[Libre-silicon-devel] OK. That happened (Funding)
eegerferenc at gmail.com
Fri Dec 11 22:34:38 CET 2020
I digged a bit deeper into the topic. Unfortunately, it turns out that
coherence and monochromaticity indeed matter.
Light source in I-line and G-line steppers is usually mercury vapor
discharge tube, that is polychromatic and incoherent. However,
polychromaticity results in dispersion in lens material and chromatic
aberration, which reduces modulation index and usable resolution
significantly. Also, incoherent illumination (i.e. light source with
nonzero size) results in the existence of plane wave components in
pre-reticle illumination propagating with an offset angle to the optical
axis. After diffraction on the reticle, higher order diffraction modes of
these off-axis components would not be captured by the projection optics,
resulting in incomplete image reconstruction (reduced spatial bandwidth,
"blurring") at the wafer level, further reducing modulation and resolution.
To counter this, illumination systems in steppers contain a narrow-band
filter (Bragg-filter or similar) to filter out exactly one narrow spectral
line (hence I-line and G-line), and additional optics to eliminate off-axis
components and increase coherence (therefore, what reaches the reticle is
more or less "laser-like", highly coherent and monochromatic radiation with
very narrow FWHM). Also, the projection optics in the stepper is usually a
"narrow-band" system in terms of wavelength due to dispersion, and
fine-tuned to work only at the wavelength of the line it is specified for.
The micro-LEDs in the JBD display, as are all other LEDs, are polychromatic
(the FWHM of the one I considered in the calculation is 15nm), and it also
can be assumed that coherence is not too good, too (as high coherence would
mean near-zero viewing angle, which is a problem for a display). In
addition, the center wavelength of the display falls half-way between
I-line and G-line.
In my opinion, coherence may be improved by using micro-optics at each
micro-LED that produce highly collimated beams (JBD patents 10304375 and
9977152 describe microlenses and microreflectors, I don't know if they are
really capable of and committed to doing this or it is just
"shelling-patenting"). The wavelength of the emitters can be fine-tuned to
match the nominal wavelength of the projection optics (GaAs/InGaAs LEDs are
tunable to some degree by changing the Ga:In ratio). However, the effects
of residual incoherence and of the 15nm polychromaticity need to be
evaluated further, which is beyond my level of competence, and possibly
requires the detailed knowledge of the actual projection system and/or
fiddling around with one.
On Fri, Nov 20, 2020 at 12:28 PM Pavel Nikulin <pavel at noa-labs.com> wrote:
> On Fri, Nov 20, 2020 at 2:11 PM ludwig jaffe <ludwig.jaffe at gmail.com>
> > - Effect of incoherent illumination on projection optics
> I want to note that I-line mercury lamps were incoherent, and worked
> just fine before the industry switched to excimer lasers.
> I have doubts if effect on quality at such power levels, and
> resolutions is a thing.
> A 1µm process should be very viable for things like smart interposers,
> or other backend business. I'd say it should have an even higher
> bearing on things like WLCSP because you eliminate the mask shop from
> the loop every time you need to make a package modification, which can
> go up to 100+ for most popular chips.
> Libresilicon-developers mailing list
> Libresilicon-developers at list.libresilicon.com
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Libresilicon-developers