Glass substrates should be thoroughly cleaned by using the vapor degreasing methods, and should be left untouched on the surface after cleaning. Try to use good quality clean tweezers.  Cleaning the substrates well is the best way out.

K. Sreenivas

SU-8 can be tricky. There are a number of things that can influence adhesion. Exposure has to be just right, too much and too little can both cause adhesion problems. If you have a vapor priming oven, HMDS is one of the best adhesion promoters. HMDS is, in my opinion, better than Omnicoat for adhesion even when spin coating. Too much Omnicoat can be bad. I sometimes use Omnicoat as a release layer although my preference is PVA. Try plasma treating your glass before applying your adhesion layer. Watch your coverglass during your post exposure bake. If the glass starts to bow on the hotplate then you have probably over exposed the SU-8. Glass and SU-8 have different thermal expansion coefficients which make it difficult to put thick layers on thin glass like cover slips.  You may want to try multiple thinner layers to achieve the desired thickness. Thermal shock can also be problematic. Try ramping up and down the temperature on you bakes and never take your substrate off of a 95oC hotplate. Let it cool below the Tg (~60oC) on the plate.

Good Luck!

I thoroughly clean glass substrate with acetone and isopropanol. I treat the substrate in Piranha solution at 80 degree Celcius for 20 minutes.

http://dspace.cc.tut.fi/dpub/bitstream/handle/123456789/20657/kunwar.pdf?sequence=3

@ Ron Reiserer

Thank you very much for your inputs. I initially tried with HMDS. I quickly moved to Omnicoat, as the structures were not adhering properly with HMDS. As you pointed out precisely, the cover glass was bending after the exposure. I thought it was due to vacuum force in the spinner, but it was not. I will try again with your suggestions.

@ K. Sreenivas

Thank you for your input. Earlier, I have not cleaned the substrate properly before applying adhesion promoter. I will clean wafers by Piranha cleaning this time.

@ Puskal Kunwar

Thank you Puskal. I will do proper cleaning this time.

Piranha not only cleans the substrate but also leaves hydrogen moieties on glass substrate which promotes adhesion. However, my suggestion is 'Do not forget the step post exposure bake'. In case of SU-8,  polymerization completes not after exposure but after post exposure bake.

Dear Anvesh,

SU-8 photoresist is a non-polar solution. To make good adhesion of SU-8 the glass surface should be hydrophobic. Thus. plasma cleaning may useful to make better surface for your requirement.

Dear all,

I have done piranha cleaning this time before spin coating and lithography. As you can seen in the image, the structures were not clearly formed. Ideally, a parallel set of lines of width 20 microns should be formed. But, the lines were disturbed. Is it because of the  exposure? I maintained exposure based on the SU8 2050 document. Even I tried to over expose. But as Ron said, there was bowing of glass slide when I over expose. 

Please share your views.

Did you do post exposure baking?? I used to get partially polymerized line when I skipped this step.

 Hi Puskal, 

Yes, I do post baking after the exposure. For the post bake, I follow the SU8 2050 data sheet @ 65 degrees C for 2 mins and 95 degrees C for 6 mins.

In that case, increasing the exposure doze might help.

Try to use BARC (backside anti reflectant coating) like PGME, it'll be both adhesion promoter as well as blocking the UV reflection from the glass.

PEB is increasing adhesion but in the other hand also induce internal stress. Considering the CTE different between glass & SU8 is very high, the thermally induced stress will also increase. I suggest to do a low PEB temperature in a longer time, i.e 65C for 20 minutes, but make sure you don't undercrosslinked. I did this PEB temperature for 50um thick feature and never get undercrosslinked

For a strong bond between SU8 and glass, there are critical elements that you need to consider:

First, you need to make sure your substrate is super clean and super dry. I recommend using Piranha solution if possible. If not, clean with Acetone/IPA in an ultrasonic bath. I do the latter as I have features that get destroyed in Piranha. Then, you need to make sure your sample is completely dehydrated. For that, an overnight bake in oven at 150C is recommended. An overnight bake might be overdoing it but I get perfect adhesion. 

Finally, if your features are entirely detached from the surface, you might need to lower the exposure dose. 

The bake time either pre-bake or PEB is also very important. If you post your recipe, it will be easier to know what step you need to improve. 

@ Paresa Modarres

Thank you for your inputs.

I am trying to fabricate SU-8 features of 20 micron size on a 0.17mm glass slide. 

I cleaned glass slide with Piranha and then dehydrated for only 30 mins at 100C. Then I spin coated HMDS before spin coating SU-8 2050. I followed pre-bake procedure according to the document. Then I played with exposure steps since the glass is transparent. Exposure levels varied from 1 -1.5 times compared to what I generally follow for silicon. Then the normal post-bake procedure was followed according to document again. However, I did not find a significant improvement in the features. Most of the times once the development was done, features used to washed away. Or features seemed under exposed after the UV exposure.  Can you suggest any improvement?

Thank you.

According to your recipe, I think the first thing would be allowing more time for dehydration at higher temperatures in an oven. For my specific case, I use SU8 2035 for 40 um features and my bake times are nowhere close to the data sheet. For soft bake, I do the following: I bake at 67C for 9 minutes on a hot plate and then I raise the temperature to 97C on the same hot plate. This way, the temperature is slowly raised to 97C. I leave the sample at 97C for 30 min. For the PEB bake, I have: 67C (2 min) followed by 97C (5 min). 

The bake time on the data sheet does not apply to glass substrates. For glass substrates, you need longer time. For your information, I use 1.1 mm glass slides.

I had the same problem, my entire SU8 layer came off in one piece! Adhering to all proper cleaning, dehydration, and baking times, the last thing that resolved my issue was lowering the exposure does. If your problem is exposure dose, you should observe noticeable changes by changing the dose. For example, I saw significant differences going down from 400 mJ/cm2 to 350 and then down to 300 at which I had no delamination. By the way, I use no filters. 

As suggested by Paresa, The data sheet is worthless for glass. The data sheets may be a good starting point but I think most peoples working protocols differ greatly from the data sheet, even on silicon wafers. To start with, the bake times on the data sheet are way off. Unlike most other resists, SU-8 has no thermoset characteristics to worry about. The soft bake can be very long and you won't change your develop characteristics like with thermosetting resists. Too short though will cause you to under expose and possibly cause the mask to stick. The point of the soft bake for SU-8 it to drive out solvent and raise the glass transition temperature. No cross linking occurs in this step. What the data sheet fails to inform you about is that the solvent is a photo-inhibitor. The reason the exposure dose is so high in the data sheet is that the suggested bake does not get rid of all of the solvent and you have to expose more to compensate. You may also notice that the exposure dose is a quite wide range rather than a fixed line. This is most likely because the amount of solvent is highly variable in their experiments. If you drive it all out with a longer bake the exposure dose is much more consistent. I usually expose about 70% of the bottom line on the data sheet.

Next, there is a problem with the way the data sheet tells you to step the bake. The whole point of stepping a bake is to reduce film stress.  In the case suggested by the data sheet you step up but not down on all bakes. This is literally meaningless for the soft bake because you exceed the glass transition by 30oC and remain at that temperature long enough for the stress to relax. Most of the residual stress is caused by cooling the material too fast. If anything, you should ramp the temperature down after the bake. The same applies to the post exposure bake. When the post exposure bake begins, there is not a significant amount of cross-linking and you will again exceed the glass transition temperature for some time that can allow relaxation. The cool down is again where you need to be concerned about residual film stress. Only in the hard bake is the warm up important and here you should both ramp up and down at about 1-3oC perminute.

The bottom line here is that you really have to do a process optimization for each individual use case for SU-8 and the data sheets are only a close approximation of a starting point.

I started using the filter plate from Omega that is recommended in the SU-8 data sheet and that really helped. What wavelengths are you exposing with? SU-8 is very sensitive to wavelengths below 350nm. If you filter those out it will (1) reduce any T topping effect and (2) make the expose dose a lot close to the values on the data sheet. Before using the filter I was exposing at about 30% to 40% below the value on the data sheet. Now with the filter I am about 10% below that data sheet. Also I was also getting an issue with delamination/lifting of the entire SU-8 structure from my glass substrate and I found that this was related to an over-exposure. So that may also be the case for you as well.