Looks like over exposure. Some film stress and cracking is normal in thick SU-8 layers, but if you get too much, you'll get adhesion failure. A hard bake will normally reflow some of the un-crosslinked SU-8 into the cracks and make them disappear but your structure must survive the develop process to make that useful. Use your exposure step to give yourself just enough selectivity to get your pattern developed and then do the rest of the crosslinking in the hard bake. Also, try using an anti-reflective layer under your glass for better control of the exposure dose.

Hello Khashayar,

As a frequent contributor to ResearchGate with lots of experience, Ron has provided a solid answer.

MicroChem suggests a hardbake above 150 oC for reflow (see attached data sheet). I would also suggest using a good quality hot plate with a large solid metallic block to provide good heat capacity for your process flow. I found there was a massive difference in bake times for different substrates (e.g. silicon and glass) and also for different glass thicknesses. If possible create some test structures and see what parameters work best for you.

In terms of exposure, as Ron suggests, definitely try using an anti-reflective layer under the substrate (even wafer dicing tape should help). For thicker films (to help prevent T-topping), you should use a long-pass filter to remove wavelengths below 350 nm. We use one from Omega Optical (pn: ZJB360, or now PL-360LP). 

Let us know how you get on.

Jules

http://www.microchem.com/pdf/SU-82000DataSheet2000_5thru2015Ver4.pdf

Dear Ron and Jules thank you very much for your technical guidance. To alleviate the crack problems I ramped the temperature variations during prebake and postbake which really decreased the crack intensity and dimensions but it still exists. My exposure dose is adopted from microchem's protocol according to my resist thickness (50 microns), nonetheless, we lowered the exposure time but the cracks didn't decrease.

We have studied bake and postbake time duration effects on delamination which did not give us even a hint to the problem. I guess since I am not working in a clean room, maybe it is a contaminant issue. In fact, as we don't have a chemical hood, we can't use sulfuric acid and H2O2 solutions for our substrate preparations. We just use acetone and IPA rinsing followed by a dehydration on a hot plate. 

We also lowered development time but it led to under development (indicated by white color after dispensing some Isopropyl alcohol to it).

Another probability is the glass material. Our substrate is a glass slide, cut by a glass cutter. Do you recommend changing the glass type? maybe using conventional glass pieces being 3 or 4 milimeters thick?

I like to think of the data sheet only as a starting point for developing a protocol. If you are doing anything other than exactly what they are doing in the data sheet, then you will need to adjust your parameters. This includes using exactly the same tools as they did when creating the data sheet (which I don't believe they disclose). Having said that, I also feel like the data sheet misses a couple key points about their own material and the basic mechanics of thermal expansion and film stress. To their defense, it is often important in a production facility to find the fastest possible protocol in order to maintain reasonable production line times, but in a research setting, this is often not important.

In the data sheet, they give very short soft-bake times. With other resist types this bake step will chemically change the resist, changing the exposure and develop parameters as well. With SU-8, this step is only removing the solvent that allows the resist to flow in the spin step. No crosslinking is happening during this step. This bake can (and should) be longer to insure that all of the solvent is removed from the SU-8. This will help reduce film stress, improve exposure, and increase the reflow (Tg) temperature of the SU-8.

In the data sheet, They use a two step bake on every bake step which is really a bit backwards for the purpose. On the soft-bake, there is no need to pre-warm the substrate since you are going well over the glass transition temperature (Tg) of the material, you simply need to allow ample time for relaxation while above the Tg. You do need to make sure that the substrate does not cool too quickly. A point completely missed by the data sheet. If you remove your substrate from the hotplate right after the bake, where do you put it? On a 20oC countertop or wafer chuck? Allowing the substrate to cool slowly on the hotplate to around 40oC or below will help to reduce film stress and subsequent cracking. As a side note, you should probably cover your substrate while on the hotplate if you are not working in a cleanroom. By having a longer soft-bake, you will find that you need a lower exposure dose to achieve proper develop selectivity. I typically use about 70% of the  exposure dose suggested by the data sheet.

On the post exposure bake, things are a bit more finicky. At this point you have produced the Lewis acids that are going to start the cross linking process. Here you need to balance diffusion and thermal process temperatures a little more carefully. Again, the warmup is probably less critical than the cool down but you need to do it all fast enough that there is not significant diffusion of Lewis acids before the develop step or your pattern will start to loose fidelity.

The Hard bake needs to be done with a ramp up, ramp down procedure. In this case, you''l introduce film stress on any temperature change, so you need to do everything slowly.

One comment on exposure intensity. If your intensity is too high, you can cause heating that is localized to the exposed areas. This can also introduce stress asymmetries that will cause cracking at edges of your pattern.

I have produced may SU-8 patterns on glass of varying sorts (slides, cover glass, mask plates, etc...) so I don't think it is the glass that is the problem, just that the process required for glass is different than silicon. For instance, glass has very different thermal properties than silicon wafers, so bake times must be adjusted to compensate for these variations.

I hope you find some of this useful and best regards!

Dear Mr. Reiserer

I am so grateful for your replies. I used your advice. We finally managed to overcome the problem of delamination. The answer was in the one day rest after the prebaking. We simply let the prebaked SU8 stay in a place without light and did the same procedure we used to after 24 hours. No resist separations from the substrate were detected anymore. 

By the way, don't think I am done with it. I need more help and advice.

Thank you very much (to Mr. Reiser and Mr. Hammond) for your kind replies.

Thank You, everyone, for Nobel note.