Scatterometry Overlay

Scatterometry overlay measures how precisely two patterned layers line up by analyzing the spectral signature light produces when it reflects or diffracts off a specialized overlay target — rather than imaging the target directly with a microscope. It's the umbrella category covering both diffraction-based overlay (DBO) and spectroscopic overlay, both of which extract registration error from how light behaves optically rather than from a visual image of the target itself, and both of which trade image-based overlay's intuitive directness for genuinely faster throughput and smaller target footprints.

Scatterometry Overlay: Reading Light Instead of Images Two different ways to pull overlay error out of a target's optical response. DBO -1 order (bright) +1 order (dim) intensity imbalance between +1st/-1st orders ∝ overlay fast, direct, requires calibrated swing curve SPECTROSCOPIC full spectrum shape shift, fit to a model to extract overlay richer signal, more modeling required WHY SMALLER TARGETS WIN µDBO targets shrink to fit in-die, not just in the scribe line — capturing local overlay variation that a scribe-line-only target would simply never see, while 2D grating targets measure X and Y in one shot. Scatterometry's speed advantage over imaging comes from this: less area to scan per measurement point. Faster, smaller, and increasingly as accurate as traditional image-based overlay at advanced nodes.

Diffraction-based overlay reads registration error from a simple intensity imbalance between two diffraction orders, making it the more direct of the two scatterometry approaches, computationally. A DBO target is built from two gratings with deliberate, known, opposite offsets, and when overlay error is present, it breaks the symmetry between the +1st and -1st diffraction orders the target produces under illumination — one order comes back measurably brighter than the other, and that intensity imbalance translates directly, through a calibrated swing curve, into an actual overlay number. The simplicity of this relationship — essentially one measured intensity difference mapping to one overlay value — is exactly what makes DBO fast enough for high-volume production use.

Spectroscopic overlay extracts registration information from the full shape of a target's spectral response rather than from a single intensity comparison, trading DBO's simplicity for a richer signal that can extract more information from the same physical target. Instead of comparing two discrete diffraction orders, spectroscopic overlay measures how a target's entire reflected or diffracted spectrum shifts in shape as overlay error changes, fitting that measured spectrum against a model to back out the actual overlay value. This approach demands more computational modeling than DBO's direct intensity comparison, but it can extract overlay information even from target geometries or measurement conditions where a simple two-order intensity comparison wouldn't provide enough signal on its own.

Micro-DBO targets solve scatterometry overlay's biggest practical limitation by shrinking the measurement footprint small enough to fit directly within the die, rather than needing dedicated scribe-line space. Traditional overlay targets, scribe-line-sized, consume area between dies that can't otherwise be used productively, and they only tell you about overlay at the scribe line — not necessarily representative of overlay inside the actual functional die area. Shrinking the target down to roughly ten microns on a side, with multiple measurement pads packed into that compact footprint, lets it sit directly within the die, capturing overlay behavior much closer to where it actually matters for device performance.

Two-dimensional grating targets measure X and Y overlay simultaneously from a single target structure, rather than requiring two separate measurements at two different target orientations. A target patterned as a genuine 2D grating, rather than two separate one-dimensional gratings oriented perpendicular to each other, captures both the X-direction and Y-direction registration error from one combined optical measurement. This directly cuts the number of distinct measurement sites and measurement time needed to characterize overlay fully across both axes, compounding scatterometry overlay's existing throughput advantage over slower measurement approaches.

Scatterometry overlay's throughput advantage over image-based overlay comes down to a straightforward physical reason: it simply doesn't need to resolve a detailed image, just measure an aggregate optical signal from the whole target at once. Image-based overlay has to resolve individual target features clearly enough to locate their precise centers, which takes real imaging and focus time per site. Scatterometry overlay instead collects an aggregate spectral or intensity signal from the entire target in one measurement, without needing to resolve individual features within it — a fundamentally faster measurement operation that's exactly why scatterometry-based overlay has become the production throughput workhorse at advanced nodes, even as image-based overlay remains the reference technique scatterometry models get calibrated against.

Take scatterometry overlay 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