Etch Timed Depth Control Budget
# Etch Timed Depth Control Budget
An etch timed depth control budget bounds the depth of a partial etch that has no stop layer, where depth is set by etch rate multiplied by time, and records the rate-stability and feedback needed to hit the target. It treats a blind, timed etch as a rate-control problem, because there is no landing layer to stop on and no clear endpoint signal to stop at.
Many etches clear a film and land on an underlayer, letting a selectivity or endpoint budget define the stop. A partial or recess etch instead stops partway into a material at a target depth. With nothing to signal the stop, the depth depends entirely on the product of rate and time, so any change in rate moves the depth. This budget does not restate the depth-uniformity budget; it addresses the open-loop timed depth itself.
## See why a fixed time is fragile
In a timed etch, the recipe fixes the time and assumes a rate. If the actual rate differs from the assumed one, the achieved depth differs in proportion, and there is no stop layer to absorb the error. The depth lands wherever rate times time puts it.
The sketch shows a partial trench in a thick film with no underlayer, a target-depth marker and an error band that opens if the rate drifts. It is not measured data; a blind timed etch simply has no layer to stop or signal on.
Reported work shows the limitation directly. A recess-etch study of transistor devices requiring nanometre-scale thickness control states that because the etch rate can change due to byproducts in the chamber, a timed etching approach is not effective, and the authors moved to in-situ thickness monitoring instead. The timed method fails exactly when the rate is not perfectly stable.
## Qualify rate stability as the real control variable
Because depth tracks rate, the budget's core is rate stability. The rate can shift with the chamber wall state, with how much area is being etched, and with chamber history, so a rate qualified once does not stay valid indefinitely. Treat the assumed rate as a quantity with its own drift and uncertainty, not a constant.
Record when and how the rate was qualified and how it is monitored. A timed depth that was correct at qualification can drift as the chamber changes, which is why the wall-memory and first-wafer conditions connect to this budget.
## Account for loading and feature dependence
The rate is not one number across the wafer or across features. A deep-silicon loading study reported the wafer-average etch rate falling sharply as the etched area increased, and aspect-ratio-dependent etching makes narrow features etch slower than wide ones. A timed etch therefore produces different depths in different places unless these dependences are included.
State the open-area fraction and the feature widths the timed recipe assumes, and budget the depth spread they imply. A single depth target without a layout context cannot be met everywhere by one fixed time.
Separate a rate-drift depth error from a loading or ARDE depth difference. One is a change over time; the other is a difference across the pattern at the same time, and the corrections differ.
## Add feedback when the tolerance is tight
When the depth tolerance is tighter than the open-loop rate stability allows, the process needs feedback rather than a fixed time. Options include run-to-run control that adjusts time from measured depth, or in-situ monitoring such as interferometry that stops on a measured thickness, as the recess-etch study adopted. Each adds cost and has its own requirements.
No single approach is universal. A timed etch can be adequate for a loose depth on a stable process and inadequate for a nanometre-scale recess, so the method must be matched to the tolerance and qualified on the real film and tool.
## Close the budget with traceable evidence
Specify the target depth and tolerance, the assumed rate and its qualification, the open-area and feature assumptions, the control method, and the measured depth against the target, at a named process state. Retain chemistry, recipe revision and chamber history so a later result can be compared with the earlier one.
A timed depth control budget treats a blind partial etch as rate-stability control, with feedback added when a fixed time cannot hold the tolerance. Rate stability, loading dependence and depth verification are separate questions, and each needs its own evidence.
## Sources and scope
- [Recess Etching Process for AlGaN/GaN-HFET Devices Using In-Situ Monitoring (SAMCO, CS MANTECH)](https://csmantech.org/wp-content/acfrcwduploads/field_5e8cddf5ddd10/post_2136/108.pdf): states that because the etch rate can change due to chamber byproducts, a timed etching approach is not effective for nanometre-scale recess control, motivating in-situ thickness monitoring.
- [Loading effects in deep silicon etching (ResearchGate record)](https://www.researchgate.net/publication/228601870_Loading_effects_in_deep_silicon_etching): reports a strong dependence of etch rate on loading, with wafer-average rate falling markedly as the etched area increases.
Original diagrams and explanatory synthesis. No universal rate, depth tolerance or qualified control method is implied.