Etch Bowing Budget

# Etch Bowing Budget

An etch bowing budget defines the allowed outward deviation of an etched sidewall from a stated reference and the measurements used to verify it. It keeps buried widening visible when top and bottom dimensions alone look acceptable.

Here bowing refers to local feature geometry. It is distinct from whole-wafer bow or warp, which describes the shape of the wafer itself.

## Inspect the width between the endpoints

The illustrated opening has equal top and bottom widths but expands between them. A measurement at only those endpoints would miss the widest section. The reference lines in this example are straight walls joining the top and bottom positions.

The drawing is schematic. Real features may have a maximum width away from mid-depth, different deviations on the two walls, or several local extrema. A measurement called middle CD is not necessarily the maximum CD unless the profile search establishes that relationship.

## Name the metric before reporting a value

A bow metric might be maximum opening-width expansion, maximum one-wall displacement from a reference, or another specified profile quantity. Those metrics have different meanings. State the reference construction, depth range and sign convention.

In the invented symmetric example, the top and bottom are 40 nm wide and the maximum width is 52 nm. Total width expansion relative to the 40 nm reference is 12 nm. Only with the stated symmetry does that imply 6 nm outward deviation on each wall.

For an asymmetric feature, the same total expansion could be distributed unequally. Measure the walls individually when the requirement applies to each wall, to centerline displacement or to nearby feature spacing. Dividing a width difference by two does not establish symmetry.

For a tapered reference, the comparison width can vary with depth. Do not subtract a top width from every buried width and call the result a reference-wall bow without explaining that convention.

## Connect widening to the integration

Bowing can reduce material separation between neighboring openings or produce a reentrant shape that matters to later coating or filling. The consequence depends on the layout, materials and downstream process.

An acceptable top opening does not establish the buried spacing or the complete fill path. Likewise, a single bow number does not establish bottom clearing, local necking, surface condition or dielectric integrity.

Derive acceptance limits from the intended integration. There is no universal total expansion or percentage bow that establishes manufacturability across different geometries and materials.

## Separate the observation from the mechanism

Primary profile simulations have linked bowing to scattered-particle interactions in defined high-aspect-ratio contact etch conditions. Sidewall protection, mask shape, transport and local electrical conditions can also be relevant in process investigations.

A bowed image alone does not uniquely identify which mechanism dominates in the actual feature. Keep the measured geometry separate from the proposed cause, and test explanations with controlled evidence.

A recipe change that reduces bow can affect rate, selectivity, clearing or another part of the profile. Qualification should include those affected requirements rather than treating bow reduction as the only objective.

## Measure the relevant local geometry

Specify the section location and orientation, edge-detection method, calibration, resolution and uncertainty. An oblique section or preparation artifact can distort apparent dimensions.

For a three-dimensional hole, one section may not capture the largest diameter or worst local wall deviation. Include appropriate spatial sampling and report its limits. Retain representative widths, pitches, pattern densities and wafer positions.

Record whether the measurement is as-etched, after resist removal or after cleaning. Different process states can have different visible boundaries and material conditions.

## Close the budget with a defined acceptance rule

Retain the reference profile, measured widths or wall positions, depth of the extremum and applicable local limits. State whether the rule applies per feature, to a sampled maximum or to another justified statistic.

Carry measurement uncertainty and incoming variation through the decision. Do not replace a local excursion with a favorable average or change the reference after seeing the result.

A useful bowing budget states what moved, relative to which reference, and where the largest deviation was found. Endpoint widths support that description; they cannot replace the buried-profile requirement.

## Sources and scope

• [Profile simulation of high aspect ratio contact etch](https://www.sciencedirect.com/science/article/pii/S0040609006011667): primary profile simulation examining sidewall necking and bowing mechanisms in specified contact-etch conditions.
• [Lam Research — Flex G Series announcement, July 2015](https://investor.lamresearch.com/2015-07-13-Lam-Research-Launches-Flex-TM-G-Series-for-High-Aspect-Ratio-Dielectric-Etch?asPDF=1): historical manufacturer discussion of depth-dependent CD and profile control for high-aspect-ratio memory features.

Original diagrams and symmetric numeric example. Dimensions are teaching values, not a universal bow limit or qualified process specification.