Generally speaking there are two components to this problem: poor adhesion and building up stress in structures, which makes them detatach. To promote adhesion clean wafers well (e.g. plasma), dry them by baking to remove absorbed water before applying resist, you can also use adhesion promoter, like HMDS. It is important to follow the recipes for pre- and post-baking (to have good results having both higher and lower temperatures are important). You can do with two hot plates, but sometimes it is even better to use one (the one with lid and vacuum contact is best to use) and try to change temperature gradually. Instead of placing wafer from one plate to another, you can use same plate and ramp temperature from 65 to 95 and then let it cool down on the same plate. This helps to reduce stress in resist. If you manage to develop the structures, then before molding PDMS it is very important to hard bake the structures to get good mechanical strength and adhesion! You should also ramp temperature slowly up and down. From RT to 200C and back. This makes usually strong and long lasting masters! Good luck!

Indeed Alar tells everything, Piranha etching of wafers, si dessication and slow thermal ramps. We found that post exposure bake at 85°C (a little longer than standard recipe) and slow coool down to ambent temperature can help reducing internal stress due to thermal dilitation coefficient mismatch with Si.

Completely agree with Alar, reducing stress in the SU-8 during the pre- and post-baking process is paramount to success.

I would recommend using a hot plate with a metal block on top - this will provide good heat capacity and a more uniform heat distribution. I often found using a glass dish to cover the sample also provided good results.

You should also check the SU-8, it is common to store in a fridge and allow to acclimatise before opening and application. Incorrect handling can lead to lots of problems.

Thanks Alar and Benoit and Jules, I appreciate this, I will definitely try to reduce stress in SU-8 and care about the poor adhesion.

dip the wafer in BOE buffer and then bake it to 120 degree can help this detach problem.

I once had the same problem, and I am mostly agree with Alar. From my own experience:

1- HMDS does not really make a difference with SU8.

2- Pre-cleaning Wafer with Acetone/IPA/DI water and then having Dehydration bake is crucial (I was doing it DHB at 200 C for at least 30 min)

3- Lower PEB and SB temperatures with longer durations are really useful (as Alar said, avoid sudden temperature changes)

4- I have seen delamination many times during Hard Bake. Therefore, I would suggest you to skip that step or do it at no more than 60 C.

Also check this really useful website: http://memscyclopedia.org/su8.html

Thanks, Pei-Yin and Utku, I appreciate your helps.