A few common etch recipes using e-beam resist
PMMA can be used as an etch mask in fluorine plasmas, if the RF power is low enough. PMMA has a low glass-transition temperature (~110C) so you simply have to avoid melting it. CSAR and ZEP have better etch resistance mainly because they have a higher glass transition temperature. They are basically the same sort of organic polymers as PMMA, so the etch resistance is not drastically better.
HSQ has very good etch resistance to oxygen plasmas, but very poor resistance to fluorine plasmas. HSQ will etch considerably faster than CVD-deposited silicon nitride or silicon oxide. CSAR and even PMMA will work better when etching silicon or silicon compounds.
I-line photoresist has even better etch resistance than CSAR, and you can use photoresist in an e-beam process - but not as the resist. Spin a layer of photoresist, hard-bake it at ~ 180C, then spin on thin HSQ. Expose the HSQ with electrons, develop in TMAH, then transfer the pattern into the photoresist using an oxygen plasma. You must choose an appropriate thickness for the photoresist layer, which might require you to dilute a small amount. You will also have to work out the etch rate in oxygen. Next, switch the etcher to your favorite fluorine gas to etch the nitride, oxide, niobium, or whatever is under there.
The rates listed here are approximate. You should use them as a starting guess, and confirm the rates in your own etcher. Rates will vary with the diameter of the plates, the spacing of the plates, and the material (carbon or quartz) of the bottom plate.
Etch silicon oxide or silicon nitride using a mask of PMMA or CSAR
Here is an etch recipe for an ordinary parallel-plate etcher such as the Oxford-80, with a carbon bottom plate:
| CHF3 | 40 sccm |
| O2 | 2 sccm |
| pressure: | 30 mTorr |
| RF power: | 100 W |
This etches silicon nitride at about 10 nm/min, while etching the PMMA mask at about 5 to 10 nm/min.
Clean off resist with oxygen
After the fluorine etch, strip the remaining PMMA or CSAR with this:
| O2 | 30 sccm |
| pressure: | 30 mTorr |
| RF power: | 200 W |
Two minutes of this would be typical. This plasma develops a self-bias in a parallel-plate etcher (such as the Oxford-80). This bias aides etching by providing some physical sputtering. Barrel etchers do not strip resist as well, because they do not have any bias. Barrel etchers work very poorly for stripping resist from insulating substrates.
Deep holes in silicon for use as e-beam alignment marks
Here is an etch recipe for an ordinary parallel-plate etcher such as the Oxford-80, using a carbon bottom plate:
| CF4 | 20 sccm |
| SF6 | 20 sccm |
| pressure: | 30 mTorr |
| RF power: | 100 W |
Etch the silicon for about 6 minutes to get a 1.5 um deep hole. The resist layer can be 2 um thick PMMA. You could probably use CSAR, but we have not measured the etch rate.