Critical-Dimension Scanning Electron Microscope
CD-SEM and CD-AFM both measure a pattern's critical dimension directly, by imaging the actual feature rather than inferring its size from an optical model the way scatterometry does — CD-SEM with a scanning electron beam, CD-AFM with a physical probe tip tracing the feature's profile.
CD-AFM reaches a sidewall profile CD-SEM fundamentally cannot, because an electron beam looking down from above simply can't see a feature's hidden undercut. CD-SEM images from essentially a top-down view, so it measures top-surface width well but has limited ability to characterize a sidewall angle, footing, or undercut hidden beneath an overhang; CD-AFM's physical probe tip traces the actual 3D surface contour as it moves, capturing sidewall shape and undercut geometry that a top-down electron image can't resolve — at the cost of being far slower per measurement than CD-SEM's fast electron scan.
That speed gap is exactly why CD-SEM runs as the routine, high-volume CD monitor while CD-AFM stays reserved for targeted characterization work. A fab needs to check critical dimension on a large sample of production wafers continuously, which only CD-SEM's throughput realistically supports; CD-AFM's much slower point-by-point probe trace makes it impractical for that volume, so it gets used selectively — qualifying a new process, investigating a specific sidewall-shape problem, or providing ground-truth validation for a faster, optically-modeled technique like scatterometry.
CD-SEM and CD-AFM both measure dimension directly rather than inferring it from a model, and the choice between them comes down to exactly the tradeoff CD metrology always runs into: how much of the feature's full 3D shape you actually need to see, against how fast you need to see it.