If I understood correctly, the top part is fine (vertical sidewalls)
but the bottom is wider. This may be more to do with poor collimation
of the UV source or diffraction/reflection effects of the mask. (Poor
filtering / low exposures tend to generate structures wider at the top
than at the bottom).
How do the widths of the features compare with the mask?

2008/4/23 Andrew Sarangan :
> This sound like under-exposed SU8. With a broad band source and no i-line
>  filter, I have found that it needs a higher dose than the prescribed amount.
>  Try increasing the exposure time, and I'm sure that would solve your
>  sidewall problems.
>
>
>
>  On Tue, Apr 22, 2008 at 2:21 AM, ֣????(Ruilin Zheng)

>  wrote:
>
>
>  > Hello, everyone,
>  >
>  > I am doing some fabrication work with SU-8 2150 ultrathick layer, up to
>  > 400
>  > microns at a single layer spin coating.
>  > Because I just start my work, so there are a lot of problem stay in my
>  > way.
>  > After scanning the sidewall with laser scanning microscopy, I found that
>  > the
>  > side wall profile is partially vertical. At the part near the substrate up
>  > to about 1/3 of its height, the profile is concave.
>  >
>  > >From 1/3 to total height (up to surface ), the profile is fairly
>  > vertical.
>  > The difference between the concave part and the vertical part is about 40
>  > microns.  The UV source I am using doesn't have a filter for UV-light, so
>  > it is not a very coherent source.
>  >
>  > Does any one meet this problem before? Please give me some advice.