Inverse lithography technology works backward from the desired wafer pattern to compute the mask shape that will best reproduce it under a given optical process, rather than starting from a rectangular design and correcting it. That computation naturally produces curved, freeform mask edges instead of the Manhattan (rectilinear) geometry used in conventional masks. Curvilinear masks can hit a larger process window and lower edge placement error, but they are harder to manufacture and to verify, which is why manufacturability terms like mask rule check and runtime appear alongside the core ILT claims in this dataset.

This 15-record dataset ranks 9 assignees in total, with the leader holding 4 filings and the top 5 combined covering 73.3% of all records in scope. That is a genuinely concentrated field rather than one with a long tail of small filers — the ranked 9 account for 100.0% of the records. Anyone assessing competitive position should treat the top few filers as the ones to check first for overlapping claims.

The raw trend shows a drop from 2 filings in 2021 to 0 by 2024, a -100% change the dataset flags explicitly, after a peak of 4 filings in 2023. However, patent publication typically lags actual filing by around 18 months, so 2025 and 2026 figures in any such dataset are necessarily incomplete and should not yet be read as a sustained decline. The more defensible read is that filing activity peaked in 2023 and the post-peak years are still being counted.

US20240280891A1 is Samsung Electronics' OPC method that builds a curvilinear mask layout by dissecting a rectangular mask into segments, generating shape-variable points shifted toward a rounded target, extracting a contour, and evaluating edge placement error against it. It blocks implementations that follow this specific segment-to-curvilinear-point pipeline with EPE verification built in, but it does not foreclose other routes to curvilinear mask generation — such as full-chip hierarchical ILT approaches, which appear separately and earlier in this dataset's citation record. A design that reaches curvilinear geometry through a different point-generation or contour-extraction method may sit outside its claims, though that needs a claim-by-claim comparison rather than a summary judgment.

Relative to the dense core claims around ILT mask synthesis and OPC, this dataset shows thinner coverage in AI-assisted mask rule checking, multi-beam mask writer calibration, curvilinear contour EPE verification tooling, and sub-resolution assist feature placement automation. G06N (AI-based computing) appears on only 6.7% of the 15 records, versus 93.3% for core G03F photolithography claims, which is a wide enough gap to suggest the AI-assisted verification angle is under-claimed rather than saturated. Any first claim there should be checked against the leading assignees' existing filings before committing resources.