Etch Open Area Endpoint Budget
# Etch Open Area Endpoint Budget
An etch open area endpoint budget accounts for how the strength of an optical endpoint signal scales with the exposed area of the film being etched, and bounds whether a reliable endpoint can be detected at all. It treats the open-area fraction as a first-class input to endpoint planning, not an afterthought.
Optical emission endpoint detection watches a plasma emission line tied to a product or reactant of the film being cleared. The size of the change at endpoint depends on how much of that film is exposed, so a pattern with very little open area produces a small signal that can be lost in noise. This budget does not restate the endpoint-margin budget; it concentrates on the detectability problem set by open area.
## See the signal shrink with open area
When a large fraction of the wafer is the film being etched, clearing it changes the monitored emission sharply and the endpoint is easy to see. When only a small fraction is exposed, the change at endpoint is proportionally smaller and sits closer to the baseline fluctuation of the signal.
The sketch contrasts a clear endpoint drop at large open area with a trace buried in noise at small open area. The traces are schematic and carry no measured values; the usable signal depends on the film, the optics and the detection algorithm.
Reported work is explicit about the threshold. A study of plasma-etch endpoint detection states that for critical contact and via steps the exposed area of the film is very low, often under one percent with the rest masked, and that traditional endpoint detection fails there because of the low signal-to-noise ratio at endpoint.
## Treat open area as a signal-to-noise problem
The limiting quantity is signal-to-noise, not open area alone. A via-etch study identifying signal-to-noise as the main critical parameter tied it to the hardware, the minimum open area to be etched, and the detection algorithm together, and noted the difficulty when open area is below about one percent.
Estimate the exposed fraction for the actual layout and compare the expected endpoint change against the measured baseline noise of the signal. A pattern that endpoints cleanly at high open area on a test structure may not produce a usable endpoint on a low-open-area product pattern.
## Plan around a weak or absent endpoint
When the open area is too small for a reliable optical endpoint, the process needs a defined fallback rather than a hope that the signal appears. Options include a timed etch calibrated on test wafers, improved collection optics or wavelength selection to raise the signal, a more sensitive detection algorithm, or a different endpoint method such as interferometry where the film and geometry allow it.
Each fallback has its own requirements. A timed stop depends on rate stability and the loading budget; interferometry depends on the film stack and refractive properties; algorithm changes depend on having a repeatable baseline. State which fallback is relied upon and what qualifies it.
Separate an open-area detectability problem from a true process drift. A missing or weak endpoint can mean the signal is below noise, not that the etch behaved differently, and confusing the two misdirects the investigation.
## Tie the endpoint plan to the result
Endpoint detection serves a correct stop, so the budget closes on the outcome. Specify the open-area fraction, the expected signal-to-noise, the chosen detection method or fallback, and the resulting stop accuracy for the actual pattern, then confirm it rather than assuming an endpoint will be available.
No single method is universal. A method that works at high open area can fail at low open area, and raising sensitivity can also raise false triggers, so the choice needs qualification on the real film, pattern and tool.
## Close the budget with traceable evidence
Specify the open-area fraction, the monitored signal and its noise, the detection method or fallback, and the stop accuracy achieved, at a named process state. Retain film, pattern, optics and recipe revision so a later result can be compared with the earlier one.
An open area endpoint budget ties the detectability of an endpoint to the exposed area and the signal-to-noise it produces, and names a fallback when that is too low. Signal strength, detection method and stop accuracy are separate questions, and each needs its own evidence.
## Sources and scope
- [Multivariate endpoint detection of plasma etching processes (Goodlin, MIT)](https://mtlsites.mit.edu/researchgroups/Metrology/PAPERS/ECS02-Endpoint-Goodlin.pdf): reports that for contact and via etches the open area is under about one percent and that traditional endpoint detection fails there because of low signal-to-noise at endpoint.
- [Endpoint Detection Using OES in Via SiC/GaN Fabrication (CS MANTECH)](https://csmantech.org/wp-content/uploads/2022/09/6.2.2021-Endpoint-Detection-Using-OES-in-Via-SiC_GaN-Fabrication_Final_Paper_07.04.21.pdf): identifies signal-to-noise as the main critical parameter, tied to hardware, minimum open area and algorithm, with difficulty below about one percent open area.
Original diagrams and explanatory synthesis. No universal open-area threshold, signal level or qualified method is implied.