To add comments or start new threads please go to the full version of: 50nm pattern with EBL
PhysOrgForum Science, Physics and Technology Discussion Forums > Technology > Silicon technology

stefano.park
20th June 2006 - 07:57 AM
Hi all

I am trying to pattern a grating structure of 50nm linewidth. The distance between patterns is also 50nm. The linelength is 1um. There are 50 grating structures in 1um X 1um dimension.
I know each system has different characteristics but please give me some adivce.
My main problem is proximity effect.
Parameters.

Si wafer
PR: I tried both 2% PMMA (100nm) and 3% PMMA (150nm)
The acceleration voltage is 30keV
Working distance is 7mm
CCD and LS: 12
Beam current : I tried 5pA to 10pA
Doses: 200 to 300

How can I reduce the proximity effect to get 50 linewidth structure?

Another question

When I pattern 100nm or 200nm grating structures, there are roughness along the line edge. I need straight and clear edge. What parameter mainly cause this problem?

Thank you.

guiding_light
20th June 2006 - 09:09 AM
For your first question, you realize that the edges of your 1 um square grating receive less backscattering dose than the center of your grating (50% dense). The edges are ~25% dense (average of 50% and 0), and the corners are ~12.5% dense (25% 1:1 and 75% 0).

You will certainly need to compensate the dose, either reducing the grating dose or increasing the edge/corner dose. You will need to generate an artificial dummy pattern to differentiate the dose locations (like large blocks covering the edge). 30 keV proximity region may be several microns (?? you can check with a test pattern).

For your second question, at some point, it is inherent to the resist. Increasing dose will help reduce the shot noise contribution to LER. "Shot noise" means if you have too few particles hitting the resist, a random fluctuation makes a bigger contribution to the end result.
stefano.park
20th June 2006 - 09:28 AM
QUOTE (guiding_light+Jun 20 2006, 09:09 AM)
You will certainly need to compensate the dose, either reducing the grating dose or increasing the edge/corner dose. You will need to generate an artificial dummy pattern to differentiate the dose locations (like large blocks covering the edge). 30 keV proximity region may be several microns (?? you can check with a test pattern).


Thanks guiding_light!

I don't understand how to make an artificial dummy pattern to differentiate the doses location.



Thank you.
guiding_light
20th June 2006 - 10:06 AM
You have your original grating pattern, 1 um by 1 um. Maybe the proximity effect won't be so severe if the correlation length is several microns.

You can check by first writing a much larger grating (e.g., 100 um) and see how far from the edge the linewidth begins to change. Then you can decide how much additional exposure the edge needs, and how far from the edge you want to apply this extra dose.
Amit
13th July 2006 - 04:48 AM
Stefano were u able to generate 50 nm grating ?? If yes what are the parameters u are using?
Nikhil
1st August 2006 - 01:39 PM
For Gratings, I dont think proximity is a huge issue, essentially because the density of features is normally constant through out the surface. If you are experiencing problems exposing such patterns, my first concern would be absolutely perfect focus and stig correction, all over the field that you are exposing.

Other tricks that you can use are using a very small field size and stich the patterns if your ebeam can do this well, and using a very thin resist (1000A od PMMA or slightly less)

Keep in mind that the flatness of the substrate is of primal importance.

You can try general proximity correction for your structures using a program like caprox etc.

For LER in 200-300 nm grating, this can be reduced by using thinner layers of resist and optimizing your development/etch cycles.

fejones
1st August 2006 - 04:15 PM
You may try a procedure described by:
Hu, W.; Sarveswaran, K.; Lieberman, M.; and Bernstein, G., J. Vac. Sci. Technol. B [/B]22[B],1711 (2004)
Hu et al. use a higher dose (2-3X), then develop using a cold developer (4-8 deg C "IPA: Methyl isobutyl ketone MIBK 3:1 with 1.5 vol% methyl ethyl ketone (MEK)" (Hu et al., 2004).
The idea is to develop only the lowest MW of the fragmented PMMA. By going to lower temperature, the selectivity is increased and the development time is slowed down. This should decrease the problem with proximity dosage and allow you to get smaller lines with a smaller aspect ratio.
Guest
1st August 2006 - 05:43 PM
[QUOTE]Doses: 200 to 300
QUOTE]
Units?

[QUOTE]You will need to generate an artificial dummy pattern to differentiate the dose locations (like large blocks covering the edge). 30 keV proximity region may be several microns (?? you can check with a test pattern).[/QUOTE]

Or just try the original pattern and see how uniform across the grating the feature width is.

[QUOTE]Other tricks that you can use are using a very small field size and stich the patterns if your ebeam can do this well, and using a very thin resist (1000A od PMMA or slightly less) Keep in mind that the flatness of the substrate is of primal importance.[/QUOTE]

Stitching is probably bad for the 100 nm pitch grating he is making, since there are already so few periods in his 1 um grating.

[QUOTE]The idea is to develop only the lowest MW of the fragmented PMMA. By going to lower temperature, the selectivity is increased and the development time is slowed down. This should decrease the problem with proximity dosage and allow you to get smaller lines with a smaller aspect ratio. [/QUOTE]

Reduced development results in worse LER since a smaller fraction of exposed PMMA is removed. Sometimes it is recommended to descum the resist using an oxygen plasma, especially prior to liftoff.
Guest
1st August 2006 - 05:45 PM
QUOTE
Doses: 200 to 300

Units?

QUOTE (->
QUOTE
Doses: 200 to 300

Units?

You will need to generate an artificial dummy pattern to differentiate the dose locations (like large blocks covering the edge). 30 keV proximity region may be several microns (?? you can check with a test pattern).


Or just try the original pattern and see how uniform across the grating the feature width is.

QUOTE
Other tricks that you can use are using a very small field size and stich the patterns if your ebeam can do this well, and using a very thin resist (1000A od PMMA or slightly less) Keep in mind that the flatness of the substrate is of primal importance.


Stitching is probably bad for the 100 nm pitch grating he is making, since there are already so few periods in his 1 um grating.

QUOTE (->
QUOTE
Other tricks that you can use are using a very small field size and stich the patterns if your ebeam can do this well, and using a very thin resist (1000A od PMMA or slightly less) Keep in mind that the flatness of the substrate is of primal importance.


Stitching is probably bad for the 100 nm pitch grating he is making, since there are already so few periods in his 1 um grating.

The idea is to develop only the lowest MW of the fragmented PMMA. By going to lower temperature, the selectivity is increased and the development time is slowed down. This should decrease the problem with proximity dosage and allow you to get smaller lines with a smaller aspect ratio.


Reduced development results in worse LER since a smaller fraction of exposed PMMA is removed. Sometimes it is recommended to descum the resist using an oxygen plasma, especially prior to liftoff.
fejones
3rd August 2006 - 04:32 PM
QUOTE (Guest+Aug 1 2006, 05:45 PM)
Reduced development results in worse LER since a smaller fraction of exposed PMMA is removed. Sometimes it is recommended to descum the resist using an oxygen plasma, especially prior to liftoff.

I believe Hu et al. used the higher dosage to increase the amount of fragmented PMMA removed; and the higher dosage (based on an earlier post by guiding_light) should improve the LER.
Also, if you look at the paper, the profile for liftoff appeared to improve.
plasma_guy
19th August 2006 - 01:21 AM
One other thing to remember is that, especially with these automated proximity correction schemes like GHOST or dedicated boards (hardware Gaussian calculators), the correction is never 100% correct. As a result, you still have residual proximity effect, perhaps at new scales due to errors in the correction model. These errors not only result in linewidth error but also resist thickness differences. The latter is especially bad for liftoff.
guiding_light
19th August 2006 - 02:08 AM
QUOTE
One other thing to remember is that, especially with these automated proximity correction schemes like GHOST or dedicated boards (hardware Gaussian calculators), the correction is never 100% correct. As a result, you still have residual proximity effect, perhaps at new scales due to errors in the correction model.


All these vendor packages suck, period.

QUOTE (->
QUOTE
One other thing to remember is that, especially with these automated proximity correction schemes like GHOST or dedicated boards (hardware Gaussian calculators), the correction is never 100% correct. As a result, you still have residual proximity effect, perhaps at new scales due to errors in the correction model.


All these vendor packages suck, period.

These errors not only result in linewidth error but also resist thickness differences. The latter is especially bad for liftoff.


If you are really sensitive to resist thickness effects, the best solution is to use a thin (few nm at most) hardmask (native oxide is best) directly underneath the resist, with a short wet etch. Or the hardmask can be somewhat thicker than a few nm, and you use a short dry etch, assuming your resist thickness is at least a few tens of nm.
EBLer
20th November 2006 - 05:27 PM
QUOTE (guiding_light+Aug 19 2006, 02:08 AM)

All these vendor packages suck, period.



If you are really sensitive to resist thickness effects, the best solution is to use a thin (few nm at most) hardmask (native oxide is best) directly underneath the resist, with a short wet etch. Or the hardmask can be somewhat thicker than a few nm, and you use a short dry etch, assuming your resist thickness is at least a few tens of nm.

Agree! For those who don't have a good PEC software and other correction method, the best solution is use bi-layer (like HSQ resist on Photoresist) or trilevel (PMMA (or ZEP520, if you have)/hardmask layer/photo resist). It reduces proximity effect greatly. With a minor adjustment at the edge, you will be able to pattern 50nm half-pitch gratings. This method also helpful for you second question.
tryout
12th December 2006 - 08:54 AM
How about using contamination (like from vacuum pump oil) as the resist? It has actually been taken seriously at a number of places, though cant bring self to.
plasma_guy
5th June 2007 - 04:16 AM
QUOTE
One other thing to remember is that, especially with these automated proximity correction schemes like GHOST or dedicated boards (hardware Gaussian calculators), the correction is never 100% correct. As a result, you still have residual proximity effect, perhaps at new scales due to errors in the correction model. These errors not only result in linewidth error but also resist thickness differences. The latter is especially bad for liftoff.


Came across a review paper which may be worth reading:

H. Niedrig, "Electron backscattering from thin films", J. Appl. Phys. 53, R15 (1982).

The cited experimental data show significant variation of the backscattering coefficient, although as an 'average' description it still may be useful.
Guest
5th June 2007 - 07:16 PM

Same with secondary electron yield (as in SEMs), variable as hell. Electron scattering is very statistical in nature.
PhysOrg scientific forums are totally dedicated to science, physics, and technology. Besides topical forums such as nanotechnology, quantum physics, silicon and III-V technology, applied physics, materials, space and others, you can also join our news and publications discussions. We also provide an off-topic forum category. If you need specific help on a scientific problem or have a question related to physics or technology, visit the PhysOrg Forums. Here you’ll find experts from various fields online every day.
To quit out of "lo-fi" mode and return to the regular forums, please click here.