The most-cited prior art in this dataset comes from equipment makers such as KLA-Tencor, alongside ASML and semiconductor manufacturers like TSMC and AMD. Recent-year momentum figures show all of these holders at zero filings in the latest tracked year, which given the 18-month publication lag likely reflects reporting delay rather than an actual drop in R&D. The concentration of citations on a handful of early 2000s-era filings suggests the foundational measurement-model claims are already staked out by these incumbents.

Filings rose from 19 in 2017 to a peak of 58 in 2022, then fell in subsequent years. Part of that decline is real — filing cycles in mature metrology techniques tend to slow once core claims are established — but part of it is a publication-lag artefact, since patent applications typically take about 18 months to publish. The most recent year in the dataset reports zero filings, which should be read as incomplete data rather than a stopped field.

The IPC breakdown shows G03F (photolithography and photomechanics) in 402 of 461 families, by far the largest subclass, followed by G01B (dimensional measurement) at 183 and G01N (material analysis) at 162. H01L (semiconductor devices) appears in 123 records and G06T (image processing) in 41, showing a smaller but present overlap with computational and device-level claims. This spread confirms the field is driven by optical measurement method claims more than by semiconductor device structure claims.

US20180252514A1, filed by KLA-Tencor and published in 2018, covers a method of training an overlay-measurement model from scatterometry data collected across Design-of-Experiments wafers with programmed overlay and process variations, so that actual overlay can be separated from process and system noise. It sits within a densely cited part of the landscape, alongside older KLA and related art on scatterometry-based overlay detection. Anyone filing a new scatterometry-trained-model claim in this space should expect this record, and its cited predecessors, to surface in prior art search.

The IPC data points to lighter claim density outside the core G03F/G01B/G01N overlap — particularly where G06T image-processing methods intersect with defect classification, and in alignment-mark design tuned for multi-patterning schemes rather than single-exposure lithography. Co-assignee data also shows almost no cross-company collaboration, with only 10 co-filing pairs across 461 families, suggesting joint development claims in this area remain largely unexplored. Machine-learning-assisted defect classification and process-window prediction from focus-exposure data are two specific branches worth a targeted novelty search.