Few points:

- Any reason you're using bilayer resist? It seems your metal film is pretty thin, and one layer ought to be enough to lift it off.

- Double check the temperature of the hotplate, sometimes overbaked resist can have trouble easily lifting off. I usually bake PMMA at a lower temp (~135-140 C), and it works just fine for me.

- What does your preprocess plasma ash step involve? Is it necessary? What is the substrate? In my experience, plasma chambers in academic clean rooms have a lot of cross-contamination and any "clean" step usually ends up with deposition of unwanted stuff like fluorocarbon films from F-plasmas. How this affects the adhesion properties of the metal film depends on what substrate and what kind of contamination it is. If you are only worried about organic contamination, you can substitute with standard cleaning steps (Piranha + HF or just Acetone + IPA).

- Aluminum can stick pretty well to standard substrates such as Si, although it shouldn't be that much more difficult to remove than Ti. Sometimes when you leave samples in stagnant solvents the resist film can dissolve underneath the metal, but if the metal film is stubborn it may not liftoff cleanly and actually redeposit on the substrate, which makes it more difficult to get rid of. You may want to sonicate the sample right after metal deposition, or squirt acetone on the sample from a squeezy bottle with some force to get rid of most of the film first. You can then clean it in solvent to remove other gunk and smaller flakes of metal.

- Hot acetone is always better at removing gunk and easier liftoff. Try 55C acetone overnight. Better yet, try 85C remover PG overnight. You'll get a much cleaner liftoff. Just make sure to cover the beaker with aluminum foil and move the hotplate to a fume hood! Also check with your safety coordinator about leaving samples unattended, some clean rooms have strict rules against that to prevent accidental fires and releases.

John,

Following up on Tanuj's answer: I agree that for the metal thickness you're interested in, using a bi-layer resist should not be necessary. I routinely lift off 100-nm-thick films with a single layer of PMMA. However, if you want to continue with a bi-layer process, I would suggest you check whether your lab stocks 950 PMMA in MIBK rather than in Anisole. (This is to avoid intermixing of the two layers.)

As for why lifting off Al is more finicky than Au, have you looked at your samples under SEM? The grain size of the evaporated metal can affect how easy lift-off is. See, for example, Figure 1 in APL 109, 201101 (2016) http://dx.doi.org/10.1063/1.4967791

Other possibilities: Are you using the same evaporator for the different metals? Does the geometry vary at all? i.e. are both metals evaporated from the same spot in the chamber or is one in the center and the other off to the side? You should always try to have the evaporation be normal to the sample surface. Do you find that you need significantly more current in order to evaporate Al than Au? (This is sometimes possible if your Al source is "dirty" because it's not used as often as Au.) This can cause the metal to evaporate "hot" and melt your resist, thereby making lift-off nigh impossible.

@Tanuj

I inherited this recipe, so no reason in particular. However, I have seen a few online sources that claim resist stack should be ~2-3x higher than evaporation thickness for reliable liftoff. Perhaps that's just false voodoo? We spin at 3500 RPMs, so a single layer of the resists I mentioned would not in fact be 3x taller than 40nm. It may not even be 2x (they both appear to spin on at ~70nm according to the MicroChem datasheet). I also was under the impression that a bilayer gives more reliable liftoff, since the bottom layer of resist is more sensistive to e-beam, thus is blows out, while the top layer does not blow out as much. This gives a nice undercut profile, which is good for liftoff obviously. 

I'm using the temperature/time recommended on the MicroChem datasheet. I have checked hotplate temperature, and it is very much 180 C. I think this is less of a resist problem, and more of the metal I'm using (or how this metal is being evaporated). 

I am currently questioning how useful the pre-evap plasma ash step is. We do it for 6 seconds on fairly low power. I know this will lightly etch PMMA, which is why I'm wary of it. We do have our own plasma asher so it is not shared, but it is old and probably not cleaned often enough. It's just an oxygen plasma. In the future maybe I'll forgo this step, as I really don't know how useful a 5 second plasma ash could be. 

I like the ideas for trying to liftoff the majority of the metal after a short time, so I'll give that a shot one of these days. I also have seen many people recommend hot acetone liftoffs, the problem is our lab does not have a fire quenching system in place for our fume hood, so I'd have to do this in our clean room. No biggie, and probably worth a try. 

Thanks!

@Andreas

We unfortunately do not stock the PMMA in MIBK rather than in Anisole. I'm confused how you could be doing liftoff of ~100nm with a single layer of PMMA, unless it is a very different polymer than the ones I listed. At 3500 RPMs, these resists spin on at 70nm, so evaporating 100nm would definitely give me trouble. 

The reason I use the bilayer is because one resist is more sensitive than the other (bottom resist), so it blows out more than the top layer. This is so we get a nice undercut profile. I would take this type of resist stack any day over a sloped sidewall resist for liftoff. However, I don't know much about the intermixing of these two different layers, maybe that could play a role. 

I think it has more to do with aluminum than the resist profile itself. I have looked at plenty of samples under the SEM, and indeed the grain size for aluminum is huge compared to Ti/Au (can sometimes be ~50nm or so). I see similar effects as that figure shows, where the boundaries of aluminum are grainy/non-ideal. We do plan to get a nitrogen-cooled evaporator stage in the near future to reduce aluminum grain size, so hopefully that helps. 

It is the same evaporator, and since the crucibles are on a rotating chuck, they are evaporated from the same spot. It is a relatively long distance to get to the sample, and there is a circular hole cut very close to the sample so that incident metal particulates are very perpendicular to the sample. The sample platen cannot be rotated, and the crucible locks in place so that they are always at the same angle (~0 degrees). 

I do find that the power needed to get aluminum to start evaporating is significantly higher than that of Ti or Au. In fact, it's about 4x higher than Ti and 2x higher than Au. I always thought this was more of an inherent property of Al though, as this power doesn't change much from when I initially pre-melt the crucible with fresh aluminum. But I'm sure it's possible that Al is evaporating "hot", which I could see causing damage to the resist. 

John,

If your evaporator gets hot then you certainly run the risk of polymerizing the resist and making it difficult to lift off. You may consider a couple of options:

- Use a newer Al target, with relatively lesser Al in the crucible. Rule of thumb is that the Al melt should be covering ~2/3 of the crucible, to prevent wicking and cracking. You can consider using crucible with thicker walls, so that the internal volume is smaller and you can use less Al. This will help heat the entire melt more quickly and uniformly. Because Al rapidly expands, it usually always cracks graphite crucibles, which can lead to some carbon contamination. You could consider using a Mo crucible instead, which is much stronger but also much more expensive.

- If the evaporation is still running too hot, then consider lifting off in hot Remover PG. PG is specifically designed for removing resists that polymerize as a result of plasma etches or hot processes.

I'm assuming you use PMMA A2 for resolution purposes. If your critical feature resolution is greater than 100 nm, you could consider using PMMA A4, which is typically 200 nm after spincoating @ 3k rpm. If you're more worried about the shape of the features and/or overdeveloping, then you could do a "cold development" in IPA/water for a significantly better resolution and edge roughness profile (I know, it sounds strange): http://nano.yale.edu/sites/default/files/files/pmma_develop_ipa_water.pdf

Dear John,

I would like to add a few suggestions to the above discussion.

When depositing Al, it can wet the resist profile more isotropically than other metals. Thus, increasing resist thickness upto 5x metal height might be helpful. More than that might compromise resolution.

Al grain size, as well as its diffusion into polymers, depends a lot on deposition rate Hence keeping slower deposition rate could give you less diffusion and smaller grain size.

In addition to thermal crosslinking of the resist due to high temperature, e-beam evaporation may cause problems with lift off due to stray radiation generated by the e-beam causing unpredictable exposure in resists. PMMA when overexposed can turn into negative resist and get extremely hard to remove during lift off. It may be useful to try using thermal evaporation instead of e-beam evaporation.

Just to get back to others who may be having problems.

It sounds strange, but it turned out a previous processing step of doing a long UV-Ozone clean is what seemed to be causing the difficult liftoff. We were trying to grow a small amount of oxide using UV-Ozone, so we were doing hour+ UV-Ozone exposures.

This seemed to alter the surface chemistry enough that PMMA had difficulty getting removed from the surface. The recipe I listed works great for aluminum liftoff if you don't do ultra long UV-Ozone exposures. I can't speak for typical UV-Ozone cleans in the ~1-5 minute range.

I think better to use Remover PG

Ravindra Bhardwaj This is what we ultimately used. Remover PG and 1165 heated to 75 C were both very good. Additionally we added in a ~5-10 minute plasma ash to clean up things after the lift off.

Ravindra Bhardwaj Can you tell how many bath you used to do? Generally I read people do two bath at 80 oC followed by IPA and DI water. But, I found residue of PMMA on my sample after 3 bath. And I did AFM to confirm it. Still I am looking convenient way to get rid of completely with this PMMA.

Increasing the resist thickness could help. It did in my case. The recipe was almost the same except for using Sputter instead of E-beam. I agree with Nikhil E-beam can harden the resist. Use A4 instead of A2 in one of the layers, preferably 450K A4 as 450 K has better resist contrast.

My samples were kept for around 20 hrs in PG remover at 75 oC and then used a syringe to shower the PG remover on patterns while still in the bath.

Gap between my features is 50nm.