward FPGA's, but the on-chip connect problem gets bigger
and bigger.

Now, the answer could be, "Don't worry about it, champ.  We've fiddled
and diddled our way this far and who knows how far it will take us."
If that's the best answer there is, that's the best answer there is.

And, no, I wouldn't bother trying if I didn't think I could forsee the
kinds of gains you claim you never see. 

RM

"Robert Myers" <rmyers@rustuck.com> wrote in message
news:h38pfvcdg0ptd6bi2kdf4q4oh1khabbnl1@4ax.com...
snip
>
> 2. Getting high energy electrons to give up their energy isn't all
> that easy.  Alpha particles from space make it all the way through 15
> psi of atmospheric burden with enough energy left to invade RAM and
> flip bits.

Sorry.  The alpha particles in question come from the package and chip
materials.  Cosmic rays be a whole other thing.    Alpha particles are
stopped by a sheet of paper.
>
> The DOE, which has even more flexible funding restrictions than either
> IBM or Intel, has looked at multiple ways of delivering energy to a
> small target for purposes of simulating fusion and always seems to
> come back to photons.

Photons are easy to make in mass quantities.
>
> See Del's parallel post for someone with experience with the actual
> problem at hand.  Multiple beams seem like a much better way to get
> high energy densities without high charge densities.
>
> RM
>
>
>
>
>
>
>
> >If the photoresist requires one activation electron per 100 square
> >angstroms, then I'll need 100 million electrons to expose each square
cm
> >of wafer.  If I space my electrons roughly 1 cm apart (1 electron per
> >cc), then my relativistic ebeam takes 1/3 of a second to make that
> >exposure.  This does not seem like a very long time, and 1 electron
per
> >cc seems like it would produce a fairly small electrical field.
> >
> >A 300mm diameter relativistic electron beam with 1 electron per cc is
> >3.3 uA, which doesn't seem like it would produce a big magnetic field
> >either.
> >
> >It sure doesn't seem like this beam wants to disassemble over short
> >distances.  What else is the problem?
>

"del cecchi" <dcecchi@msn.com> wrote in message
news:Hr9La.929$IP6.41890@eagle.america.net...

....

> Photons are easy to make in mass quantities.

So to speak.

- bill

Bill Todd (billtodd@metrocast.net) wrote:
: 
: "del cecchi" <dcecchi@msn.com> wrote in message
: news:Hr9La.929$IP6.41890@eagle.america.net...
: 
: ...
: 
: > Photons are easy to make in mass quantities.
: 
: So to speak.
: 
: - bill

Oh, that was worth a good chuckle. Thanks. 
-Scot

On Sat, 28 Jun 2003 00:08:08 -0500, "del cecchi" <dcecchi@msn.com>
wrote:

>
>"Robert Myers" <rmyers@rustuck.com> wrote in message
>news:h38pfvcdg0ptd6bi2kdf4q4oh1khabbnl1@4ax.com...
>snip
>>
>> 2. Getting high energy electrons to give up their energy isn't all
>> that easy.  Alpha particles from space make it all the way through 15
>> psi of atmospheric burden with enough energy left to invade RAM and
>> flip bits.
>
>Sorry.  The alpha particles in question come from the package and chip
>materials.  Cosmic rays be a whole other thing.    Alpha particles are
>stopped by a sheet of paper.

I've read the "alpha particles are stopped by a sheet of paper" thing.
At some energy level, that is certainly true.  If you give a charged
particle enough energy, though, it will go through anything, although
with some finite probability of being scattered, depending on how much
mass it is going through.  When it does scatter, it isn't stopped, it
produces a shower of lower energy particles, which are in turn
scattered, and so on.  The fact that significant numbers of particles
reach the earth's surface is a testament to the fact that it takes a
lot of mass to get high energy charged particles to give up their
energy.

It is true that unscattered (primary) cosmic rays rarely reach the
earth's surface.  What we usually measure at ground level are the much
lower energy results of multiple collisions in the atmosphere.  Were
that process not going on, it is unlikely that life would be possible
anywhere on earth except in the deep ocean or underground.

http://www-istp.gsfc.nasa.gov/Education/wcosray.html

"The alpha particles in question come from the package and chip
materials," I am interpreting to mean that gamma rays, not some
charged particle, cause a local ejection of an alpha particle.  I
don't know about relative fluxes of charged particles and high-energy
photons that reach the earth's surface, but it is clear that
measurable quantities of charged particles reach the earth's surface
and that a significant part of the incident energy starts out as
charged particles at the top of the atmosphere.

It is also possible that you are quoting a result to the effect that
natural decay of packaging materials is more important than cosmic
radiation.  If that's true, it would be an interesting thing to know.

Another way of stating the penetrating power of very high energy
charged particles is that merely flying in an airplane frequently
raises your lifetime risk of cancer by an amount that shouldn't be
enough to stop you from flying, but isn't zero, either.  Living in
brick homes is dangerous, too. :-).

>>
>> The DOE, which has even more flexible funding restrictions than either
>> IBM or Intel, has looked at multiple ways of delivering energy to a
>> small target for purposes of simulating fusion and always seems to
>> come back to photons.
>
>Photons are easy to make in mass quantities.

It's a little more complicated than that.

RM

"Robert Myers" <rmyers@rustuck.com> wrote in message
news:vafqfv8a6fep9dgilv98ufli2bickv3b58@4ax.com...
> On Sat, 28 Jun 2003 00:08:08 -0500, "del cecchi" <dcecchi@msn.com>
> wrote:
>
> >
> >"Robert Myers" <rmyers@rustuck.com> wrote in message
> >news:h38pfvcdg0ptd6bi2kdf4q4oh1khabbnl1@4ax.com...
> >snip
> >>
> >> 2. Getting high energy electrons to give up their energy isn't all
> >> that easy.  Alpha particles from space make it all the way through
15
> >> psi of atmospheric burden with enough energy left to invade RAM and
> >> flip bits.
> >
> >Sorry.  The alpha particles in question come from the package and
chip
> >materials.  Cosmic rays be a whole other thing.    Alpha particles
are
> >stopped by a sheet of paper.
>
> I've read the "alpha particles are stopped by a sheet of paper" thing.
> At some energy level, that is certainly true.  If you give a charged
> particle enough energy, though, it will go through anything, although
> with some finite probability of being scattered, depending on how much
> mass it is going through.  When it does scatter, it isn't stopped, it
> produces a shower of lower energy particles, which are in turn
> scattered, and so on.  The fact that significant numbers of particles
> reach the earth's surface is a testament to the fact that it takes a
> lot of mass to get high energy charged particles to give up their
> energy.
>
> It is true that unscattered (primary) cosmic rays rarely reach the
> earth's surface.  What we usually measure at ground level are the much
> lower energy results of multiple collisions in the atmosphere.  Were
> that process not going on, it is unlikely that life would be possible
> anywhere on earth except in the deep ocean or underground.
>
> http://www-istp.gsfc.nasa.gov/Education/wcosray.html
>
> "The alpha particles in question come from the package and chip
> materials," I am interpreting to mean that gamma rays, not some
> charged particle, cause a local ejection of an alpha particle.  I
> don't know about relative fluxes of charged particles and high-energy
> photons that reach the earth's surface, but it is clear that
> measurable quantities of charged particles reach the earth's surface
> and that a significant part of the incident energy starts out as
> charged particles at the top of the atmosphere.

The alpha particles of classic soft error fame were emitted by package
materials such as lead and ceramic that contained trace amounts of
radioactive elements such as Uranium and Thorium which underwent alpha
decay, giving off particles with, if I recall, 4 to 5 Mev of energy.
People were actually going and searching for "old lead" that wasn't
contaminated.


>
> It is also possible that you are quoting a result to the effect that
> natural decay of packaging materials is more important than cosmic
> radiation.  If that's true, it would be an interesting thing to know.

It was.  Not such a big deal nowdays.  Once you figure out you need low
radioactivity packaging materials.
>
> Another way of stating the penetrating power of very high energy
> charged particles is that merely flying in an airplane frequently
> raises your lifetime risk of cancer by an amount that shouldn't be
> enough to stop you from flying, but isn't zero, either.  Living in
> brick homes is dangerous, too. :-).

And so a Alar on apples.  You are falling for a common piece of
statistical bullshit.  "If dose x causes 7 cases of cancer, then dose
x/10,000 causes y/10,000 cases of cancer", which may or may not be true,
depending on the substances and mechanisms involved.


>
> >>
> >> The DOE, which has even more flexible funding restrictions than
either
> >> IBM or Intel, has looked at multiple ways of delivering energy to a
> >> small target for purposes of simulating fusion and always seems to
> >> come back to photons.
> >
> >Photons are easy to make in mass quantities.
>
> It's a little more complicated than that.
>
> RM

Run current through a wire or a gas and out they come.  run current
through through a direct bandgap junction and out they come.  cake.  :-)
>
del cecchi

On Sat, 28 Jun 2003, del cecchi wrote:

>
> "Robert Myers" <rmyers@rustuck.com> wrote in message
> >
> > "The alpha particles in question come from the package and chip
> > materials," I am interpreting to mean that gamma rays, not some
> > charged particle, cause a local ejection of an alpha particle.  I
> > don't know about relative fluxes of charged particles and high-energy
> > photons that reach the earth's surface, but it is clear that
> > measurable quantities of charged particles reach the earth's surface
> > and that a significant part of the incident energy starts out as
> > charged particles at the top of the atmosphere.
>
> The alpha particles of classic soft error fame were emitted by package
> materials such as lead and ceramic that contained trace amounts of
> radioactive elements such as Uranium and Thorium which underwent alpha
> decay, giving off particles with, if I recall, 4 to 5 Mev of energy.
> People were actually going and searching for "old lead" that wasn't
> contaminated.
>

4 to 5 MeV is about right for classic alpha particles. I also recall
reading about old battleship armor used for shielding low background rooms
specifically because they were made before the start of atmospheric
testing.

Turns out that Bismuth is naturally radioactive with a half life on the
order of 10^18 years or so.

>
> >
> > It is also possible that you are quoting a result to the effect that
> > natural decay of packaging materials is more important than cosmic
> > radiation.  If that's true, it would be an interesting thing to know.
>
> It was.  Not such a big deal nowdays.  Once you figure out you need low
> radioactivity packaging materials.
> >
> > Another way of stating the penetrating power of very high energy
> > charged particles is that merely flying in an airplane frequently
> > raises your lifetime risk of cancer by an amount that shouldn't be
> > enough to stop you from flying, but isn't zero, either.  Living in
> > brick homes is dangerous, too. :-).
>
> And so a Alar on apples.  You are falling for a common piece of
> statistical bullshit.  "If dose x causes 7 cases of cancer, then dose
> x/10,000 causes y/10,000 cases of cancer", which may or may not be true,
> depending on the substances and mechanisms involved.
>

There's some indications that small doses of radiation are better than no
radiation at all - indications, not proof.  Needless to say, this stirs
up a lot of controversy.

Erik

In article <7SfLa.955$IP6.43152@eagle.america.net>, 
dcecchi@msn.com says...
> 
> "Robert Myers" <rmyers@rustuck.com> wrote in message
> news:vafqfv8a6fep9dgilv98ufli2bickv3b58@4ax.com...
> > On Sat, 28 Jun 2003 00:08:08 -0500, "del cecchi" <dcecchi@msn.com>
> > wrote:
> >
> > >
> > >"Robert Myers" <rmyers@rustuck.com> wrote in message
> > >news:h38pfvcdg0ptd6bi2kdf4q4oh1khabbnl1@4ax.com...
> > >snip
> > >>
> > >> 2. Getting high energy electrons to give up their energy isn't all
> > >> that easy.  Alpha particles from space make it all the way through
> 15
> > >> psi of atmospheric burden with enough energy left to invade RAM and
> > >> flip bits.
> > >
> > >Sorry.  The alpha particles in question come from the package and
> chip
> > >materials.  Cosmic rays be a whole other thing.    Alpha particles
> are
> > >stopped by a sheet of paper.
> >
> > I've read the "alpha particles are stopped by a sheet of paper" thing.
> > At some energy level, that is certainly true.  If you give a charged
> > particle enough energy, though, it will go through anything, although
> > with some finite probability of being scattered, depending on how much
> > mass it is going through.  When it does scatter, it isn't stopped, it
> > produces a shower of lower energy particles, which are in turn
> > scattered, and so on.  The fact that significant numbers of particles
> > reach the earth's surface is a testament to the fact that it takes a
> > lot of mass to get high energy charged particles to give up their
> > energy.
> >
> > It is true that unscattered (primary) cosmic rays rarely reach the
> > earth's surface.  What we usually measure at ground level are the much
> > lower energy results of multiple collisions in the atmosphere.  Were
> > that process not going on, it is unlikely that life would be possible
> > anywhere on earth except in the deep ocean or underground.
> >
> > http://www-istp.gsfc.nasa.gov/Education/wcosray.html
> >
> > "The alpha particles in question come from the package and chip
> > materials," I am interpreting to mean that gamma rays, not some
> > charged particle, cause a local ejection of an alpha particle.  I
> > don't know about relative fluxes of charged particles and high-energy
> > photons that reach the earth's surface, but it is clear that
> > measurable quantities of charged particles reach the earth's surface
> > and that a significant part of the incident energy starts out as
> > charged particles at the top of the atmosphere.
> 
> The alpha particles of classic soft error fame were emitted by package
> materials such as lead and ceramic that contained trace amounts of
> radioactive elements such as Uranium and Thorium which underwent alpha
> decay, giving off particles with, if I recall, 4 to 5 Mev of energy.
> People were actually going and searching for "old lead" that wasn't
> contaminated.

Don't forget the gasses (stupidly) used to test plastic package 
integrity.  I don't remember the details (Kr vs. Ar?), but one 
was a long-term alpha emitter, rather than a very short term beta 
emitter. IIRC, it was Intel that got royally burned by this.  Of 
course, another (ahem) got bit by polonium anti-static sources in 
the water supply.

-- 
  Keith

RM> 2. Getting high energy electrons to give up their energy isn't all
RM> that easy.  Alpha particles from space make it all the way through 15
RM> psi of atmospheric burden with enough energy left to invade RAM and
RM> flip bits.

Um, isn't an alpha particle actually a helium nucleus?  Betas are electrons,
right?  And betas don't get very far, I've heard, compared to alphas.  Betas
have hardly any mass.

And the stuff raining down from the heavens isn't generally the actual
charged particles that hit the upper atmosphere, right?  Any relativistic
particle entering the atmosphere generates an avalanche of particles, and
of those, it's mostly the photons that make it down to the surface.

On 28 Jun 2003, Iain McClatchie wrote:

> RM> 2. Getting high energy electrons to give up their energy isn't all
> RM> that easy.  Alpha particles from space make it all the way through 15
> RM> psi of atmospheric burden with enough energy left to invade RAM and
> RM> flip bits.
>
> Um, isn't an alpha particle actually a helium nucleus?  Betas are electrons,
> right?  And betas don't get very far, I've heard, compared to alphas.  Betas
> have hardly any mass.

Betas penetrate more than alpha's - alpha's do a better job of interacting
with the electron clouds surrounding atoms and have a much lower velocity
than a beta with the same kinetic energy.  A really high energy He nucleus
(say 10e9 eV or higher) will do a pretty good job of penetrating, but the
classic alpha particle will be less than 10e7 eV.

>
> And the stuff raining down from the heavens isn't generally the actual
> charged particles that hit the upper atmosphere, right?  Any relativistic
> particle entering the atmosphere generates an avalanche of particles, and
> of those, it's mostly the photons that make it down to the surface.
>

and muons (IIRC aka mu mesons)

On 28 Jun 2003 22:07:38 -0700, iain-3@truecircuits.com (Iain
McClatchie) wrote:

>RM> 2. Getting high energy electrons to give up their energy isn't all
>RM> that easy.  Alpha particles from space make it all the way through 15
>RM> psi of atmospheric burden with enough energy left to invade RAM and
>RM> flip bits.
>
>Um, isn't an alpha particle actually a helium nucleus?  Betas are electrons,
>right?  And betas don't get very far, I've heard, compared to alphas.  Betas
>have hardly any mass.
>
I assume that to get anything to happen, you have to couple energy
into the target.  Beta particles lose their energy mostly by ionizing
atoms.

From

http://physics.nist.gov/PhysRefData/Ionization/EII_table.html

I take an approximate ionization cross section for high energy
electrons to be about 10e-17 cm2.

Assuming a solid density of 10e23 atoms/cc, an ionization energy of
10eV, and the thickness of the photoresist to be 0.1 mm, I make the
absorbed energy for your relativistic beam with 1 electron per cc to
be 480 microwatts per cm2.

>And the stuff raining down from the heavens isn't generally the actual
>charged particles that hit the upper atmosphere, right?  Any relativistic
>particle entering the atmosphere generates an avalanche of particles, and
>of those, it's mostly the photons that make it down to the surface.

According to

http://www.research.ibm.com/journal/rd/401/curtis.pdf

a whole zoo of particles makes it to the suface, but it is neutrons
and pions that cause LSI failures.

RM

"Robert Myers" <rmyers@rustuck.com> wrote in message
news:2fhtfvgusc74ridtcoa0n8cuffvjr22g44@4ax.com...
> I assume that to get anything to happen, you have to couple energy
> into the target.  Beta particles lose their energy mostly by ionizing
> atoms.
>
> From
>
> http://physics.nist.gov/PhysRefData/Ionization/EII_table.html
>
> I take an approximate ionization cross section for high energy
> electrons to be about 10e-17 cm2.
>
> Assuming a solid density of 10e23 atoms/cc, an ionization energy of
> 10eV, and the thickness of the photoresist to be 0.1 mm, I make the
> absorbed energy for your relativistic beam with 1 electron per cc to
> be 480 microwatts per cm2.

In the case of circuits, energy or power are, in general, not too
helpful a unit.  It is minority carriers that get collected at junctions
and in potential wells over certain time intervals that count.  On the
other hand, one might, by considering size and critical charge or
current to do some rough calculations, with assumptions about how much
energy to create a hole-electron pair and how those pairs are
distributed longitudinally along the track of the particle.

(snipping above, beware)

del
>
> According to
>
> http://www.research.ibm.com/journal/rd/401/curtis.pdf
>
> a whole zoo of particles makes it to the suface, but it is neutrons
> and pions that cause LSI failures.
>
> RM

>

On Sun, 29 Jun 2003 11:50:58 -0500, "del cecchi" <dcecchi@msn.com>
wrote:

>
>In the case of circuits, energy or power are, in general, not too
>helpful a unit.  It is minority carriers that get collected at junctions
>and in potential wells over certain time intervals that count.  On the
>other hand, one might, by considering size and critical charge or
>current to do some rough calculations, with assumptions about how much
>energy to create a hole-electron pair and how those pairs are
>distributed longitudinally along the track of the particle.
>

I think there is a disconnect here.  Iain was (I thought) wondering
about relevance of the evidence I had presented to support the
assertion that it's hard to get high-energy electrons to give up their
energy.

Rather than muddy the waters further by trying to compare the physics
of the absorption of alpha rays and beta rays, I decided it might be
more helpful to make some numbers about Iain's original proposal for a
low-electron-density beam to show that such a beam would deliver a
very small amount of energy to a thin target.

I think you are saying that such a calculation has no relevance to the
likelihood of an intruding particle causing an unwanted current to
flow in an active semiconductor device, but that was not the intended
purpose of the calculation.  The IBM journal articles do such a
thorough job on that subject that I think it unlikely that I would
have anything to add.  I might add, editorially, that it speaks
volumes about the diligence of IBM that it published such thorough
work on an arcane but important subject in 1996.

RM

"Robert Myers" <rmyers@rustuck.com> wrote in message
news:7l9ufvs2eqmvokk36sdldeqgk749k92naa@4ax.com...
> On Sun, 29 Jun 2003 11:50:58 -0500, "del cecchi" <dcecchi@msn.com>
> wrote:
>
> >
> >In the case of circuits, energy or power are, in general, not too
> >helpful a unit.  It is minority carriers that get collected at
junctions
> >and in potential wells over certain time intervals that count.  On
the
> >other hand, one might, by considering size and critical charge or
> >current to do some rough calculations, with assumptions about how
much
> >energy to create a hole-electron pair and how those pairs are
> >distributed longitudinally along the track of the particle.
> >
>
> I think there is a disconnect here.  Iain was (I thought) wondering
> about relevance of the evidence I had presented to support the
> assertion that it's hard to get high-energy electrons to give up their
> energy.
>
> Rather than muddy the waters further by trying to compare the physics
> of the absorption of alpha rays and beta rays, I decided it might be
> more helpful to make some numbers about Iain's original proposal for a
> low-electron-density beam to show that such a beam would deliver a
> very small amount of energy to a thin target.
>
> I think you are saying that such a calculation has no relevance to the
> likelihood of an intruding particle causing an unwanted current to
> flow in an active semiconductor device, but that was not the intended
> purpose of the calculation.  The IBM journal articles do such a
> thorough job on that subject that I think it unlikely that I would
> have anything to add.  I might add, editorially, that it speaks
> volumes about the diligence of IBM that it published such thorough
> work on an arcane but important subject in 1996.
>
> RM

Published Internally some time before that.  :-)

I recall the studies which took memory boards to Colorado for soft error
testing to confirm the calculations.  Tough duty, but someone had to do
it.  Not me unfortunately.  :-(.

del cecchi

del cecchi wrote:
> "Robert Myers" <rmyers@rustuck.com> wrote in message
> news:vafqfv8a6fep9dgilv98ufli2bickv3b58@4ax.com...
>>> Del wrote :-
>>>Photons are easy to make in mass quantities.
>>
>>It's a little more complicated than that.
>>
>>RM
> 
> Run current through a wire or a gas and out they come.  run current
> through through a direct bandgap junction and out they come.  cake.  :-)

It depends on the energy of the photons, though; I thought it was you, Del,
who was telling us only last week how much IBM had spent on a synchrotron,
which is, after all, just a way of producing photons...

-- 
-- Jim

James Cownie	<jcownie@etnus.com>
Etnus, LLC.     +44 117 9071438
http://www.etnus.com

"James Cownie" <jcownie@etnus.com> wrote in message
news:DnSLa.11007$6F4.79856768@news-text.cableinet.net...
> del cecchi wrote:
> > "Robert Myers" <rmyers@rustuck.com> wrote in message
> > news:vafqfv8a6fep9dgilv98ufli2bickv3b58@4ax.com...
> >>> Del wrote :-
> >>>Photons are easy to make in mass quantities.
> >>
> >>It's a little more complicated than that.
> >>
> >>RM
> >
> > Run current through a wire or a gas and out they come.  run current
> > through through a direct bandgap junction and out they come.  cake.
:-)
>
> It depends on the energy of the photons, though; I thought it was you,
Del,
> who was telling us only last week how much IBM had spent on a
synchrotron,
> which is, after all, just a way of producing photons...
>
> --
> -- Jim
yep.  but a match and a puddle of gasoline works too. :-)  Or a tungsten
filament, now that someone figgered it out.

del cecchi

In article <bdhktb$ngi$1@news.rchland.ibm.com>, 
cecchi@signa.rchland.ibm.com says...
> In article <bdg5jg$mcj$1@cesium.transmeta.com>,
>  H. Peter Anvin <hpa@zytor.com> writes:
> |> Followup to:  <bdf5fm$uhm$1@news.rchland.ibm.com>
> |> By author:    cecchi@us.ibm.com
> |> In newsgroup: comp.arch
> |> >
> |> > In article <lntlfvschqhod2c45ikqqnmp5q43t511q8@4ax.com>,
> |> >  Robert Myers <rmyers@rustuck.com> writes:
> |> > snip
> |> > |> The conclusion of the article sounds right to me: at some point, chip
> |> > |> designers will have to figure out how to do it with photons.
> |> > |> 
> |> > |> RM
> |> > |> 
> |> > Yep, just like "they" said in articles that we would all be using x-ray
> |> > lithography to do .25 micron parts.  IBM even spend 1 super size (billion
> |> > dollars) to build a synchrotron in East Fishkill so it could do x-ray
> |> > lithography.  Anyone need a syncrotron?  I know where you can get time on one. :-)
> |> > 
> |> 
> |> It seems to me that electron beam lithography would be a lot easier to
> |> deal with than X-ray lithography.  We're still a few generations away;
> |> current fabs use mostly DUV (deep ultraviolet), and I think the 65 nm
> |> generation will need EUV (extreme ultraviolet)... but at some point it
> |> *will* run out of steam regardless of improvements.  However, it's not
> |> clear that going to sticking with photons soft X-rays is the next
> |> step... X-ray optics is messy, whereas charged particules can always
> |> be dealt with electromagnetically.
> |> 
> |> 	-hpa
> |> -- 
> If you are talking about direct write e-beam, that was used back a long time ago
> to personalize bipolar gate arrays for mainframes on the remnants of the QTAT
> line.  The three members of the line were Apsra, Poca, and Hontas.  The main
> problem is one of throughput.  It was much lower than conventional masks.  
> 
> To use electrons in a high throughput situation one either needs a bunch of beams
> or a projection type system, which has mask problems. 

I thought the purpose of the synchrotron was to produce enough 
electrons for direct write E-beam (on many tools).  I didn't 
think it was for X-ray lithography at all. I didn't deal with the 
E. Fishkill types at the time though. 

-- 
  Keith

> There's some indications that small doses of radiation are better than no
> radiation at all - indications, not proof.  Needless to say, this stirs
> up a lot of controversy.

However, the dose you accumulate when flying regularly - say, once a working
day on the Concorde 8-) - across the Atlantic is a substantial fraction of
the allowed dose for working in radioactive control zones. And the tendency
in the past years has been to lower that limit, if anything.

	Jan