每日资讯Daily Notes
从 AI-SEM 叠对到原子尺度写入：把辅助判断与样品验收分开确认From AI-SEM overlay to atom-scale writing: separate assisted judgement from sample acceptance
依据 NIST 的公开项目整理。物理模型与 AI 可帮助缩小量测和对准问题；样品放行仍应回到可追溯的工艺、原始数据和验收规则。Based on NIST public projects. Physics models and AI can narrow metrology and alignment questions; sample release still requires traceable process records, raw data and acceptance rules.
叠对量测：先写清模型可用范围与不确定度Overlay metrology: state model scope and uncertainty first
NIST 于 2026 年更新的项目，聚焦以物理支撑的 AI-SEM 用于叠对、关键尺寸与缺陷检查，并明确研究目标包含性能边界、速度和准确度。对试制项目而言，算法输出可作为排查线索；应同时保存图像采集条件、标尺/校准、模型版本、置信度、复核位置及人工判读规则，不能把“模型判断正常”直接写成样品合格。NIST's 2026 project develops physics-backed AI-SEM for overlay, critical-dimension and defect inspection, including performance limits, speed and accuracy. Treat algorithm output as a troubleshooting lead and retain acquisition conditions, calibration, model revision, confidence, review sites and human-review rules; a normal model result is not sample acceptance.
原子尺度写入：每一段图形都需要回读与对齐证据Atom-scale writing: each region needs readback and alignment evidence
NIST 的原子尺度器件项目说明，将大版图切分为可扫描区域，并在每段写入后成像、分析其光刻质量，再据特征进行下一段对齐。这个思路也适用于常规微纳试制：版图版本、写入/曝光参数、对准标记状态、局部 SEM 或光学复检和后续工艺变更，应按同一批次链路保存，避免把跨版本的图像或参数混作同一次验收依据。NIST's atom-scale-device workflow slices layouts into scan regions, images and analyzes lithographic quality after each region, then aligns the next region from detected features. For routine prototyping, retain layout revision, write/exposure settings, alignment-mark state, local SEM/optical review and downstream process changes in one lot traceability chain.
一个工艺观察Process Note
把“辅助判断—复核量测—放行结论”拆成三个字段Use three fields: assisted judgement, review metrology, release decision
素微编辑观察：在样品记录中单列算法或仿真建议、原始/复核量测结果和最终放行人/规则。三者可以相互支持，却不应互相替代；当版图、材料、涂胶、写入或量测任一版本变更时，应重新说明旧结论是否仍适用。素微's editorial observation: record algorithm/simulation advice, raw or review metrology, and the final release owner/rule in separate fields. They may support each other but are not substitutes; any layout, material, coating, writing or metrology revision requires an explicit applicability check.
项目准备提醒Project Prep
提交量测或图形验证需求时，先给出可复核的判定表For metrology or pattern verification, provide a reviewable decision table
请提供样品与批次编号、材料/层结构、版图和掩模版本、关键结构与量测位置、目标值和容差、设备/成像条件、抽样规则、原始数据格式、复核方法，以及异常时的处置或放行阈值。Provide sample and lot IDs, material/stack, layout and mask revisions, critical features and sites, targets and tolerances, tool/imaging conditions, sampling rule, raw-data format, review method, and exception or release thresholds.
来源Sources