Pattern Transfer

NEXT>

The pattern exposed in resist must be transferred somehow to the substrate. You could etch the substrate in a plasma of reactive gas (e.g. chlorine) or you could evaporate metal and then strip the resist in a solvent. Other pattern transfer techniques include implantation and polishing.

NEXT>

Pattern Transfer Page 2

NEXT>

Liftoff of metal wires requires soaking the wafer in solvents to “lift off” the unwanted metal. This is a common technique in research, because it is so simple, but liftoff is not usually used in production since there are more reliable and less messy ways of patterning wires. Acetone is a common solvent used for dissolving resist, and it becomes more effective if you heat it. But it also becomes explosive if you heat it! So if someone tells you to boil the acetone, tell them no- that’s foolish and unsafe. Instead, use warm NMP, which can be heated up to 150C without becoming dangerously explosive. (If you heat NMP above 200C it becomes acidic. Isn’t that strange?)

NEXT>

Pattern Transfer Page 3

NEXT>

In electron-beam lithography you have a limited selection of resists. The positive resists PMMA and ZEP are polymers that are broken up by electron irradiation. After exposure, the smaller bits dissolve faster in the “developer”. This is a very simple contrast mechanism; there isn’t any complex chemistry as you find in photoresist. (There are more complicated chemically amplified e-beam resists, but they are way too fussy and not worth the effort.)  

The negative resist HSQ contains monomers which are crosslinked by the electron beam. The unexposed regions wash away in developer.  UVN and ma-N are two other negative resists, but are far more difficult to use. For no apparent reason, ma-N negative resist has become more popular than UVN. Both have at best 50 nm resolution, and both use chemical amplification to induce crosslinking. HSQ has the best resolution of any e-beam resist, in the range below 8 nm.

Multilayer resist systems usually involve PMMA and the copolymer P(MMA-MAA). The copolymer is more sensitive to electron exposure, and so we typically put a layer of copolymer UNDER a layer of PMMA to form a large undercut resist profile. A large undercut is handy for doing metal liftoff.

YINQE provides PMMA, ZEP, HSQ and copolymer. We do not stock UVN and ma-N, and so if you want to use one of these then you’re on your own. ZEP and HSQ are quite expensive, and so we have to charge extra for those.

NEXT>

Pattern Transfer Page 4

NEXT>

Pick your developer. The best developer for PMMA is cold IPA+water (3:1), which we keep in a dedicated tank in the YINQE lab. If someone tells you to use MIBK-IPA mixtures instead, they are very much out of date. IPA-water is better in every way: higher resolution, less residue, less swelling, higher aspect ratio, lower toxicity, and lower cost.

ZEP can be developed in a wide assortment of chemicals, depending on your needs. There is a tradeoff between sensitivity (speed) and resolution. If you need to expose large features quickly, then you can use hexyl acetate as the developer. But most people use cold xylene to develop ZEP, since they need the best resolution.

When developing HSQ, there is a tradeoff between resolution and stress. That’s right: strong developers are used for the highest resolution, and these developers require very high electron doses. High doses shrink the resist more than low doses, leading to more stress in the resist film. Depending on the pattern, HSQ might peel off, crack, or “unzip” due to stress. That’s why people will use patterns of dots and short lines to show the highest resolution. For more practical applications we use a weaker developer. Around here we usually develop HSQ in MF312 (which is 0.54N TMAH) for 4 minutes.

NEXT>

Pattern Transfer Page 5

NEXT>

You might not want to think about it now, but at some point you’ll have to remove the resist from your wafer. The liftoff process dissolves resist in solvents, so that’s simple. If PMMA or ZEP is used as an etch mask, then you can remove the remaining resist simply by programming an oxygen etch step in the same plasma etcher. Just hit it with an oxygen plasma after (say) chlorine or fluorine. (Note that a low-bias barrel etcher will not do. You need a bit of bias to break up the polymer.)

Removing HSQ can be difficult, since it does not etch at all in oxygen. HSQ is a low-density silicon oxide, and so it will etch readily in fluorine plasmas or in hydrofluoric acid. If your device is incompatible with those etchants, then you could put a sacrificial polymer layer under the HSQ. A thin layer of PMMA or photoresist will allow the HSQ to be stripped in solvents.

NEXT>

Pattern Transfer Page 6

NEXT>

Our colleagues to the north seem to think that PMMA cannot be used as an etch mask. They are badly misinformed! Just keep the plasma power low (<= 100W) to prevent the resist from flowing. ZEP is a somewhat better etch barrier than PMMA, but only by 20% to 30%. That’s not very substantial, so you should be skeptical of exaggerated claims about ZEP. PMMA and ZEP are actually very similar polymers, and so you would not expect them to be terribly different as etch masks.

NEXT>