Etch CD Budget

# Etch CD Budget

An etch CD budget tracks how a defined critical dimension changes through pattern transfer and whether the final width stays inside its allowed range. It separates the nominal transfer change from process variation and measurement uncertainty.

CD means critical dimension. It might be the width of a material line, an opening diameter or a selected width within a cross-section. Those are different measurements; name the feature, location and measurement method before combining numbers.

## State the sign convention

This guide defines signed transfer change as final material CD minus the incoming reference CD. The incoming reference can be a patterned resist or hard-mask width, provided that reference and the final target measurement are explicitly specified. This definition does not imply that the two measurements are taken in the same material or at the same physical height.

For the illustrated material line, a 102 nm reference transfers into a 100 nm final line, giving a change of -2 nm. If the reduction is symmetric about an unchanged centerline, the total 2 nm narrowing corresponds geometrically to 1 nm at each edge. Symmetry is an assumption, not something a single linewidth measurement proves.

An opening may widen while an adjacent material line narrows. A process document may also use the opposite sign or quote a per-side bias. Record the convention rather than treating every number called etch bias as interchangeable.

## Carry nominal values and variation separately

Suppose the required final width is 100 ± 5 nm. Choose a nominal incoming reference of 102 nm and an assumed nominal etch change of -2 nm. Their nominal sum is 100 nm.

For this deterministic teaching example, let the incoming width be bounded by 102 ± 2 nm and the transfer change by -2 ± 2 nm. The lowest modeled result is 100 + (-4) = 96 nm; the highest is 104 + 0 = 104 nm. Thus the modeled final bound is 100 ± 4 nm, within the assumed 95–105 nm specification.

The 1 nm remaining on each side is only a nominal unallocated bound in this simplified model. It is not a demonstrated manufacturing margin. Other transfer steps, drift and measurement uncertainty still need to be included without counting the same contribution twice.

These ± values are assumed bounds, not standard deviations. Adding their magnitudes gives a conservative interval bound when both bounds apply. Do not replace that calculation with a root-sum-square combination unless the inputs are defined statistically and the relevant correlations are addressed. Conversely, a bounded interval calculation does not establish a probability of passing the specification.

## Budget the actual transfer sequence

A resist-to-BARC opening, hard-mask transfer, main etch, overetch and post-etch clean can each alter the chosen CD. Measure at defined stages or use a qualified model. Avoid adding a stage correction to a full-sequence bias that already includes that stage.

The transfer change may depend on incoming width, local density, sidewall passivation, mask erosion and exposure history. A nominal correction measured on one pattern should not be applied automatically across every layout. Check isolated features, dense patterns and relevant wafer positions.

## Match geometry and metrology

Top, middle and bottom CD can differ when the wall tapers or bows. A top-down measurement does not by itself establish a buried bottom width. Resist and target materials can also interact differently with a measurement method. Separate a physical transfer change from tool or material-dependent measurement bias.

Validate the modeled width window with suitable reference measurements, profile information and repeatability checks. CD alone does not establish edge placement, alignment or line-edge roughness; those can require separate budgets and acceptance criteria.

## Engineering takeaway

Specify the width, reference and sign first; then budget the mean change, variation and measurement uncertainty. A final nominal width on target is useful only when the relevant feature distributions and profiles also meet the process requirements.

## Sources and scope

• [NIST — Accuracy Considerations for Critical Dimension Semiconductor Metrology](https://www.nist.gov/publications/accuracy-considerations-critical-dimension-semiconductor-metrology): reference measurements and dimensional accuracy.
• [2024 IRDS Metrology](https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=959302): critical-dimension measurement capability and etch-bias challenges.
• [Lam Research — Etch Essentials](https://newsroom.lamresearch.com/etch-essentials-semiconductor-manufacturing): etch profile, selectivity and CD control context.

Original illustrations and calculated interval example. Dimensions are teaching values rather than a device-specific process specification.