SPR 955 is an I-line series of photoresists developed by microchem, it's similar to SPR 220 but is formulated for thinner viscosities and higher resolution. We've been able to consistently get 500nm gratings in 0.97µm of SPR 955 in the stepper, on silicon. The 0.97µm thickness was chosen because it's one of the peaks listed in the datasheets swing curve. At this thickness substrates with moderate to high reflection of 365nm light will be more sensitive to the exposure time / actual thickness of the resist.
SPR 955 (0.97µm)
- Vapor prime wafer with HMDS
- Spin 0.97µm in ACS, CEE 200X #1 or CEE 100CB
- Bake at 100°C for 90 sec
- Verify thickness if possible, for thin resists small thickness variations drastically change the necessary exposure dose.
|Silicon||0.30 sec||5.5 sec|
|Low Reflection||0.34 sec - 0.50 sec|
|Borosilicate||0.30 sec - 0.34 sec||Severe sidewall degradation <1µm due to back reflection|
- Bake at 110°C for 90 sec
Notes: When photoresist is this thin the swing curve will have a larger effect; small thickness variations will have a larger change in the necessary exposure. Also at the end of the process you will have closer to .9-.94µm of resist. The softbake needs to be at 100C, not 115. Doing the softbake at 115 will burn out a lot of the PAC's, increasing dose to clear by ~40% (.24 sec vs .17 sec on the stepper).
These gratings are on silicon and exposed via the stepper:
Low Reflection Substrate
With gratings smaller then ~500nm it's very difficult to tell if they came out just via a microscope, the failed 350nm gratings below can look like they resolved just from an optical microscope.
The minimum size of an isolated line in a stepper is really limited by how well the substrate is in focus since it's really two large features (that can pass many orders through the lens) next to each other. Especially below 500nm you start to get pretty strong CD loss. And although 350nm & 400nm lines can typically be printed, small changes if how well the substrate is focused will have large changes in CD. Notice how some of these lines are around half the height they should be, this is probably just variation in how focused it was at that location due to wafer flatness, topology, etc.
Although you can get a nominal 350 / 400nm isolated trench to print, the physical size of it is going to end up closer to 500nm, as shown below.
This plot shows the ideal exposure time for several develops in the ACS. The 0.3 sec exposure with a AZ 300 25 sec develop in the ACS seems to be a good point.
The next plot shows the CD loss vs the amount of defocus in the stepper. ±900nm is about the limit before it's no longer going to work. Larger features will of course have a wider focus range.