Etch Recipe Transition Budget

# Etch Recipe Transition Budget

An etch recipe transition budget defines how changes between recipe segments are timed, observed and accepted. It records what happens between a command change and the process condition required for the next operation.

A programmed segment duration is not automatically the duration of a stable wafer condition. Equally, a useful cyclic process does not necessarily require a fully settled state in every segment. The acceptance rule must match the intended operation.

## Follow the transition through three levels

Separate the commanded setpoints, the equipment response and the wafer result. A controller log establishes what was requested. Suitable telemetry can characterize response. Relevant wafer measurements establish the resulting material and feature condition.

For time-multiplexed silicon etching, the Bosch process alternates etching and passivation. The cited primary study describes equipment with different gas-delivery characteristics. Its timing observations belong to those configurations; they are not universal transition specifications.

## Define the start and end events

State the event that starts the interval: a controller command, valve action, recorded segment boundary or another defined marker. State the event that ends it and the timestamp source used for each event.

Do not subtract timestamps from unsynchronized clocks without accounting for their relationship. Sampling rate, transport delay, filtering and missing data can affect the timing evidence.

A command can precede the measured response. A sensor can also observe a different location from the wafer. Document what each signal measures and the limits of using it as a proxy for the process condition.

## Make acceptance explicit

In this invented teaching example, three signals enter their respective accepted intervals at 10, 15 and 20 seconds and remain there. Their first simultaneous overlap is therefore 20 seconds. The example has no additional dwell requirement and recommends no timing for an actual etcher.

A real acceptance rule may require continuous persistence for a specified dwell, additional signals or a different event-based criterion. If a signal leaves its accepted interval, the rule must define whether the dwell resets or the segment is rejected. A single threshold crossing does not prove sustained qualification.

Keep limits, signal validity, sampling and persistence requirements together. Do not declare the wafer condition qualified solely because the slowest displayed signal has crossed a threshold.

## Count transient exposure

Material can experience changing conditions during a transition. Omitting that interval from the steady segment label does not remove it from the wafer history.

For a quantity that can legitimately be integrated, distinguish the measured or modeled time-dependent value from a setpoint multiplied by elapsed time. The validity of that calculation depends on the quantity, measurement and process model. There is no universal scalar exposure that predicts every etch outcome.

Qualify the complete sequence against the required removal, profile, selectivity and material integrity. Do not assume transition effects can be corrected by adding the same amount of main etch time.

## Treat cyclic operation on its own terms

A deliberately pulsed or alternating process can use repeatable transient behavior as part of its operation. Requiring each pulse to reach an assumed steady state can misrepresent the intended process.

For such sequences, characterize the relevant cycle timing, repeatability and accumulated wafer result. Specify the evaluated cycle population and any startup or ending cycles separately when their behavior differs. A qualified transition criterion for a continuous segment cannot automatically be reused for a short pulse.

## Preserve the equipment context

Record the recipe revision, relevant hardware configuration, signal definitions and timing method. A changed delivery path, control setting or measurement configuration can require requalification of the transition evidence.

Assess representative layouts and wafer locations when qualifying the material result. Stable telemetry alone does not establish uniform geometry or functional acceptance across a wafer.

## Close the budget with evidence

A complete record identifies the requested change, observed interval, applicable acceptance rule and resulting wafer measurements. Keep uncertainty and unresolved proxy limitations visible.

The budget connects recipe timing to demonstrated process behavior. It provides a defined basis for accepting a transition without treating every setpoint change as an immediate wafer-state change.

## Sources and scope

• [Tillocher et al. — Comparison between Bosch and STiGer Processes for Deep Silicon Etching](https://pmc.ncbi.nlm.nih.gov/articles/PMC8537062/): primary study of alternating silicon etch processes and equipment-specific implementation.
• [Oxford Instruments — Deep Silicon Etch](https://plasma.oxinst.com/products/dsie/): manufacturer description of alternating Bosch etch and passivation operation.

Original diagrams, illustrative timeline and explanatory synthesis. No universal settling time, signal tolerance or qualified process recipe is implied.