Mask and Optical Models, Evolution of Lithography Process Models, Part IV

[Daniel Payne]

John Sturtevant explains the difference between Mask and Optical models. Very important at the 28nm node and below. <s class="hash" style="margin-top: 0px; margin-right: 0px; margin-bottom: 0px; margin-left: 0px; padding-top: 0px; padding-right: 0px; padding-bottom: 0px; padding-left: 0px; opacity: 0.7; text-decoration: none; display: inline-block; ">#</s>S…http://lnkd.in/Aagj_6

<script src="http://platform.linkedin.com/in.js" type="text/javascript"></script>
<script type="IN/Share" data-counter="right"></script>

 

[LinkedIn]

Quan Gan<q>

1.mask modeling: from practical experience, I don't think current Mask3D model has any advantage against the binary. I even met a case that A company use their own M3D modeling to fight against B company's T&P searching method, but finally failed to find any advantage, and brought extra massive computing load, as the mask shall need decompose the pupil by 4x. So, I don't think M3D shall be accepted in really production, need to be adopted by the lab </q>

2.Optical kernel: although, Hopkins equation and the decomposition of TCC told me that more kernels, more accuracy. But in so many cases, even in Foundries' productionlized OPC only use ~18 kernels. I scratch my head many times, I got my own coclution: fit method hide the difference into these fitted freedom dimensions.. So my question is need we use so many kernel(100) ? Just tell client our tool works so hard to satisfy them?