Hi! I am creating the recipes for our spin coater and also monitoring the thickness of the resist over 1 year already. Target thickness is 4.7um and the control window is +/-50nm. This resist is always drifting: I am adjusting the spinning speed to lower rpm to keep the thickness back in limits and then wait another month until it drops by 100nm. I talked to the tool vendor and they claim that the problem might be in the bowl exhaust, although some recipes have the exhaust step and some don't and the thickness behavior is similar. The resist manufacturer claims that it is only a factor of 2 parameters: spinning speed (its uniformity) and the exhaust of the solvent from the bowl. However the monitor is done in exact same conditions every week, there should not be any clear drift of thickness (if the exhaust is bad for example, then the thickness would have value x and it would stay there).
Do you have a deeper experience about it?
Many thanks in advance!
Hi Olga, do you control for spin process time? Maybe some of the wafers sit in the spin cup for a longer time before going to the bake plates, and those wafers become drier ... i.e. thinner. Others leave faster and look thicker because they have more residual solvent. What I would suggest is double your spinning time during the main spin and see if that improves process consistency.
In addition to the answer from Dustin W. Janes I would also suggest you look at the resist itself. Are you continuing to use the same bottle? If yes solvent could be evaporating over time.
1) Usually resist coats wafer non-uniformly, the film is thinner in the center. So, if you have similar wafers, but measuare in different points, you'll get different thicknesses. E.g. non-uniformity of ~25-30 nm for 4.7 um film is typical for such thick resist. Also, it depends on wafer size.
2) How do you measure thickness? Ellipsometry results depend on refractive index of your film - something could happen and index will be changed, e.g. new bottle of resist, or substrate were not cleaned well and some sublayer left, etc.
AFM/profilometry can make scratches on resist, so the real thickness is large than measured. For this measurements openings in resist are needed, it means you have to make lithography and develop. Unexposed areas (where thickness is measured) are developed very slow, too. This "dark erosion" depend on your coating parameters and developer characteristics, so remaining thickness could be varied.
Hi all and thank you for your answers.
I use an automatic tool that dispenses the same dose of resist, from the same point on all wafers, and all wafers have the same time of spinning. Also the softbake time is theoretically the same in between all wafers. The resist bottle in this tool is not having an opening, hence not exposed to air. I expect to have up to 40nm across wafer THK variation, but not more.
The thickness is measured right after the softbake (and cooling) in a reflectometer, so there is no etching in development.
As you can see it took me long time to reply, but I was just looking at how the tool will behave in order to understand if there is a pattern. And it was. It takes about 5 -6 weeks for the tool to stabilize after changing a spinning parameter (even if changing the acceleration rate only). Now the results look stable without my intervention.
Conclusion: it is not clear what are the tool internal variations, but it takes a month + to get stable resist thickness results.
- Badges
- Science topic
- Similar topics
- Psychometrics
- Measurement