durusmail: mems-talk: SU-8 Cracking
SU-8 Cracking
SU-8 Cracking
Michael Larsson
2008-02-24
Hi Sonic,

Have you considered the increased appearance of cracks during
development to be a function of mask pattern rather than development
time? As you develop a pattern into SU-8, you effectively introduce
holes in a stressed sheet. The holes/pattern act as stress
concentrators that increase local stresses above nominal levels,
making cracking more likely. When your layer is partially developed,
the boundary conditions are not the same as they eventually become
when fully-developed; hence the progression in cracking with time in
developer. Any notion of the developer solution inducing cracking in
SU-8 through some sort of chemo-mechanical process is a red herring.

By 'over'-exposing your layer, you are effectively smoothing abrupt
changes in geometry, reducing stress concentrators. This could be an
explanation for the things you are observing.

Aside from modifying baking and exposure conditions in your SU-8
process, perhaps a mask re-design may pay dividends. Smooth sharp
changes in geometry through the use of filleted radii, and blank off
regions that do not carry a pattern so that you don't lock-in problems
at the design stage.

Regards,

Michael


> Hi,
>
> In my case, after 400mJ/cm2 exposure, only few small cracks appeared on su8
surface. These cracks disappeared after hard bake. I don¡¯t think su8
reflowed during hard bake. But the hard bake did anneal su8 and removed the
small cracks.
>
> About the bubbles, no matter how carefully I operated, there is still one or
two tiny bubbles produced. In my case, this did not matter much, since I only
used su8 as a mask.
>
> There is another thing I do not understand. When I exposed 800mJ/cm2, the su8
can be developed more than 5mins without any cracks. Of course my structure
looks ugly with this exposure dose. On the other hand if I use 400mJ/cm2
exposure, I will get nice features after 2mins developing. But if I keep
developing to 3mins or longer, big and deep cracks show up again. Is this
phenomenon reasonable?  The developing time should be controlled very precisely?
>
> Sonic
reply