【24h】

Yet One more Dwell time Algorithm

机译:还有一个停留时间算法

获取原文

摘要

The current demand of even more powerful and efficient microprocessors, for e.g. deep learning, has led to an ongoing trend of reducing the feature size of the integrated circuits. These processors are patterned with EUV-lithography which enables 7 nm chips [1]. To produce mirrors which satisfy the needed requirements is a challenging task. Not only increasing requirements on the imaging properties, but also new lens shapes, such as aspheres or lenses with free-form surfaces, require innovative production processes. However, these lenses need new deterministic sub-aperture polishing methods that have been established in the past few years. These polishing methods are characterized, by an empirically determined TIF and local stock removal. Such a deterministic polishing method is ion-beam-figuring (IBF). The beam profile of an ion beam is adjusted to a nearly ideal Gaussian shape by various parameters. With the known removal function, a dwell time profile can be generated for each measured error profile. Such a profile is always generated pixel-accurately to the predetermined error profile, with the aim always of minimizing the existing surface structures up to the cut-off frequency of the tool used [2]. The processing success of a correction-polishing run depends decisively on the accuracy of the previously computed dwell-time profile. So the used algorithm to calculate the dwell time has to accurately reflect the reality. But furthermore the machine operator should have no influence on the dwell-time calculation. Conclusively there mustn't be any parameters which have an influence on the calculation result. And lastly it should take a minimum of machining time to get a minimum of remaining error structures. Unfortunately current dwell time algorithm calculations are divergent, user-dependent, tending to create high processing times and need several parameters to bet set. This paper describes an, realistic, convergent and user independent dwell time algorithm. The typical processing times are reduced to about 80 % up to 50 % compared to conventional algorithms (Lucy-Richardson, Van-Cittert ...) as used in established machines. To verify its effectiveness a plane surface was machined on an IBF.
机译:当前对甚至更强大和高效的微处理器的需求,例如深度学习已导致减小集成电路的特征尺寸的持续趋势。这些处理器采用EUV光刻图案化,可实现7 nm芯片[1]。生产满足所需要求的镜子是一项艰巨的任务。不仅对成像特性的要求不断提高,而且新的镜片形状(例如非球面镜片或具有自由曲面的镜片)都需要创新的生产工艺。然而,这些透镜需要在过去几年中建立的新的确定性子孔径抛光方法。这些抛光方法的特征在于根据经验确定的TIF和局部去毛刺。这种确定性的抛光方法是离子束成形(IBF)。通过各种参数将离子束的束轮廓调整为接近理想的高斯形状。利用已知的清除功能,可以为每个测量到的误差曲线生成停留时间曲线。这样的轮廓总是精确地以像素的形式生成到预定的误差轮廓,其目的始终是将现有的表面结构最小化,直至达到所用工具的截止频率[2]。校正抛光运行的处理成功与否取决于先前计算的停留时间曲线的准确性。因此,用于计算停留时间的算法必须准确反映实际情况。但是此外,机器操作员不应对停留时间的计算产生影响。总之,一定不能有任何对计算结果有影响的参数。最后,应该花最少的加工时间来使剩余的误差结构最少。不幸的是,当前的停留时间算法计算是分散的,取决于用户的,往往会产生高处理时间,并且需要多个参数进行投注设置。本文介绍了一种现实的,收敛的和用户独立的驻留时间算法。与已建立的机器中使用的常规算法(Lucy-Richardson,Van-Cittert ...)相比,典型的处理时间减少到大约80%到50%。为了验证其有效性,在IBF上加工了一个平面。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号