Etch Undercut Budget
# Etch Undercut Budget
An etch undercut budget defines how much lateral etching beneath a mask or stop layer an integration can tolerate, and the evidence needed to measure and attribute it. It separates the designed lateral bias of a process from an unintended loss of material that narrows a supporting film or widens a cavity beyond the patterned opening.
Undercut is a geometric consequence, not a single defect type. Isotropic chemistry, ion scattering, mask lift and layer-selective wet steps can all remove material from beneath an edge that the top-down pattern never exposed.
## Separate designed bias from unwanted undercut
A process can specify a deliberate lateral bias to meet a target critical dimension; that bias is designed and controlled. Undercut, by contrast, is the portion of lateral removal that exceeds the intended bias or appears where no lateral removal was planned, most often beneath a stop layer, liner or adjacent film with different selectivity.
Isotropic components of an otherwise directional etch, galvanic or micro-masking effects, and wet steps with poor directionality are common contributors. Attributing a given recession to one mechanism requires the stack, the chemistry history and the inspection geometry together, not the image alone.
## Distinguish undercut from related lateral effects
| Observation | What it describes | What it does not establish |
|---|---|---|
| Undercut | Lateral loss beneath a masking or stop feature | The removal mechanism or chamber source |
| Bowing | Mid-trench lateral bulge in the etched sidewall | A recession beneath an edge above the bulge |
| Notching | Lateral loss concentrated at one interface during an etch step | Loss distributed along the full sidewall |
| Microtrenching | A localized deeper groove near a sidewall base | Any lateral dimension at all |
Each term names a distinct geometric pattern. A report that says only "lateral attack" without specifying which of these shapes was observed cannot be compared against a numeric budget written for one of them specifically.
## Build the undercut budget around a measured reference
Before comparing a measured recession to a limit, fix the reference edge the measurement starts from and the process checkpoint it was taken at. A number taken after a strip step is not interchangeable with one taken immediately after the etch, since cleaning chemistries can themselves attack an already-weakened edge.
Report lateral loss per side and specify whether the reported value already excludes designed bias. A total opening-width change is not the same quantity as a per-edge undercut, and conflating them produces a budget that cannot be traced back to a single mechanism.
## Qualify the consequence, not only the dimension
A stop-layer or liner that loses lateral support can affect downstream fill, contact alignment or mechanical integrity of the structure above it, independent of whether the recession itself exceeds a dimensional spec. Evaluate both the measured lateral loss and its functional consequence for the specific stack before closing the investigation.
An undercut budget ties a measured lateral recession to a fixed reference edge and the specific consequence it threatens, rather than to a single dimension read off one image. Designed bias, mechanism attribution and functional impact remain separate questions that each need their own evidence.
## Sources and scope
- [Lam Research — Plasma Etch Fundamentals](https://www.lamresearch.com/plasma-etch): process description of isotropic and lateral etch components in plasma etch chemistries.
- [Oxford Instruments — Etch Process Control](https://plasma.oxinst.com/media-centre/blog/etch-process-control): discussion of selectivity-driven lateral loss at stop layers and liners.
Original diagrams and explanatory synthesis. No universal undercut limit, mechanism attribution or qualified process fix is implied.