The model is part of the measurement.
Ellipsometry measures how reflection from a sample changes the polarization of light. The material properties we want come from interpreting that response through a model of the sample.
At Intel, I developed spectroscopic ellipsometry models for novel semiconductor materials on 300 mm wafers. Those models supported mapping of film thickness, optical constants, and optical band gap, and were transferred to the inline metrology team for process control.
This modeling work sat within my responsibilities for industrial CVD/ALD tool ownership and novel back-end-of-line (BEOL) film development. I worked through tool bring-up, qualification, and process stabilization; metrology provided the feedback needed to evaluate the films and processes I was developing.
From research analysis to a repeatable workflow.
A research model has to become something other people can use. My work connected optical modeling and characterization with the practical requirements of pilot-line process development.
At HRL, I developed optical models and production recipes for ellipsometry and reflectometry, together with calibration and qualification methods. The task was to turn an expert-dependent interpretation into a measurement that other users could execute reproducibly.
- 01 / MEASUREUnderstand the instrument and the sample.
- 02 / MODELConnect the optical response to physical parameters.
- 03 / INTERPRETMake the result useful in its process context.
A low fitting error can hide the wrong physical answer.
In multilayer ellipsometry, correlated thickness and optical-property parameters can produce a mathematically excellent fit that is not physically unique. I validated models using known individual-layer measurements, structured design-of-experiments samples, and secondary characterization where necessary.
Those checks exposed models that fit the data well but allowed physically incorrect solutions. Constraining the solution space made it possible to turn the model into a qualified production recipe that other engineers and technicians could use.
Carry that validation into production.
Calibration, reference structures, documented workflows, and ongoing statistical monitoring keep a qualified recipe useful across repeated measurements and realistic sample variation. Python and SQL automation connected those measurements to visualization, reporting, and process feedback.