Global Water Intelligence
An example of predictive yield modelling, as presented by Dan Wilcox, Principal – Director Process Engineering at Page, at the last UPM Community Event.
Taken from a presentation by Dan Wilcox, Principal of Director Process Engineering at Page, at the last UPM Community Event. Watch the April UPM Community Event recording: 'Deconstructing the Challenges of Facility 2.0: Yield and Reliability Drivers'
The chart shows a model which was created to measure total organic carbon (TOC) in the ultrapure water (UPW) used in the last rinse of the wafer. Instead of directly counting the particles in the water, the method counts the number of carbon atoms on the wafer surface and correlates this with the expected level of water contamination. The method can be useful for circumventing the detection limits of metrology for counting particles in water. Experimentation and modelling in such a way allows engineers to determine what treatment is needed for the UPW before the rinse by monitoring the wafer to avoid reaching the critical number of carbon atoms which would create a defect. Algorithms based on such modelling can then be related back to see what kinds of maintenance might be needed, such as filter, resin, valve or pump replacements.
The variables shown in the graph are the dryer spin speed and time, which determine the thickness of the water shell on the wafer. The number of moles of carbon on the wafer surface is then compared to the thickness of the water shell, allowing for a comparison to TOC levels in the UPW, and to see how far UPW contamination results in potential defects on the wafer.
Wilcox explained that the accuracy of this model was confirmed by experimentation based on dosing the water with different quantities of organics to determine the resulting contamination on the wafer surface.
Haley Oryell
Global Water Intelligence
Orla McCoy
Global Water Intelligence
glen sundstrom
Evoqua Water Technologies