[Beowulf] OT: speculation regarding the 22nm fab process

[Alan Louis Scheinine]

If you are not interested in optics, press delete now.

Peter St. John wrote
  > Imagine building a process at say 44 nm, then measuring
  > it's output at 22nm precision. I'm considering the 22nm scale
  > measurement as a distortion. Then compute the inverse; apply
  > the inverse to your design; and feed the distored, or as it
  > were encoded, design to the input of the process; it's
  > (measured, not built) effect could be to produce a correct
  > feature at 22nm.

Well, manipulation of an image digitally is not quite the same
as making modifications on the light source, because the
wave nature of light and the wavelength used puts a limit
on what can be done.  While this is not related to Beowulf,
there are many physicists among the esteemed readership,
so I'll summarize what I found with Google.

Future Fab. has an article (subcribers only)
"We discuss how one can find an optimal photomask by rigorously
solving the inverse lithography problem."

An overview of articles seems to me to indicate that
X-Ray lithography has some problems and is behind
schedule so 22nm using UV is being stretched to its
limit.
Warren [of IBM] says the X-ray techniques - which are
sometimes called extreme UV lithography, or EUV - are not
working properly yet. At least not at prices that chip
makers or their customers can bear.

In general, the next generation lithography
(around 32 nm) will use

UV light

water immersion to reduce the wavelength

high numerical aperature, which
increases intensity and also reduces diffraction effects

With regard to 22 nm
Fabtech speculated:
significant increases in mask costs due in part to
the need for double patterning

In June of this year:
Last week Toppan announced what it believes is the industrys
first 32-nm photomask manufacturing process, set to begin volume
production this month.
[I cite this to indicate where we are in fabrication that anyone
can purchase (if you have several tens of millions) as a baseline
in comparison to 22-nm.]

In Feb. 2007, EE Times of India writes:
The current limit for 193nm immersion, with a refractive
index of 1.44, is around the production of devices at 40nm,
according to experts.

EDN published in June
The joint IBM-Toppan 22-nm photomask process will include
argon fluorine (ArF) immersion lithography, which is the
current mainstream technology for the manufacture of 32-nm photomask.

Canon write
[Immersion] When combined with an ArF excimer laser (193-nm wavelength),
which provides the shortest wavelength available for lens optics today.

Mark LaPedus of EE Times writes
    "A key to ''computation scaling'' is a partnership between
IBM and Mentor, which plans to devise a new resolution enhancement
technique (RET) to enable 22-nm designs and perhaps beyond.
This RET technology, know as source-mask optimisation, will
NOT [emphasis mine] eliminate dreaded and expensive 193-nm
immersion -- with double-patterning techniques.
Source-mask optimisation is said to optimise both mask layout and
illumination simultaneously to maximise image contrast in a scanner.
The technology is said to provide a means to minimise the use of
double-patterning by employing customised sources within the
scanner, along with optimised mask shapes."

The above repeats the idea that immersion and double-patterning
is used.  Other articles also mention that the computed mask
is associated with controlling the light source.

At Pallab's Place we read
Source Mask Optimization (SMO)
This SMO will reduce the high amount of complexity of the data on
the mask, and shift some of the complexity to the masking equipment
and the light source.  The impact will now be both mask and source
simulation data processing.

To summarize, with 193-nm light (in vacuum) and using water
with an index of refraction of 1.44, even 40 nm resolution
sound like a miracle.  I realize that near-field effects
might be relevant, but how flat could the wafer be to finely
control the distance to the mask over a 30 cm diameter?
A proposal in 2006 summarizes the state of the art, saying
"By taking advantage of the unique optical properties of water
at ultraviolet wavelengths, resolution nearly 1/20th of the
wavelength of visible light is possible."  What might those be?
Just below an absorption band, a tranparent material has a
large amount of scattering.  What else?
On page 18 of
http://www.almaden.ibm.com/u/mohan/R&D_in_IBM_India_2nd_Roundtable_on_Indo-US_Perspectives_in_Science_&_Technology_Mohan_Bangalore_2-16-2008_2-2008.ppt
there is a picture of a highly pixelated mask, very tiny pixels.
The idea of doing "tricks" with diffraction using pixels with
size on the order of a few wavelengths is normal, for example
a hologram.  Being able to do such tricks with pixels much smaller
than the wavelength is surprising.

Best regards,
Alan

-- 

  Alan Scheinine
  5010 Mancuso Lane, Apt. 621
  Baton Rouge, LA 70809

  Email: alscheinine at tuffmail.us
  Office phone: 225 578 0294
  Mobile phone USA:   225 288 4176  [+1 225 288 4176]