Critical Dimension AFM

Critical dimension AFM (CD-AFM) measures the width, height, and sidewall shape of semiconductor features by physically tracing a specially-shaped probe tip across the feature's cross-section, reconstructing the full profile with sub-nanometer reproducibility. It functions as the NIST-traceable reference standard that faster, non-destructive techniques like optical scatterometry get calibrated against — the ground truth every other CD metrology method ultimately has to agree with.

Why CD-AFM Needs a Boot-Shaped Tip A straight tip can't reach behind an overhanging sidewall — a flared one can. STRAIGHT TIP — MISSES THE OVERHANG re-entrant sidewall straight tip can't probe the undercut area BOOT-SHAPED (FLARED) TIP — REACHES IN flared tip profile follows the undercut directly THE MEASUREMENT IS A DILATION, NOT A PURE PROFILE What CD-AFM actually records is the tip's own shape convolved with the sample's shape — tip calibration against known reference gratings is essential, because the raw trace alone can't separate tip from sample. CD-AFM's sub-nanometer reproducibility is exactly why it's the reference OCD models get calibrated against.

Boot-shaped, flared tips exist specifically to solve a geometric problem a straight AFM tip simply can't handle: probing a re-entrant sidewall that curves back underneath itself. Many real semiconductor features — lines, trenches, contact holes — don't have perfectly vertical sidewalls; some curve inward or outward in ways that create an overhang a straight-sided probe tip can't physically reach underneath. A flared, boot-shaped tip has a wider section near its apex specifically so it can swing into that overhang and trace the sidewall's true shape, including the part a simple vertical probe would have no way to contact at all.

The full cross-sectional profile CD-AFM reconstructs — top CD, bottom CD, middle CD, sidewall angle, and overall height — delivers far more information than a single width measurement, which is exactly why it functions as a true reference standard rather than just another CD-measuring tool. A feature's dimension genuinely can vary at different heights through its cross-section, and knowing only a single "the CD is X" number hides exactly that variation. CD-AFM's complete profile captures the entire shape at once, which is precisely the kind of comprehensive ground truth needed to validate whether a faster technique's simplified output actually represents the real feature geometry accurately.

Tip-shape calibration against known reference structures is absolutely essential, because what CD-AFM actually measures is a mathematical dilation — a convolution — of the tip's own shape with the sample's true shape, not the sample's shape in isolation. Every raw AFM trace is inherently a blend of the probe's own geometry and the sample's real geometry, meaning the measured profile only reflects the sample's true shape once the tip's own contribution has been mathematically separated out. Without accurate, up-to-date knowledge of the tip's actual current shape, there's no way to perform that separation correctly, which is exactly why rigorous, frequent tip-shape calibration is treated as a mandatory, non-negotiable part of every CD-AFM measurement rather than a one-time setup step.

Tip wear during normal use directly degrades measurement accuracy over time, which is why characterization gratings — reference structures with precisely known geometry — get measured routinely to track exactly how a tip's actual shape has changed since it was last calibrated. Every physical contact an AFM tip makes with a sample surface gradually wears and reshapes the tip itself, meaning a tip's effective shape today isn't necessarily the same as it was when it was first calibrated, or even yesterday. Measuring a known reference grating periodically lets the measurement system detect and correct for this gradual tip-shape drift, maintaining measurement accuracy over the tip's operational life rather than assuming its shape stays fixed indefinitely.

CD-AFM's role as the calibration anchor for optical scatterometry (OCD) models is exactly why its own accuracy matters so disproportionately, well beyond any single measurement it performs directly. Optical scatterometry infers dimensions from a model relating structure geometry to optical spectrum, and that model needs real, trusted dimensional data to calibrate against — CD-AFM provides exactly that reference data, meaning any systematic error in CD-AFM's own measurement propagates directly into every scatterometry model calibrated against it, across potentially the entire production flow that scatterometry serves. This outsized downstream dependency is precisely why CD-AFM's NIST traceability and rigorous tip-calibration discipline matter so much more than they would for a technique used only in isolation, rather than as the trusted foundation other, faster production techniques are built on top of.

Take critical dimension afm further

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