Etch Sidewall Roughness Budget

# Etch Sidewall Roughness Budget

An etch sidewall roughness budget defines the allowed texture of an etched wall, the measurement scale and the evidence needed to assess it. It keeps the roughness metric attached to its reference surface, sampling and process state.

A number labeled roughness is incomplete if the direction, spatial scale, filtering or measurement method is missing. It is also not interchangeable with overall bow, taper or tilt.

## Separate overall shape from texture

Etch Sidewall Roughness Budgetoverall wall shape and small-scale texture are different measurementsOverall shapeCheck bow, taper and tiltLocal textureDefine the reference and scaleSchematic wall traces, not to scale · form removal and filtering must be specified; they change the reported roughness.

A smooth wall can still be bowed or tilted. A nearly straight wall can have fine texture. The distinction depends on the stated form-removal and spatial-scale convention; it should not be chosen after seeing which calculation gives a passing result.

Specify whether the measurement follows depth, follows the wall along a line, or covers a two-dimensional surface. Conventional top-down line-edge roughness does not by itself establish the complete buried sidewall texture.

Cyclic DRIE scallops are one form of sidewall texture. Scallop depth, scallop spacing, profile RMS and areal surface roughness describe different quantities and should not be silently substituted for one another.

## Name the metric and its reference

RMS and Peak-to-Valley Differinvented equal-spaced residuals after a specified reference subtractionPoint 1-2 nmPoint 2+2 nmPoint 3-2 nmPoint 4+2 nmMean residual0 nmRMS residual2 nmPeak-to-valley range4 nmFour teaching points are not an adequate sampling plan · no confidence bound, spectrum or acceptance limit is implied.

For the invented residuals -2, +2, -2 and +2 nm, the mean is zero. RMS is the square root of the mean squared residual: sqrt((4 + 4 + 4 + 4)/4) = 2 nm. The peak-to-valley range is 2 - (-2) = 4 nm.

Those values describe four stated samples relative to a specified reference. They do not establish a population distribution or an adequate measurement plan. RMS and peak-to-valley measure different properties; a limit written for one cannot be applied directly to the other.

Distinguish a one-sigma or RMS quantity from a reported three-sigma metric. For width measurements, the relationship between the two edges also matters: linewidth roughness is not automatically equal to the roughness of either individual edge.

## State the spatial bandwidth

Record the measured length or area, sample spacing, resolution, form removal and any filters. These choices determine which spatial variations contribute to the reported value.

NIST's work on line-edge and linewidth roughness shows that measured RMS depends on sampling length and interval, while imaging noise can add bias and uncertainty. Those metrology principles motivate explicit measurement definitions; they do not make every line-edge method directly applicable to a buried sidewall.

A lower value obtained after stronger smoothing is not automatically evidence of a smoother physical wall. Comparisons need compatible measurement settings or a justified conversion.

## Account for measurement limits

Document instrument calibration, image or probe response, edge identification, noise treatment and uncertainty. Limited access or probe geometry can prevent a method from resolving some sidewall features.

Cross-section preparation can introduce artifacts, and a single section may miss texture elsewhere on a three-dimensional surface. Report where the surface was sampled and how representative coverage was established.

Do not subtract a noise estimate from a measured roughness value without a validated model. Preserve raw or suitably traceable data so the processing choices can be reviewed.

## Connect the requirement to the next step

Sidewall texture can matter to coating, fill, optical behavior or electrical performance depending on the integration. Establish the relevant acceptance metric and spatial scale from that application rather than reusing a universal RMS limit.

A treatment that reduces texture may also change feature width, depth, local profile or material integrity. Qualify the affected dimensions and functional requirements together. A reported smoothing result in one DRIE application does not establish compatibility with every material stack.

## Preserve process state and scope

Record whether the measurement is as-etched, after strip, after cleaning or after a separate smoothing step. Retain mask condition, feature layout, recipe revision and chamber history where relevant.

Include representative depths, feature sizes and wafer positions. A favorable mean roughness does not rule out a local defect or a narrow section, so keep applicable local profile limits visible alongside texture statistics.

A useful roughness budget specifies the surface, metric and scale. The value becomes meaningful when the measurement and acceptance rule refer to the same physical quantity.

## Sources and scope

  • [Villarrubia, NIST — Issues in Line Edge and Linewidth Roughness Metrology](https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=822535): sampling dependence, spatial bandwidth and measurement-noise effects in roughness metrology.
  • [Park et al. — Low-temperature smoothing method of scalloped DRIE trench](https://link.springer.com/article/10.1186/s40486-020-00116-x): primary study of sidewall scallop measurement and smoothing in a specified silicon process.

Original diagrams and four-point arithmetic. No universal roughness limit, statistical confidence or qualified smoothing recipe is implied.

Take etch sidewall roughness budget further

Ask the copilot about this term, or have our engineers assess it against your process.

Ask CFSGPT about this → Request a technical assessment → Open the CFS app